ML20212K702
ML20212K702 | |
Person / Time | |
---|---|
Site: | Seabrook |
Issue date: | 01/23/1987 |
From: | Thompson G MASSACHUSETTS, COMMONWEALTH OF |
To: | |
Shared Package | |
ML20212K635 | List: |
References | |
OL, NUDOCS 8701290158 | |
Download: ML20212K702 (316) | |
Text
.
, Affidwit of Gordon Thompson 23 January 1987 pap i Affidavit of Gordon R Thompson PhD I, Gordon Thompson, hereby depose and say:
My qualifications are set forth in an attached resume (Attachment 1). That resume indicates that I have experience in assessing th~e potential for accidental releases of radioactive material from nuclear power facilities, and in assessing the potential role of emergency planning in mitigating the public health effects arising from such releases.
This affidavit addresses a request by New Hampshire Yankee Division, Public Service Company of New Hampshire (hereaf ter, the " licensee"), that the emergency planning zone (EPZ) at the Seabrook nuclear power plant be reduced. Specifically, this affidavit responds to a schedule set by the relevant NRC Atomic Safety and Licensing Board (hereaf ter, the " Board"),
whereby intervenors must outline their arguments against the request by 27 January 1987.
The arguments outlined here can be summarized as follows:
(1) According to the emergency planning criteria set out in NUREG-0654, the licensee has not established, on the basis of its own technical analyses, a case for reducing the EPZ; (11) The licensee's technical analyses have not correctly accounted for certain reactor accident phenomena which can lead to large, early releases of radioactivity; (iii) Considerable uncertainty surrounds these phenomena, which are the .
subject of ongoing research; (iv) The Seabrook plant is of fairly typical design, and is therefore at risk l from these phenomena in the same manner as other nuclear plants;
~
(v) Additional phenomena and potential accident sequences may lead to substantial delayed releases for which public protection would be degraded if the EPZ were reduced; (vi) Because of scientific uncertainty surrounding the above-mentioned phenomena, and uncertainty regarding the strength of reinforced concrete containments, any consideration of the licensee's request is premature, at least until publication of the forthcoming NUREG-1150 in its final form; (vil) If the Board considers the licensee's request at this time, it should allow at least 6 months for intervenor expert witnesses to prepare a rebuttal case, this time being necessary to prepare independent analyses 8701290158 870126
{DR ADOCK 05000443 PDR
4 Affidwit of Gordm Thompson 23 January 1987
, page 2 of the nature and probability of severe accident outcomes;
. (vill) Aside from the above-mentioned phenomena with their accompanying scientific uncertainty, there are various considerations arguing against the conclusions of the licensee's technical analyses, including the effects i- of human error and malice; and (1x) The Board will fall in its duty if it does not allow adequate time for preparation of independent analyses by intervenor expert witnesses, since the NRC Staff's review of the licensee's technical analyses has been
- inadequate in terms of process and scientific objectivity.
The above-summarized arguments are treated at greater length in the
- following narrative. This discussion is supported by two further attachments. One of these (Attachment 2) is testimony written by myself and presented by the Commonwealth of Massachusetts to the Advisory Committee on Reactor Safeguards, the other (Attachment 3) is a report co-authored by myself and Steven Sholly.
It must be emphasized that arguments are presented here only in an outline form. Nor should the absence of any argument from this discussion 1- be regarded as limiting the scope of issues to be considered by the Board.
The schedule set by the Board has simply not allowed for preparation of an adequate rebuttal to the licensee's technical analyses.
l The Licensee's Case According to its Own Analyses
, in Attachment 2, pages 3 to 4, I have argued that the licensee's own L
technical analyses have not established a case for EPZ reduction according to the criteria by which NUREG-0654 has justified a 10-mile plume exposure EpZ. I make the argument here by reference to Attachment 2, and offer two supporting arguments.
l First, a draf t review by Brookhaven National Laboratory of the licensee's l
analyses (available to me in a version dated 5 December 1986; hereaf ter designated the "Brookhaven review") has offered modifications of licensee curves such as are presented in Exhibits 2 to 5 of Attachment 2. These modifications (see, for example, Figure S.7 of the Brookhaven review) are in the general direction of a higher frequency for exceedance of a given ,
dose at a given distance. I do not accept the findings of the Brookhaven review as an accurate assessment of the risk associated with the Seabrook plant, but note that the Board may choose to do so. If e
. , - . - ....m. ,.,--.h.3, , _,_ --..g.- ,. .,_ ,. , , ,--#,e 7 , m ,y-. - , , . - - - - . . , ,. . - . - . .- . . ._ ._ _ _ _ _ _ _ _
,O Affidsvit of Gordon Thompson 23 January 1987 Pop 3 l
Brookhaven-modified curves were used, my argument as presented in pages 3 to 4 of Attachment 2 would have even greater validity.
Second, NUREG-0654 states, at page 6, that: The overallobjective of emergency response pins is to provide dose savings (ad in some cases immediate life saving) for a spectrum of xcidents that couldproduce offsite doses in excess of protective Action Guides (pAGs) No single specific xcident sequence shouldbe isolatedas the one for which top /m because exh xcident couldhave different consequences, both in nature anddegree. Ftrther, the rage ofpossible selection foraplanningbasis is very large, starting with a zeropoint ofrequiringnoplanning at all because slynificant offsite radiological accident consequences are unlikely to occur, to planning for the worstpossible accident, regardless ofits extremelylowlikelihood The licensee does not claim that accidents leading to doses well above PAG levels at a 10-mile distance are impossible. Rather, its case for EPZ reduction rests on the purported low probability of such accidents. Such a position is not consistent with one of the potential selections for a planning basis as identified in NUREG-06S4, namely Dianning for the worst cessible accident. reaardless of its extremelv low likelihood. As I have argued in Attachment 2, page 11, this is an appropriate planning basis for nuclear reactor accidents.
Phenomena Potentially Leadina to Larae. Early Releases The phenomena of particular relevance to Seabrook are: direct heating of the containment atmosphere by ejected core material; accident-induced steam generator tube rupture (SGTR); and steam explosions. Each of these phenomena can lead to a breach or bypass of containment early in an accident sequence, with a consequent large release of radioactivity.
Attachment 2 provides a brief discussion of these phenomena, and a more complete, although earlier, alscussion can be found in Attachment 3, Chapter 9. The state of knowledge in this area is in flux, and any review of the field will soon become dated.
Both direct heating and induced SGTR require the reactor vessel to be at high pressure. For Seabrook, the Ilcensee's Probabilistic Safety Assessment (PSA) suggests (see Tables 2.3-1 and 5.1-3) that 97% of the l
1
' Affidwit of Gordm Thompson 23 January 1987 page4 estimated core melt frequency can be attributed to high pressure sequences. This indicates that direct heating and SGTR need to be carefully considered for Seabrook.
During the forthcoming proceeding, the licensee may argue that the vessel will become depressurized during most sequences, due,elther to a thermally-induced breach in the primary envelope or to deliberate operator actions. Such a claim would need to be carefully scrutinised, but it is instructive to consider the implications of core melt frequency being dominated by low pressure sequences, particularly with regard to the likelihood of steam explosion. Members of the NRC's Steam Explosion Review Group (SERG) have prepared subjective estimates for the probability of containment failure following a steam explosion at low pressure. Several of these probability estimates have a range up to 0.01 per core melt, a number wnich would undermine the licensee's case for an EPZ reduction. It should also be noted that the SERG probability estimates were subjective because research has failed to produce an objective basis for making such estimates. When matters of public safety are a stake, it is appropriate to use subjective estimates with caution.
Scientific Uncertaintv and Onacina Research Direct heating, induced SGTR and steam explosion phenomena, together with related severe accident phenomena, have been the subject of intensive research over recent years. Yet, there remain considerable uncertainties about the behavior which might be exhibited during an actual reactor accident. These uncertainties were evident at a meeting of expert review groups which was organized by the NRC in Bethesda, MD, earlier this month, and are discussed in an earlier policy paper to the NRC commissioners (SECY-86-369) which also sets out the role of these review groups.
l
! SECY-86-369 describes program schedules and budgetary requirements for relevant NRC research activities through FY 1989. It follows that the l resolution of outstanding uncertainties is not likely to occur on a shorter timescale.
! It may be helpful to illustrate the uncertaintties and their implications by two examples First, the licensee dismisses the problem of direct heating on the basis of some small-scale experiments. However, there are grounds l
l
' Affidsvit of Gordon Thompson 23 January 1987 pegs 5 for believing that scale may be important to this phenomenon, with larger-scale events being more dangerous. A test program at 1/10th-scale is under way at Sandia National Laboratories, with a supporting experimental program at Brookhaven. From these experiments, it may be possible to develop an accurate analytical capability which is relevant to the phenomenon at full scale; no such capability currently exists. Second, in studying the potential for induced SGTR lt is necessary to understand the processes of natural circulation in the primary system. The NRC has established a research schedule which may continue until at least FY 1990, in an attempt to obtain such understanding. Even if complete I understanding can be obtained in this area, it will still be necessary to establish an analytic basis for assessing the pattern of deposition of fission products in the steam generator tubes, as well as the potential for one tube failure to propagate to other tubes.
i
, Tvolcality of the Seabrook Design The licensee claims that Seabrook has an unusually strong containment.
However, similar claims are not made for other aspects of the design.
Although there is a lightweight outer containment, the licensee's PSA does not claim any significant credit for this feature.
! It is therefore important to examine the strength of the containment.
First, it will be noted that the licensee has stated (see Attachment 2, page
- 6) that Seabrook's containment is "nearly identical" to that of Indian point i Unit 2. Second, a finite element cc:ns. analysis .. a . o .. c vad in the Brookhaven review indicates containment failure at the wall-basemat intersection, at a pressure slightly above 157 psig. A similar analysis previously conducted at Brookhaven for Indian Point Unit 3 indicated containment failure at the same location, at a pressure of 125 psig. The I comparability of the two Brookhaven analyses requires examination, and
- the validity of either estimate is open to question, as discussed later in this af fidavit. However, it is clear that the licensee has no basis for arguing that Seabrook is a highly atypical nuclear plant.
Since Seabrook is not an unusual plant, it will be subject to severe accident phenomena in a similar way to other plants. The present scientific uncertainties surrounding severe accident phenomena apply to all plants, including Seabrook.
I
i i
l i ' Affidsvit of 0erdm Thompem 23 January 1987 pop 6 ,
Delaved Releases and Emergency Planning Even if large, early releases could be ruled out, it would not follow that the EPZ could be reduced. Delayed releases could be suf ficiently large that public safety would be adversely affected if the EPZ were reduced, First,-
even if the warning time were long, emergency response measures may not -
' be executed effectively for all population groups if the pre-planning otherwise conducted within the EPZ were not to occur. Second, the warning time may be short. Plant personnel and NRC staff may believe, until shortly before a release occurs, that a core melt accident can be contained l'
and that there is no need to notify citizens. This would be more likely if the Board were to ratify the licensee's claims about the safety of Seabrook.
Substantial delayed releases might occur through three mechanisms. First, if containment failure were explosive, there is the potential for flashing of contaminated pools of water and entrainment of deposited material, as discussed in Attachment 3, Chapter 9. Second, slow releases curing
- core-concrete interaction might lead to a significant presence of radionuclides in the containment atmosphere at the time of failure (the Te-132 to 1-132 chain is of interest in this context). Both of these mechanisms are surrounded by considerable scientific uncertainty. Third, l there may be slowly-developing sequences in which both core melt and
- containment failure are delayed, so that the source term is large. As an
! lllustration, NUREG/CR-4624, Volume 3, estimates a large source term for the AG accident sequence at the Surry plant, a sequence in W i h.
containment failure precedes core melt.
premature Nature of this Proceeding in view of the above-mentioned uncertaintles, and the timescale of NRC research efforts which address those uncertaintles, it is premature to ,
make any finding about the risk implications for Seabrook of direct containment heating, induced SGTR, steam explosion, and late-release source terms.
t in addition, since containment strength would be such a salient issue in this proceeding, it should be noted that there are uncertainties in this area as well. Attachment 2, pages 6 to 8, and Attachment 3, Chapter 8, address
, these uncertainties. For reinforced concrete containments such as
-,,,m_--... , _ . - - - . _--_.-,-.-.-.,.._,--_-,.__.--,..,,_..-m_
i o Affidsvit of Gordon Thompson 23 January 1987 pop 7 Seabrook, some of these uncertainties may be resolved upon completion of the 1/6th-scale Sandia test program, and subsequent analysis of its results.
Given these unresolved questions, and the potential risk to public health arising from an inappropriate decision to reduce the EPZ, a proceeding .
would be premature. Any action on the licensee's request should await the development of an improved understanding of accident phenomena, and the ratification of that understanding by the wider scientific community. This requires a delay, at a minimum, until the forthcoming report NUREG-1 ISO has received public comment and been published in its final form. An example from my own experience illustrates this point. Since December 1984 I have corresponded with M Silberberg of the NRC, regarding induced SGTR. In a letter dated 16 January 1987, Silberberg suggests that my outstanding questions would most appropriately be addressed in the context of public comment on NUREG-1 ISO.
Need for Indeoendent Analyses if the proceeding Goes Forward if the Board elects to move forward with this proceeding in spite of the existence of major unresolved issues, it is vital that the Board allows a period of at least 6 months for intervenor expert witnesses to prepare independent analyses. These analyses, taking appropriate account of the state of scientific knowledge to date, could identify the extent to which a case can be made that a reduced Seabrook EPZ would satisfy the objectives l
i and .r 9 i out in NUREG-06S4.
A period of at least 6 months would be required because the analyses would need to encompass:
l (1) a review of the present state of understanding of relevant phenomena; (11) estimates of the nature and probability of representative severe accident release classes; and (111) estimates of public exposure to radiation for various EPZ delineations.
Additional Factors Affecting Accident Risk The preceding discussion has focussed on accident phenomena for which there is incomplete scientific understanding. Thero are, however, other l considerations which argue against an EPZ reduction. The uncertainties i
i;1 l
'* Affidsvit of Orde Thompson .
23 January 1987 pop 8 which accompany these considerations generally have to do with the behavior of complex engineered systems and their associated personnel, rather than with the state of scientific knowledge.
4.
Relevant considerations include:
l (1) the degree to which design, maintenance and operation activities are subject to human error;
- (11) the extent to which studies such as the licensee's PSA and its
- . subsequent EpZ analyses have identified and correctly characterized the relevant accident sequences; (111) the ef fects of plant aging (for example, corrosion of containment reinforcement, embrittlement of major reactor components); and (iv) the potential for, and implications of, acts of malice by plant personnel or others.
[
l ' Independent analyses, such as could be conducted by intervenor expert .
I witnesses, could address the implications of these considerations in the context of EPZ reduction. It should be noted that it may not be possible,
- even in principle, to provide quantified estimates of the risk arising from all of these considerations. Sabotage, for example, is not a random event.
Yet, experience with insider sabotage and tampering events (see, for example, NUREG/CR-4462) Indicates that they cannot be eliminated from i
analyses of risk.
4 Inadeauate NRC Review of Licensee Technical Analyses I
To date, the only significant external review of the licensee's analyses has
- l. been conducted by the NRC Staf f, assisted by Brookhaven personnel. This review has been inadequate, on several grounds.
First, although meetings with the licensee have nominally been open, proper notice has not been given, record-keeping has been poor, and it has ,
been difficult for outside observers such as myself to obtain
! documentation. Second, the process has failed to identify and keep prominent the major uncertainties. It has tended to focus on details, leading to relatively minor modifications to the licensee's results. Third, the process appears to have been insulated from ongoing investigations of severe accident uncertainties by other arms of the NRC. Fourth, the process has not exhibited scientific rigor; it has allowed subjective considerations to have an equal role with the results of analyses based on
C Affidwit of Gordon Thompson 23 January 1987 pop 9 l an empirical foundation. This lack of rigor is evident in the transcript of a meeting held at Bethesda on 14 January 1987 (apparently the first such meeting for which a transcript was taken). Discussion at that meeting was poorly organized and did not adequately discriminate between subjective
! judgment and scientifically defensible understanding.
In summary, the entire process creates the impression that the NRC Staf f is going through the motions of a review. Thus, if the Board does not allow adequate time for independent analyses by intervenor expert witnesses, the licensee's technical analyses will not have been subjected to rigorous review and rebuttal. In view of the importance of the EPZ issue in terms of public safety, the Board would be f alling in its duty if it allowed this to occur.
Signed under the pains and penalties of perjury, this 23rd day of January,1987:
C:mnunvi:0!'h c Mc=chusetts OL S . \ ,z-Crn'.y el uiS2:2:( Gordon R Thompson phD institute for Resource and Security Studies Ey['Y ^*3 5;7[ Nh[j 27 Ellsworth Avenue
, i., !_ b._.,'L Cambridge, MA 02139 Notary Fubli; /
$ldSA L LE.'0,1C, "t/ P.:'
J:
tn an :: n to r:: w: n. un Attachments: (1) Resume for Gordon Thompson, September 1986.
(2) Testimony to the Advisory Committee on Reactor Safeguards, US Nuclear Regulatory Commission, by Gordon Thompson,10 October 1986.
(3)"The Source Term Debate: A Report by the Union of Concerned Scientists", by Steven Sholly and Gordon Thompson, January 1986.
ATTACHMENT 1 TO AFFIDAVIT OF GORDON R. THOMPSON 1
Resume for Gordon Thompson September 1986 professional Exoertise Consulting scientist on energy, environment, and international security issues.
Education
- PhD in Applied Mathematics, Oxford University,1973.
- BE in Mechanical Engineering, University of New South Wales, Sydney, Australia,1967.
- BS in Mathematics and Physics, University of New South Wales,1966.
Current Accointments
- Executive Director, Institute for Resource & Security Studies ( IRSS ),
Cambridge, MA.
- Coordinator, Proliferation Reform Project ( an IRSS project ).
- Treasurer, Center for Atomic Radiation Studies, Acton, MA.
- Member, Board of Directors, Political Ecology Research Group, Oxford, UK.
- Member, Advisory Board, Gruppe Okologie, Hannover, FRG.
Consulting Excertence ( selected )
- Lakes Environmental Association, Bridgton, ME,1986 : analysis of federal regulations for disposal of radioactive waste.
- Greenpeace, Hamburg, FRG,1986 : participation in an international study on the hazards of nuclear power plants.
- Three Mile Island Public Health Fund, Philadelphia, PA,1983-present :
studies related to the Three Mile Island nuclear plant.
- Attorney General, Commonwealth of Massachusetts, Boston, MA,1984-present : analyses of the safety of the Seabrook nuclear plant.
- Union of Concerned Scientists, Cambridge, MA, 1980-1985 : studies on energy demand and supply, nuclear arms control, and the safety of nuciear installations.
- Conservation Law Foundation of New England, Boston, MA,1985 :
preparation of testimony on cogeneration potential at the Maine facilities of
,- 2 Great Northern Paper Company.
- Town & Country Planning Association, London, UK, 1982-1984 : coordination and conduct of a study on safety and radioactive waste implications of the proposed Sizewell nuclear plant.
- US Environmental Protection Agency, Washington, DC, 1980-1981 assessment of the cleanup of Three Mile Island Unit 2 nuclear plant.
- Center for Energy & Environmental Studies, Princeton University, Princeton, RJ,1979-1980 : studies on the potentials of various renewable energy sources.
- Government of Lower Saxony, Hannover, FRG, 1978-1979 : coordination and conduct of studies on safety aspects of the proposed Gorleben nuclear fuel center.
Other Exoerience ( selected )
- Co-leadership ( with Paul Walker ) of a study group on nuclear weapons proliferation, Institute of Politics, Harvard University,1981.
- Foundation ( with others ) of an ecological political movement in Oxford, UK, which contested the 1979 Parliamentary election.
- Conduct of cross-examination and presentation of evidence, on behalf of the Political Ecology Research Group, at the 1977 Public Inquiry into proposed expansion of the reprocessing plant at Windscale, UK.
- Conduct of research on plasma theory ( while a PhD candidate ), as an associate staf f member, Culham Laboratory, UK Atomic Energy Authortty, 1969-1973.
- Service as a design engineer on coal plants, New South Wales Electricity Commission, Sydney, Australia,1968.
Publications ( selected )
t
- Nuclear-Weacon-Free Zones : A Survey of Treaties and Procosals ( edited with David Pitt ), Croom Helm Ltd, Beckenham, UK, forthcoming.
- International Nuclear Reactor Hazard Study ( written with fif teen other authors ), September 1986, Greenpeace, Hamburg, FRG ( 2 volumes ).
- * "What happened at Reactor Four" ( the Chernobyl reactor accident ), Bulletin of the Atomic Scientists. August / September 1986, pp 26-31.
- The Source Term Debate : A Reoort by the Union of Concerned Scientists
( written with Steven Sholly ), January 1986, Union of Concerned Scientists, Cambridge, MA.
- Checks on the spread" ( a review of three books on nuclear proliferation ),
Nature.14 November 1985, pp 127-128.
0- t 3
- Editing of Persoectives on Proliferation. Volume I, August 1985, published by the Proliferation Reform Project, Institute for Rescurce and Security Studies, Cambridge, MA.
- "A Turning Point for the NPT ?", ADIU Reoort. Nov/Dec 1984, pp 1-4, University of Sussex, Brighton, UK
- " Energy Economics", in J Dennis (ed), The Nuclear Almanac. Addison-Wesley, Reading, MA,1984.
- "The Genesis of Nuclear Power", in J Tirman (ed), The Militarization of High Technoloov. Ballinger, Cambridge, MA,1984.
- A Second Chance New Hamoshire's Electricity Future as a Model for the Nation ( written with Linzee Weld ), Union of Concerned Scientists, Cambridge, MA,1983.
- Safety and Waste Manacement Imolications of the Sizewell PWR ( prepared with the help of 6 consultants ), a report to the Town & Country Planning Association, London, UK,1983.
- Utility-Scale Electrical Storace in the USA The Prosoects of Pumoed Hydro.
Comoressed Air. and Batteries. Princeton University report PU/ CEES *120, 1981.
- The Prosoects for Wind and Wave Power in North America. Princeton University report PU/ CEES
- 117,1981.
- Hydroelectric Power in the USA Evolvina to Meet New Needs. Princeton University report PU/ CEES
- 1IS,1981.
- Editing and part authorship of " Potential Accidents & Their Effects", Chapter lit of Reoort of the Gorieben International Peview. oublished in German by the Government of Lower Saxony, FRG,1979 -- Chapter ill available in English from the Political Ecology Research Group, Oxford, UK.
- A Study of the Consecuences to the Public of a Severe Accident at a Commercial FBR located at Kalkar. West Germany. Political Ecology Research Group report RR-1,1978.
' Excert Testimony ( selected )
- International Physicians for the Prevention of Nuclear War,6th Annual Congress, Koln, FRG,1986 : Relationships between nuclear power and the threat of nuclear war.
- Maine Land Use Regulation Commission,1985 : Cogeneration potential at facilities of Great Northern Paper Company.
- Interf alth Hearings on Nuclear Issues, Toronto, Ontario,1984 : Options for Canada's nuclear trade and Canada's involvement in nuclear arms control.
- Sizewell Public Inquiry, UK,1984 : Safety and radioactive waste implications of the proposed Sizewell nuclear plant.
i
o 4
-
- New Hampshire Public Utilities Commission,1983 : Electricity demand and supply options for New Hampshire.
- Atomic Safety & Licensing Board, Dockets 50-247-SP & 50-286-SP, US Nuclear Regulatory Commission,1983 : Use of filtered venting at the Indian Point nuclear plants.
- US National Advisory Committee on 0ceans and Atmosphere,1982 :
Implications of ocean disposal of radioactive waste. -
- Environmental & Energy Study Conference, US Congress,1982 : Implications of radioactive waste management.
Miscellaneous
- Australian citizen.
- Married, one child.
- Resident of USA,1979 to present; of UK, 1969-1979.
- Extensive experience of public speaking before professional and lay audiences.
- Author of numerous newspaper, newsletter, and magazine articles and book reviews.
- Has received many interviews from print and electronic media.
n* n******n***
i
l ATTACHMENT 2 TO AFFIDAVIT OF GORDON R. THOMPSON
'e Some Comments on Recent Studies Sponsored by Public Service Company of New Hampshire, Regarding Emergency Planning at the Seabrook Nuclear Plant Testimony to the Advisory Committee on Reactor Safeguards, US Nuclear Regulatory Commission 10 0ctober 1986 by Gordon R Thompson on behalf of The Attorney General, Commonwealth of Massachusetts INSTITUTE FOR RESOURCE AND SECURITY STUDIES 27 Ellsworth Avenue, Cambridge, MA 02139
i
- Testimony to ACRS 10 0ctober 1986 pap 2 I Introduction
-Ttmrtestimony of fers some comments on two studies recently prepared for Public Service Company of New Hampshire (PSNH) by the consultants Plckard, Lowe and Garrick, Inc (PLG,1985; PLG,1986). These studies were the subject of a recent meeting of two ACRS subcommittees (ACRS,1986).
The studies purport to demonstrate that emergency planning regulations
"~
cinTi rilaxed in the case of the Seabrook nuclear plant.
The comments offered here are not exhaustive. Rather, their intent is to assist the ACRS in its understanding of the issues raised by PSNH. It currently appears that the NRC Staff review of the PSNH reports will continue at least through November 1986, and that the ACRS will give further consideration to this matter at both subcommittee and full
, committee levels before making a formal report to the NRC Chairman.
During this process, the issues will become more clearly defined, and the author of this testimony expects to make further comments as appropriate.
The principal thrust of the PSNH reports is to argue that, for the Seabrook plant, the emergency planning zone (EPZ) for plume exposure can be
- reduced from the present 10-mile radius to some smaller size -- perhaps a 2-mile radius. In support of this argument, an analysis has been conducted which is similar to that presented in NUREG-0396, the report of an NRC/ EPA Task Force which recommended a generic 10-mile radius ,
plume exposure EPZ (NRC/ EPA,1978).
In adopting this approach, PSNH raises three key questions. First, do the results of the analysis as presented by PSNH justify a reduction in the EPZ by analogy with the procedure adopted in NUREG-0396 ? Second, is the PSNH analysis credible ? Third, should such probabilistic analysis be used as a basis for emergency planning regulations ?
l The present testimony provides brief responses to the first and third of these questions, but its primary focus is on the second question.
Section 2 of this testimony compares the PSNH analysis with the findings presented in NUREG-0396. The comparison shows that the case for a i
reduced plume exposure EPZ is questionable, even on the terms chosen by
- - Testimony to ACRS 10 0ctober 1986 page 3 PSNH. Then,-Section 3 addresses the credibility of the PSNH analysis, and identifies various deficiencies in that analysis. The state of knowledge regarding various severe accident phenomena is briefly addressed in this -
discussion. Finally, Section 4 addresses the suitability of probabilistic analysis as a basis for emergency planning regulations.
- 2. Comoarison of the PSNH Aeoorts with NUPEG-0396-In NUREG-0396, the NRC/ EPA Task Force proposed EPZs of 10-mile radius for plume exposure and 50-mile radius for ingestion, and these zones were subsequently incorporated into the emergency planning regulations. The Task Force's choice of plume exposure EPZ has been subsequently characterized (NRC/ FEMA,1980) as being based primarily on the following considerations:
(a) Projected doses from the traditional design basis accidents would not exceed Protective Action Guide (PAG) levels outside the zone; (b) Projected doses from most core melt sequences would not exceed PAG levels outside the zone; (c) For the worst core melt sequences, immediate life threatening doses would generally not occur outside the zone; and (d) Detailed planning within 10 miles would provide a substantial base for expansion of response efforts in the event that this proved necessary.
In support of this choice, the Task Force presented dose estimates derived from the Reactor Safety Study (WASH-1400), and an example of those estimates is shown in Exhibit 1. This figure indicates that, given a core i melt accident, estimated whole body dose at 10 miles would exceed both lower (I rem) and upper (5 rem). PAG levels with a conditional probability l of more than 20%. Also, the dose at 10 miles is estimated to exceed 200 l rem (commonly regarded as a ' life threatening dose") with a conditional probability of about 3%.
It is questionable whether this figure is consistent with choice of a 10-mile radius EPZ on the basis of statements (b) and (c) above. Various
Testimony to ACRS 100ctober 1986 pay 4 interpretations of the phrases "most core melt sequences" and " generally not occur" are possible.
In any event, PSNH has chosen to base its case for a reduced plume exposure EPZ on a similar analysis. Some of the results are shown in Exhibits 2,3 and 4. It is certainly true that these results (specifically, Exhibit 4) suggest a very low conditional probability that the dose at 10
~ miles will exceed 200 rem. However, this is not true for the PAG dose levels.
If mean values are taken (it is arguable whether.mean or median values are more appropriate), then Exhibit 2 indicates that the dose at 10 miles-would exceed I rem with a conditional probability of 10% and would exceed 5 rem with a conditional probability of about 2%. If, as in Exhibit 3, releases with more than 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br /> of warning time are excluded, although the rationale for this is itself questionable, the dose at 10 miles still exceeds I rem with a conditional probability of more than 2%.
An accompanying sensittvity study (pLG,1986) is alleged to support the pSNH case, even though it employs purportedly more " conservative" source term assumptions. Fewer results are presented in this report, but Exhibit S provides some comparison. This figure shows curves equivalent to the median values in Exhibit 3, the curves in Exhibit 5 labeled 'This Study
- referring to the sensitivity study, and the curves labeled 'RMEpS Results*
being identical to the median Seabrook results presented in Exhibit 3. It is clear that the mean values for the sensitivity study, particularly tf releases with a warning time above 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br /> were not excluded, would show a significant conditional probability of PAG dose levels being exceeded at 10 miles.
l Thus, even if the supporting analysis were credible, PSNH could not be said to have demonstrated that projected doses from most core melt sequences l
at Seabrock wtil not exceed PAG levels outside a reduced EPZ (that is, one of less than 10-mile radius). Insof ar as the PSNH case has merit on its i
own terms, that merit is only suf ficient to address one of the considerations (statement (c) above) which motivated the NRC/ EPA Task Force -- that is, for the worst core melt sequences, immediate lif e i threatening doses would generally not occur outside the croposed EPZ.
l - - - . -
Testimony to ACRS 10 0ctober 1986 page 5
- 3. Credibility of the pSNH Analysis 3.1 The General State of Knowledge Aegarding Severe Aeactor Accidents ACRS members will appreciate that severe reactor accidents are expected to be accompanied by phenomena which are difficult to model. It is not appropriate to pursue the various phenomenological and modeling uncertainties here, and the author of this testimony has written elsewhere on this matt,er (Sholly and Thompson,1986).
In passing, it is appropriate to note that uncertainties remain large despite substantial past investments in severe accident research. This point is illustrated by Exhibit 6, which shows source term estimates for a TMLB' sequence at the Surry plant. This table summarizes an uncertainty study conducted at Sandia National Laboratories, the purpose of which was l to investigate the source term implications of incomplete understanding of relevant phenomena.
Exhibit 6 also illustrates another general point, namely that early containment failure will tend to be associated with a larger source term.
Thus, the timing of containment failure is doubly important for emergency planning; first, in terms of warning time; second, in terms of the size of the release. It is therefore particularly important to determine if the PSNH reports adequately address the uncertainties surrcunding the mode and timing of containment failure.
The remaining discussion in this section is devoted to various issues pertaining to containment f ailure. This should not be taken as implying that the PSNH analysis is adequate in other respects.
3.2 PSNH Claims Regarding probability of Early Containment Failure Exntbit 7 provides a crude summary of the PSNH claims about containment f ailure. The column headed *SSPSA" contains results from the original Probabilistic Safety Assessment (PLG,1983), while the third column shows the latest results.
It can be seen that the conditional probability of early conta:nment failure
- _ _ . ~. . . _._-._ _ . _ _ _ _ .
Testimony to ACRS '
10 0ctober 1966 page 6 (given a core melt) is now alleged to be 1x10-3, compared with 3.4x10-I ,
for WASH-1400. This difference, through its direct effect (on the
-conditional probabil ity of re ease l )and its indirect effect (on source terms), must account for much of the difference between Seabrook and NUREG-0396 results which is apparent from Exhibits 2 through 5. The remaining difference is presumably attributable to factors such as:
weather; plume modeling; release categories; and source term assumptions.
3.3 Failure of the Containment Structure it is of ten claimed that the Seabrook plant has an unusually strong containment structure. However, it is not entirely clear why this is 50. In a letter to the NRC, a PSNH official has stated that "Seabrook's containment configuration, design and construction are nearly identical to the prototype plant", the prototype in this Instance being Indian Point Unit 2 (DeVincentis,1986). Apparently this statement is not quite correct, because it has been said that the Seabrook containment contains more steel, particularly in the form of diagonal seismic reinforcement (ACRS, 1986, pp 38-40).
In any event, PSNH has not provided a convincing analysis of contairment strength. The basic structural analysis is contained in Appendix H.1 of the l
Seabrook Probabilistic Safety Assessment (PLG,1983) and is quite i sketchy. The analysis, conducted by Structural tiechanics Associates (Newport Beach, CA), yielded results as summarized in Exhibit 8.
Structural f ailure is predicted to occur by failure of hoop reinforcing bars in the cylindrical wall, at a pressure of 216 psig.
The similarity of the Seabrook and Indian Point Unit 2 containments allows some comparison of different analyses of containment strength.
Several analyses have been made for the indlan Point Unit 3 containment, which is said to be similar to the Indian Point Unit 2 containment except l for the placement of seismic reinforcement (Butler and Fugelso,1982).
For Indian Point Unit 3, three separate analyses are of interest. The first j was conducted at MIT, using a relatively crude model (Fardis et al,1982).
- if all reinforcing bars were assumed to have the same properties, and l
splice failures were neglected, the containment was found not to f all at i
t
Testimony to ACRS ,
I 10 0ctober 1986 page 7 l 170 psig, despite radial wall displacements exceeding 20 inches. Although the analysis was terminated at 170 psig, it was estimated that failure would occur at 200 psig, at about mid-height of the containment wall.
A more sophisticated model was used by analysts at Los Alamos National Laboratory (Butler and Fugelso,1982). This indicated that f ailure would occur at 118 psig, at the junction of the cylindrical wall and the basemat.
A group at Brookhaven National Laboratory, also using a more sophisticated model, likewise found that failure would occur at this junction, although their estimate of failure pressure was 125 psig (Sharma et al,1984). It is noteworthy that the Brookhaven model found all reinforcing steel at cylinder mid-height to be still in its elastic range at 125 psig. This is consistent with the finding by Structural Mechanics Associates that yield of hoop reinforcing bars will occur at Seabrook for a pressure of 157 psig.
Thus one must suspect that a re-analysis of the Seabrook containment, using the relatively more sophisticated models employed at Los Alamos and Brookhaven, would lead to a substantial reduction in the estimated f ailure pressure. Moreover, even these models should be used with caution.
There are fundamental difficulties in analyilng the failure of reinforced concrete structures (Sharma et al,1983). Some of these dif ficulties may be resolved by testing of the I/6th-scale concrete containment which is being constructed at Sandla National Laboratories, but testing is not expected to begin until early 1987 (Horschel and Jung,1986).
Axisymmetric models may not be adequate to represent a real containment. For example, the basemat at Seabrook is surrounded over 73%
l of its circumference by the annular foottng for the enclosure butiding (PLG,1983, Appendix H.1). If the cylinder-basemat junction is already a sensitive area, the discontinuities in this annular footing might be significant in terms of stress concentrations.
It is also important to investigate the ef fects of anticipated variability in
- the properties of reinforcing bars and splices. This has been done approximately in the above-mentioned MIT study of the indlan Point Unit 3 l containment, with some Interesting results. Exhibit 9 t!!ustrates the approach, showing the ef fective stress-strain relation for an 8-bar hoop as compared with that for a single bar with average properties. When this
Testimony to ACRS 10 October 1986 pegs 8 approach was used to analyse containment behavior, the estimated failure pressure fell from 200 psig to 140 psig, f ailure again occurring at cylinder mid-height. Such a reduction might not be manifested for failure at the wall-basemat junction, but is certainly relevant to the predictions of Structural Mechanics Associates as to the hoop bar failure mode for Seabrook. 0f course, the use of substandard materials or construction procedures would lead to yet further reductions in f ailure pressure.
To summarize, pSNH has not accounted for various uncertainties surrounding the structural f ailure of the Seabrook containment, and there is reason to believe that they have substantially over-estimated its f ailure pressure.
3.4 Direct Heatina of the Containment Atmosohere Direct heating following high-pressure melt ejection is widely regarded as a key contributor to uncertainty in containment performance (Haskin, 1986). A recent small-scale experiment has shown very little retention of ejected material on structures at the exit from the reactor cavity, suggesting that a large mass of finely divided core material could be available for direct heating of the containment atmosphere.
The PSNH reports provide a brief discussion of direct heating (PLG,1985, pp 4-13 to 4-16), concluding that "At Seabrook, direct heating creates no significant challenge to containment integrity". This conclusion is, at the very least, premature.
3.5 S' team Exolosion ACAS members will appreciate that the probability of containment f ailure following an in-vessel steam explosion has been the subject of prolonged debate. Yet, the debate cannot be said to be resolved (Haskin,1986).
One subject which should be carefully pursued in this context is the possibility of " accelerated meltdown
- If water were to be suddenly supplied to a steam-starved core (through recovery of safety injection capability, or through discharge of accumulators following a gradual depressurization of the vessel), the resulting zirconium reaction might precipitate a relatively rapid core slump.
Testimony to ACRS 100ctober 1966 page 9 '
Alsd, the effect of in-vessel steam explosions should be considered in terms of the entire primary circuit envelope. The steam generator tubes, which might have been thermally weakened prior to the explosion, come to .
mind in this context.
Ex-vessel steam explosions also need to be carefully considered. For example, under some conditions a steam explosion in the reactor cavity might ampitfy the offect of a high-pressure melt ejection event. Molten core material could be further fragmented by this explosion, thus increasing its ability to directly heat the containment atmosphere.
3.6 Induced Steam Generator Tube Quoture ,
p5NH does not account for the possibillty that tube rupture could be induced during a core melt sequence, although this is the subject of continuing attention by the NRC.
One-dtmensional calculations (using RELAp5 and SCDAp) have been made for a station blackout sequence at the Seabrook plant (Bayless,1984; Bayless,1985). These do not Indicate significant heating of steam generator tubes, and the primary system loop seals (between the steam generators and the pump Inlets) do not clear. However, more detailed analysts is required. Further work is under way inside the NRC, but it seems that there will be no published account of this work until li appears ,/,
as a part of the forthcoming risk report, NUREG-1150.
It is important to note that pressures and temperatures in the primary system might undergo sharp fluctuations at times when steam generator tubes are potentially at risk. For example, if the system were to gradually depressurtze, then accumulator discharge (perhaps a cyclic discharge) could induce'such fluctuations. Loop seals might be cleared in this manner, thus allowing increased heat transfer t.o the steam generator tubes.
A recent source term analysis indicates that induced tube rupture could lead to a substantial release of radioactive material. For the Sequoya plant, an induced failure equivalent in area to 5 steam generator tubes was assumed to occur at the time of core slumping (153 minutes), and the secondary side relief valves were assumed to remain open from that time
, e' i
- Testimony to ACRS 10 0ctober 1966 page 10 until vessel failure (169 minutes). A source term including 14% of lodines and cesiums was estimated (Denning et al,1986). Clearly, the source term would be larger if secondary side relief valves were to stick open. This could occur during repeated cycling of the valves during steam generator dryout, or because of loadings induced on the valves af ter tube failure.
3.7 Failure of Maior Reactor Comoonents The Seabrook reactor contains several major components (reactor vessel, pressurizer, steam generators, pump bowls, pump flywheels) whose ,
f ailure is not a design-basis event. Should such failure occur, the l
containment could be breached (by missile impact) and a core melt could be induced at the same time.
If one supposed, for the sake of illustration, that the conditional probability of core melt and containment failure, following a majcr I component f atture, is unity, then Exhibtt 7 (" Updated Results' Column) could be interpreted as stating that the probability of major component
- f allure is less than 2.7x10-7 per reactor-year. Perhaps this could be objectively demonstrated, but that would not be a trivial exercise.
l 3.8 Sabotace Sabotage is not amenable to a statistical analysis, but this does not mean
- that it should be neglected. By analogy with the immediately preceding discussion, can it be confidently assumed that the probability of a
- sabotage event involving core melt and early containment f atlure at i Seabrook is less than 2.7x10-7 perreactor year ?
3.9 Human Error It has been pointed out earlier in this testimony that the strength of the
- containment structure could be compromised by the use of substandard materials or construction procedures. Likewise, the overall integrity of
the containment envelope (including items such as isolation valves in i purge systems) could be compromised by any of a variety of human errors.
I j The analytic approach represented by Exhibit 7 assumes that errors leading to core melt are independent of errors leading to loss of
9 Testimony to ACRS 10 0ctober 1966 .
pageiI containment integrity. However, these errors might be correlated. This would be likely if there were a gradual degradation in operational standards during the plant's lifetime.
3.10 Summarv l PSNH has not demonstrated that the conditional probability of early containment f ailure, given a core melt, is very low. Accordingly, the analysis which they have presented by analogy with NUREG-0396 is not
! credible.
The preceding discussion, which is by no means exhaustive, has identified several respects in which the pSNH reports fall to adequately address some fundamental severe accident phenomena. Although the reports purport to consider uncertainties, they do not do so at the phenomenological level.
- 4. Suitability of probabilistic Analysis as a Basis for Emergencv Dlanning Because of the many uncertainties in severe accident assessment, it is not currently possible to demonstrate probabilistically that large,
! f ast-developing releases are highly unlikely at US reactors. This problem is compounded by'the importance of factors such as sabotage and long-term trends in human error, which are not susceptible to j
probabilistic analysis. The limitations of the pSNH analysis simply
! illustrate this general point.
Accordingly, one should look elsewhere for a planning basis for emergency response. The appropriate basis is to plan for those contingencies which
- j. are physically credible. In light of the Chernobyl accident, this suggests ,
that the focus of planning for plume exposure should be widened from a 10-mile radius, rather than narrowed to a smaller area. At the same time, there should be no pretense that emergency response can totally prevent adverse public health ef fects following a reactor accident.
l NUREG-0396 suf fers from the same limitation as the pSNH reports, in the l sense that it relies on probabilistic analysis. However, because the
! conditional probability of early containment failure was assumed to be
.1_.__. - ._. . _ _ _ . _ _ _ . _ . _ _ _ . _ _ _ _ _ _ _ _ _ . _
. Testimony to ACRS 10 0ctober 1986 page 12 large (see Exhibit 7), NUREG-0396 yielded a planning basis which could have been generated simply by assuming that large, f ast-developing releases are physically credible.
S. References (ACRS,1986)
Advisory Committee on Reactor Safeguards, Subcommittee on Severe Accidents and Subcommittee on Occupational and Environmental Protection Systems, transcript of meeting on 26 September 1986.
(Butler and Fugelso,1982)
T A Butler and L E Fugelso, Response of the Zion and Indian Point Containment Buildings to Severe Accident Pressures, NUREG/CR-2569, May 1982.
(Bayless,1984)
P D Bayless, Analysis of a Station Blackout Transient at the Seabrook Nuclear Power Plant, EGG-NTP-6700, EG&G Idaho, September 1984.
(Bayless,1985)
P D Bayless, Analysis of a Station Blackout Transient with .
Reactor Coolant Pump Seal Leakage for the Seabrook Nuclear Plant, Attachment I to letter of 6 June 1985 from T R Charlton (Manager, Reactor Simulation and Analysis Program, EG&G Idaho) to F L Sims (Director, Reactor Research and Technology Division, Idaho Operations
! Of fice, DOE).
(Denning et al,1986)
R S Denning and 8 other authors, Radionuclide Release Calculations for Selected Severe Accident Scenarios: PWR Ice Condenser Design, NUREG/CR-4624, Vol. 2, July 1986.
(DeVincentis,1986)
J DeVincentis (Director, Engineering and Licensing, Public Service of New Hampshire, New Hampshire Yankee Division), letter of 7 April 1986 to V S Noonan (Project Director, PWR Project Directorate No. 5, NRC).
Testimony to ACRS 10 0ctober 1986 page 13 (Fardis et al,1982)
M N Fardis, A Nacar and M A Delichatslos, Reinforced Concrete Containment Safety Under Hydrogen Explosion Loading, NUREG/CR-2898, September 1982.
(Haskin,1986)
F E Haskin, " Containment Failure Modes: Insights and Uncertainties",
Proceedings of the Topical Meeting on Reactor Physics and Safety, NUREG/Cp-0080, Vol 2, August 1986, pp 683-698.
(Herschel and Jung,1986)
D S Horschel and J Jung," Construction and Analysis of a 1/6th-Scale Concrete Containment Model , paper delivered at the Third NRC/Sandia Containment integrity Workshop, Washington, DC,21-23 May 1986.
(Lipinski et al,1985) .
' R J Lipinski and 10 other authors, Uncertainty in Radionuclide
>'elease Under Specific LWR Accident Conditions: Volume ll, 1MLB' Analyses, SAND 84-0410, Vol. 2, DRAFT, Sandia National Laboratories, February 1985.
(NRC/ EPA,1978)
US Nuclear Regulatory Commission /US Environmental protection Agency Task Force on Emergency Planning, Planning Basis for the
- Development of State and Local Government Radiological i
Emergency Response Plans in Support of Light Water Nuclear Power Plants, NUREG-0396, December 1978.
(NRC/ FEMA,1980) l US Nuclear Regulatory Commission and Federal Emergency Management Agency, Criteria for Preparation and Evaluation of Radiological Emergency Response Plans and Preparedness in Support of Nuclear Power Plants, NUREG-06S4, Rev. I, November 1980.
(PLG,1983)
Pickard, Lowe and Garrick, Inc, Seabrook Station Probabilistic l Safety Assessment, December 1983 (6 volumes).
l
Testimony to ACRS 10 0ctober 1986 page 14 (PLG,1985)
Ptckard, Lowe and Garrick,Inc, Seabrook Station Risk Management and Emergency Planning Study, December 1985.
(PLG,1986)
Pickard, Lowe and Garrick, Inc, Seabrook Station Emergency Planning Sensitivity Study, April 1986.
(Sharma ct al,1983)
S Sharma, M Reich and T Y Chang, Review of Current Analysis Methodology for Reinforced Concrete Structural Evaluations, NUREG/CR-3284, April 1983.
(Sharma et al,1984)
S Sharma, Y K Wang and M Reich,' Nonlinear Failure Analysls of a Reinforced Concrete Containment Under Internal Pressure', Proceedings of the Second Workshop on Containment integrity, NUREG/CP-0056, August 1984, pp 519-532.
(Sholly and Thompson,1986)
Steven Sholly and Gordon Thompson, The Source Term Debate, Union of Concerned Scientists, Cambridge, MA, January 1986.
Exhibit 1 Source: NRC/ EPA, 1978
( NUREG-03% )
1 ~A :
1 . . . . i . . i i i i.
. isisig i i i sing -
. i l -
C -
= -
i nsM 8
> N hI c.1 5 REM 7 -
wE -
~
oU ". -
34 .
eb 2w 52 -
50 m<
03 --
yG > o.01 7 -
.o - -
g .
g . 200 REM
' ' 'I ' ' ' ' ' I ' ' ' ' ' ' ' '
0.001 to IN 1000 1
DISTANCE (MILESI Figure I 11. Conditional Prebability of Exceeding Whole Gody Dose Versas Distance. Probabilities are Conditional on a Core Melt Accident (5 x 10-5),
j Whole body dose calculated includes: extcenal dose to the whole body due to the
' possing cloud, exposure to radionuclide< on ground, and the dose to the whole body from inhaled radionuclides.
Dose calculations assumed no protectiva actions taken, and straight line plume trajectory.
(
Exhibit 2 Source: PLG, 1985
. 6 6 . ..
1
. i 6 6 6 6 6el 6 6 4 .
s ii[ .
4%. ,
O=%
. O _
\ 1 REM
\ l 0.1 -
5 REM .
3 -
t- ~
5 5 ,
d -\ l E N _
=z \ l 3w - D I
\ o E$
gu . \ \
C4 o o d$ \ '\ i
/
\
h3 0.01 : o 0; F "M cc - t H
\ :
x8 -
1 W -
\ ~
Ez< O
~
!E
=0 -
\ l
~
\ I vi sw
<m .
o _.
l I 5c
=
5 REM
/la \ t i l
i i
E :.001 '_- NuREG - c:ss
\ MEOIAN .;
9
~ -
( \
C W
~
MEAN THIS S UCY FC9 _j
. 4 SEAB RCCK S*ATICN l
~
O__o MEDIAN -.
. . . . ....t : i e e . ...t . . . . , .. .
l 3.CC01 l 10 100 1.:CC 1
O! STANCE (MILES)
FIGURE 2-10. CCMPARISCN OF UFCATED SEAERC0K STATION RESULTS '4ITH NCRES-0 C-REM ANO l-REM WHOLE 30DY COSE PLOTS FOR NO i IMMEDIATE PROTECTIVE ACTICN j
242
Exhibit 3 Source: PL6, 1985 I - 6 i 6 6 6 66 .g i
- 6 . . ..aj i . . . . . .
1 REM
> 01 - --
c -
2 -
e a
ow - -
j5 Ds -
e9 Ed
- 'oN ~
c< g U2 -
O -
Mw w2
\
g
\
C -g g o -
>y$ 0.01 -
- Og \
u<
= -
g O -
wz
=w 0
~ ~
52
=c *
\ \
O 1 R EM
\
4 O \
au -
g :
So \ O I Ec -
\ \
\
- o O
\
5 8 0.001 -
\
O
\lC
\
<n
~
\ NU R EG - 0396 g MEDIAN -
o
- / e :
' ~
~ I SREM THIS S*JDY FOR SEA 8RCOK STATICN 0- -0 -
MEDIAN
<D
' ' ' I ' ' ' ' ' ' ' ' ' ' ' ' ' '
o.0001 1 10 100 1.000 DISTANCE (MILES)
- FIGURE 2-11. COMPARISON OF UPDATED SEABROOK STATION RESULTS WITH NUREG-0396 REM AND 1-REM WHOLE BODY COSE PLOTS FOR RELEASES WITHIN 24 HOURS OF WARNING AND NO IMMEDIATE PROTECTIVE ACTION
Exhibit 4 Source: PLG, 1985 1 . . i . . . . . . . . . .
- i i e ii6i.g ..ig 6
$ 0.1 -- 1
~
"i -
-i 9e 1
gm - -
e9 zu -
j EN ,
MRW
>d u
52 u.
C w* 0.01 -
-~
>8 -
E< -
wz -
1 c5 -
]
y5 -
j 200 R EM
, M
/
o\ 50 REM o0 -\ ~
c \
z O 0.001 - \ NUR EG - C396 -":
l C O MEDIAN a
\ i
~ \ - _'
O MEAN THIS STUOY COR _
SEABROOK STATICN
) ~
k 200 REM O- -O MEDIAN O MEDIANI . _.,
200 R EM CURVE QFF
/ l SCALE l 1 ,,,,,I . . , , ..,I . . ' ' '
n, , .
, 3 1 10 100 1.000 OlSTANCE (MILES)
FIGURE 2-9. COMPARISON OF UPDATED SEA 3 ROOK STATION RESULTS WITH NCRE3-;~96 200-REM AND 50-REM WHOLE BODY 00SE PLOTS FOR NO IMMEDIATE FROTECTIVE ACTICN l _ _
2.11
Exhibit 5 Source: PL6. 1986 l
1
. . . . ....i . .. . .... .
l
. NUREG-0386 -
"E - - ---- THIS STUDY FOR -
. SEASROOK STATION g3_ _ -
> .. ... . . . . ...* RM EPS R ESULTS FOR SEASROOK
, .J STATION wg 1 REM
$ o.1 - -
- 8 : :
E< -
55 _- s gi
~-
xy b
= .=- - - : .s%
Q
'. 3,L \
g uti \
\
325 s lk \
88 ' .. '.
- \ \
e; mu o.oi -. .,
. \ \
\
g
\ -
', \
g< _
~ .,~ . - \ -
gg _ 5.,
\1 REM .
5 REM (]
os -
. \
. . N v3 *
- 1R
. EM l \
l 5 F.E M ., g l
}
, , , , 'r. i i l l il i , , !,il - t . i t i,'
9 i 1o 100 1.000 OlSTANCE (MILES) i l FIGURE 2-4. COMPARISON OF SEAPROOK STATION RESULTS WITH NUREG-0396 -
l 1-REM AND 5-REM WHOLE BODY DOSE PLOTS FOR NO IMMEDIATE PROTECTIVE
- ACTION AND WASH-1400 SOURCE TERM METHODOLOGY l
2-11
Exhibit 6 Source: Lipinski et al, 1985
( adapted from Table 7-7 )
Estimated Release Fractions from Containment for a Surry TMLB' Sequence, Accounting for Phenomenological Uncertainties x
Containment Failure WASH-1400 Element Early Late PWR 2 Value I 0.006 to 0.8 0 to 0.004 0.7 Cs 0.004 to 0.8 0 to 0.004 0.5 Te 0.0Q2 to 0.8 0 to 0.007 0.3 Sr 3E-6 to 0.8 0 to 0.02 0.06 Ru 3E-7 to 0.2 0 to 1E-5 0.02 La 9E-9 to 0.06 0 to 7E-4 0.004
- Containment hole si::e assumed to be 0.001 m to 10 m .
i l
1 i
l Exhibit 7 Source: PLG, 1985 TABLE 2-4. ' COMPARISON OF CORE MELT FREQUENCIES AND DISTRIBUTIONS OF RELEASE TYPES WASH-1400 33p3; Upcated Risk Parameter PWR Results e Mean Core Melt Frequency (events 9.9-5* 2.3-4 2.7-4 per reactor-year) e Percent Contribution of Release Types
- Gross, Early Containment 34 1 0.1 Failure
- Gradual Containment 66 73 60 Overpressurization or Melt-Th rough
- . Containment Intact 0 26 'O I
'Sased on WASH-1400 uncertainty ranges.
MOTE: Exponential notation is indicated in abbreviated form; i.e., 9.9-5 = 9.9 x 10-5, d
1322P120585 2-19
Exhibit 8 Source: PLG, 1983 TABLE 3-3 FAILURE PRESSURES AND VARIBILITIES FOR VARIOUS_
STRUCTURAL FAILURE MODES i .
Med'ian 3 :
3 8 Pressure M 3 ',
Failure Mode (psig) ,
r
.03 .12 216 .12 I. ,-
- n. Wall Hoop Failure 12 .03 .12 223 i If 3.
Dome Hoop or Meridional Failure 12 .03 12 gg.g,p.
281 i
- . Wall Meridional Failure .14 .23 c'49-323 .18 -
Shear Failure of Sase Slab .20 .25 D. 400 .15 Flexural Failure of Base Slab 16 .30 E. 408 .25 F. Shear Failure of Wall at Base sg = Logarithmic Standard Deviation for Modeling 3
= Logarithmic Standard Deviation for Strength 3
= Ccebined Standard Deviation Computed as in Equation 2-6
! i H.I-30
If c
Stmss Single bar with average parameters
- 1. 8 - f, "y m
- 1. 6 -- _._ ~ g.
N Meridional and N %
- 1. 4 ~ diagonal bars ,
at single elevation \ "
~
- 1. 2 -
- , \ !
,/ 's, N 1.0_ , s
', \
- 0. 8 - ', hoop
',8-bar \
\
- 0. 6 - ', , ,,
s o" = nominal yleid strength
- 0. 4 - I m
, (60 ksi or 410 MPa)
', E
. K 0.2 _ s, e 0 ' ' ' ' ' ' ' '
l 1 2 3 4 5 6 7 8 9 10 Strain 1 Fig. 3.14. Effective stress-strain relations of systems of Grade 60 No.18 bars in the containment.
j ATTACHMENT 3 TO AFFIDAVIT OF GORDON R. THOMPSON h
.)
.)
THE SOURCE TERM DEBATE A REPORT BY
)
THE UNION OF CONCERNED SCIENTISTS
_/
J J
e S
O
O.
O
'O THE SOURCE TERM' DEBATE A REPORT BY O
THE UNION OF CONCERNED SCIENTISTS
'O A Review of the Current Basis for Predicting Severe Accident Source Terms With Special Bnphasis on the NRC Source Term Reassessnent Program (NUREG-0956)
O Principal Authors Steven C. Sholly, UCS Risk Analyst
'O Dr. cordon 'Ihcznpson, UCS consultant January, 1986 O
I UNION OF CONCERNED SCIENTISTS
'O 26 Church Street Cambridge, Massachusetts 02238 (617) 547-5552 Washington, D.C., Office:
1616 P Street, N.W., Suite 310 Washingt n, .C. 20036
- O (202) 332-0900
- O
.O - -
4 jO. THE SOURCE TERM DEBATE Table of Contents
.O Foreword Executive Stmanary
'O Chapter One: Source Terms - A Historical Perspective Gapter Two: Outline of the NRC Source Term Reassesment Program t
- O Chapter t ree
- Probabilistic Risk Assesment -- Source Terms in a Risk Context Oapter Four: Neglected Accident Sequences and Contaiment Failure Modes:
Internally Initiated Accidents
- O:
- Gapter Five
- Neglected Externally Initiated Accident Sequences Oapter Six: Neglected Plant Designs and Design Variations 10-Gapter Seven: Limitations.of teoretical mdels for Source Term Estimation
-Appendix to m apter Seven: Objectives, Strengths, and Limitations of the MARCH 2 Code: A Perspective from the User's Manual
.O Chapter Eight: Contalment Performance Under Internally-Initiated Accident Cbnditions Chapter Nine: Neglected Phenomena Milch Can Alter the Nature of Accident
- 0. Sequences Appendix to O apter Nine: Some Steam Explosion Incidents
!O t
IO FOREWDRD Wis repo'rt doctanents a review by the f.hion of Concerned - Scientists of the technical. basis for estimating severe accident source tenus - estimates -
9' of the magnitude and physicochenical characteristics of releases of radio-active ~ materials fran ecumercial light-ater nuclear power plants during severe accidents. 'such estimates tmderlie federal regulation of the operation of civilian comnercial- nuclear power plants by the U.S. Nuclear Regulatory Counission (NRC) .- ,
10 We current source term debate was initiated shortly after the March 28, 1979 accident at the tree Mile Island Unit 2 nuclear power plant ('IMI-2) ,
located near Harrisburg, Pennsylvania. In that accident, although extensive danage to the reactor core (incitx3ing partial melting of the fuel) occurred due to inadequate cooling, apparently very little radioactivity was released to the environnent (several millions of curies of noble gases, principally
the fonn of organic iodide) . Authorities in the nuclear industry, and same at the U. S. Nuclear Reg'ulatory Commission, took the 'INI-2 accident as an indication that little radiological release would occur fran severe accidents in general.
p Prior to the 'IMI-2 accident, source term estimates were contained in two principal references. %e first is TID-14844 which was issued in 1962 and serves as the official regulatory source term. TID-14844 has been used principally, although not exclusively, in the areas of reactor siting and environmental qualification standards for nuclear power plant safety
- . equipnent. . %e second is WASH-1400, the "neactor Safety Sttriy" (also referred
! t as the "aasmussen neport" after its director) , released in october 1975.
- O his report docunented the first detailed analysis of severe accident source terms, and later served as the basis for NBC regulations in the area of radiological energcncy planning (e.g., energency planning zone distances) .
We NRC uses source term estimates for a variety of regulatory purposes.
O Am n3 the principal NRC program areas affected by source tenu estimates are:
- a. Nuclear power plant siting regulations,
- b. Standards for qualification of safety equipnent to assure its operation during accidents when the equignent is needed.
O
- c. Bnergency planning regulations.
- d. Prioritization of nuclear power plant generic safety issues in order to direct increasingly limited NRC research funds tomrd resolution of those issues which have the greatest g estimated impact on public risk.
- e. Severe accident indennification regulations (implenenting the
" Price-Anderson Act" legislation) .
lO
,O '
- f. Safety goals for nuclear power plant operation.
- g. Contalment leak rate . testing requirenents.
- h. Nuclear power plant license restrictions on minimum safety system operability l requirements (so-called " Limiting conditions for Operation"' in the plant Technical 0- Specifications).
- 1. Design requirements for engineered safety features (e.g.,
contalments, - filtration systems, contalment spray and fan cooler systems, pressure suppression systems, etc.) .
O j. Probabilistic risk analysis, including preparation of environmental impact statements (EISs) under the National Enviromental Policy Act of 1969.
- k. Cost-benefit analysis (so-called " Regulatory Analysis") to determine whether safety improvenents should be made to O existing nuclear power plants.
- 1. NRC severe accident policy (i.e., the degree to which severe accidents should be directly addressed in NRC regulations) . -
Since the 1979 'D1I accident, NRC has devoted considerable financial O: resources to developing new source term estimates that would support relaxation of its current safety requirements in the areas of siting requirements, emergency planning standards, and other areas that the nuclear industry perceives to be hindering the preservation of the nuclear option. At-the same time, the agency has continued to ignore the host of generic safety problems that have remained unresolved for years. Given the pervasive effect O on NRC regulations of source term - estimates, any proposed revisions are deserving of careful, _ independent scrutiny. That scrutiny is the purpose for this report. We conclude that there is no technically -defensible basis for relaxing NRC's current ' safety requirements.
O The principal authors of this report are Steven C. Sholly and Dr. Gordon Thompson. Steven Sholly authored chapters one through six and is presently enployed at MIB Associates, San Jose, California. Dr. Thompson wrote chapters seven through nine; he is Dcecutive Director of the Institute for Resource and Security Studies, Cambridge, Massachusetts.
. O lO f
l 0-
O'
,O EXECUTIVE
SUMMARY
The Source Term Reassessment Program The Nuclear Regulatory Comission (NIC) defines a source term as "the quantity, timing, and characteristics of the release of radioactive material O to the envirorsnent following a core melt accident." (NURB3-0956, p. xvill An accurate knowledge of source terms for all plausible severe accidents is -
essential to estimating the overall risk to the public health posed .by nuclear power plants. In turn, a knowledge of such risks is vital in determining the
.O adequacy of NBC safety regulations.
The Commission has - been reassessing source terms for severe nuclear accidents for the past several years; the results of this primarily analytical O effort are stinmarized in its July,1985 publication, NUREG-0956, Reassessment of the Technical Bases for Estimating Source Terms. The primary motivation behind this most recent reassessment was to respond to industry claims that the 'IMI-2 accident showed that previous risk assessments greatly overestimated O risks. The principal prodtx:t of the commission's efforts is a set of computer codes that purport to predict the behavior of nuclear power- plants during
- severe accidents, more realistically, the NRC says, than those used in the Reactor Safety Study (WASH-1400) published in 1975. In the words of
- O NUREG-0956, the new codes represent "a major advance in technology." .More-over, the NBC claims that "[pirincipal omissions ard oversimplifications in the Reactor Safety Study methods have been corrected in the new source term codes." (NURD3-0956, p. 8-11
<Q The NRC nonetheless concedes, in the strnmary of NUREG-0956, that there are still uncertainties in the modeling of severe accidents:
9 o " Remaining areas of uncertainty have been identified in the new source term analytical procedures and indicate areas of research
! that should be pursued." [p. 8-1]
o "It is also clear that uncertainties persist in some of the same areas where major advances...have been made since the Reactor Safety O Study was performed." (p. 8-21
%e NRC has proceeded to use the revised codes to calculate new source terms for a few accident sequences at each of five operating plants. Based on
!O
-- - . . - .=.
\
- O- l I
these calculations the NRC staff has reached a ntnber of tentative conclusions:
. o _ " Source terms were found to depend strongly on - plant design and O construction details , thus making developnent of useful generic
- source terms difficult." [p. 8-4]
o "For most accident sequences, the largest single factor affecting -
- source terms is contairment behavior. A delay of several hours in containment failure will reduce source terms significantly."
O- (p. 8-31 o "New source terms for many accident sequences were found to be lower than . those in - the - Reactor Safety Sttrly, but some were larger."
Ip. 8-4.1
-O Using the new source term' codes, the Comission has also had recalculated the risks from a severe accident for a single nuclear plant, the Surry plant, and concluded that:
O o "A comparative risk appraisal for the Surry plant using the [new source terms) shows a reduction in estimated risk compared with the Reactor Safety Study." (p. 8-51 o "For the other plants, further analyses need to be made before any conclusion can be drawn about changes in estimated risk. The fact
- O that source terms for some accident sequences are not lower than those in the Reactor Safety Study suggest (sic) that significant reductions in estimated risk may not be found in all cases."
(p. 8-51 NUREG-0956 is replete with expressions of uncertainty and cautions on the O
, limitations of the newly revised codes now used to calculate source terms. As I
for the application of the codes for regulatory purposes: "The NRC must i attempt to reduce uncertainties and resolve technical issues with new insights lg from -a strong research program in order to complete the regulatory actions being considered."
L Despite the uncertainties and its cautionary statements, the Comission n ludes NURH3-0956 with a recomendation that the new source term codes can O.
l.
and should be used now to reevaluate existing regulatory practices:
"IIlt is recomended that the new analytical methods should be used to reevaluate current regulatory practices and revise them as needed 10
i l
O l
- 111 - l 1
'O ... . Notwithstanding the limitations and uncertainties that have l been expressed, the new methods are so much better than the Reactor Safety Study methods that their utilization is warranted." [p. 8-6]
he NRC's proposal to reevaluate its regulations using the new and/or O revised source term codes must be viewed in the cold light of the uncertain-ties ard omissions that continue to undermine the analytic techniques of probabilistic risk assessment (PRA). For, given the climate ' in which the source tenn reassesment is being carried out, " reevaluating" regulations in i
O NRC parlance will undoubtedly mean " relaxing" the regulations. However, the source term reassessment does not provide credible scientific support for relaxing safety rules. In the following paragraphs we briefly stanarize the major problens that we have identified with the NRC's risk assessment pro-
.O cedures generally and with its source term reassessment specifically. Our overall conclusion is that while some progress has been made in developing new computer codes, there are still too many important shortcomings, uncertainties, and omissions in reactor risk analysis to warrant regulatory
- O changes that would weaken existing safeguards. In fact, for some accidents the new analyses indicate source terms larger than previously calculated.
PRA SHOR'ICOMINGS
- O Probabilistic Risk Assessment (PRA) is a computer-based process for estimating the likelihood, progression, and consequences of various potential severe accidents at nuclear power plants. Source terms are an integral part lO of the PRA process in that they describe the final releases of radioactive materials into the environnent that would occur during postulated accic'ents.
(See Chapter 2ree.) However, changes in NRC regulations based on perc ~ived j reduced risks from nuclear accidents would have to be based on the total PBA O process and not just on revised source terms.
Some of the generic shortcomirgs with PRAs - reflected in uncertainties in their results - are described in Chapter Three. They include:
I O
o Their failure to take into account the aging of structures and components. Component aging could affect both the likelihood of accidents l
'O
- ,+v . - . - _.7-p ~_7__--77 -----,-7.-.y-----9-,y- - - ~ y,_ , , , _ - , - ,
1
- O.
- iv -
.O occurring and the response of one part of the plant to component failures elsewhere.
- O' o W eir -failure to take into account the fact that nuclear plants sometimes operate in violation of their tectnical specifications, either temporarily with the approval of the NRC, or in direct violation of their licenses.
- O o We failure to model partial operation of reactor systems. System failures are almost always modeled in PRAs as complete failures. In reality there are a wide variety of partial system failure modes, some of which have O more severe consequences than a complete failure.
o Large uncertainties about the strength and failure modes of contain-ments. These uncertainties lead directly to large uncertainties in subsequent risk estimates.
SHOR'IC04INGS IN ESTIMATING SOURCE TERMS
,0 Source terms are an integral part of estimating the risks of severe I nuclear accidents. They provide the essential link between the initiating i event (s) (such as a pipe rupture, valve failure, loss of feedwater, etc.) and
- O the final impact, if any, on the community surrounding a specific plant.
Given the assumed accident sequence, source terms indicate when and how much radioactive material escapes from the plant. Using this information the public health and other impacts can be estimated for the particular geograpi-
-O ical area around the plant.
The NRC source term assessment program has led to the development of new computer programs as well as the improvement of older ones to predict the release of radioactive materials resulting from severe accidents. The NRC has also supported major experimental research into the behavior of nuclear power plants during the throes of core melt accidents. Despite these efforts, however, serious problems and uncertainties remain. Thus the predicted risks lO
Oi
_y_
O.
from nuclear accidents must. be regarded as having a large measure of uncer-tainty.
'O the problem of completeness one of the most ' serious problems afflicting both PRA analysis and source term assessment is that of " complete. ness," the identification and inclusion of O. ,11 risk-significant accident sequences in the analysis. (See Chapters Four and Five.) The computer models themselves . do not ' identify possible accident
~
sequences; there is no analytical process for doing so. Rather, the identifi-cation procedure must be undertaken . " manually" by the analysts themselves O using event- and fault-tree analysis. Only after the general flow of the specific accident sequences are prescribed by the analysts can the codes be employed to predict the details (temperature, pressure, mass flow, etc.) of the accidents. As a result, if important sequences are' overlooked - as they O undoubtedly are, given the enormous complexity of nuclear plants and the countless number of ways they can break down - the calculated risk posed by reactors will necessarily be underestimated.
- O Neglected Internal Events Some of the important internally initiated sequences with potentially large source terms that were not included in the NRC's reassessment are listed below. See Chapter Four for further discussion.
fO o An Anticipated Transient without Scram (A'Iws) is a failure of the reactor to shut down automatically when it is called upon to do so. A'Iws sequences are amng the mst important accident sequences for PWRs and are important to the source term issue because they could lead to unusual, and
- O thus far unanalyzed, accident conditions. For example, the high pressure resulting from an A'IM sequence could lead to the failure of the pressure vessel head if the studs holding the head failed or if the head lifted and failed to reseat properly.
O o Steam Generator Tube Ruptures (SGTR) can occur either as the initiating j
event in an accident sequence or as a resultant failure in other accident sequences such as AlwS. In either case, a SGTR accident could lead to a l0.
)
- vi -
,j sizable source term by providing a direct release path to the environment even though containment failure is not predicted.
3 o Ioss-of-Coolant Accidents (LOCAs) involving the in-core instrument tubes in PWRs form a special case of loss-of-coolant accidents. In some plant designs such accident sequences could readily lead to containment failure and a large source term.
-)
o Pressurized Thermal Shock (PTS) sequences could result f rom the overcooling of the reactor vessel followed by repressurization of the reactor cooling system. P'IS is of particular concern with older pressurized water 3 reactors (PWRs).
o The failure of seals in reactor coolant pumps (RCPs) can lead to small IDCAs; there have been nine major RCP seal failures in U.S. PWRs since July
-] 1969. RCP seal failures are important both as LOCAs, themselves, and as complicating factors in other accidents, such as the loss of all electrical AC power.
3 o Reactor vessel rupture sequences in both PWRs and BWRs could lead to a relatively early failure of the containment and they should be included within the source term assessment.
J o There have been several thousand containment isolation failures and pre-existing leaks in containments discovered between 1965 and 1983. These routine and often undetected failures in containment integrity could lead to large source terms for many accidents and should be included in the 9 reassessment of all reactor designs. ,
o Accident sequences are typically nudeled with the assumption that safety systems are recovered prior to core melting. There are examples, 9 though, of later recovery of various safety systems -- during or even after a core melt - that could lead to large source terms. For example, the late recovery of containment sprays after the buildup of hydrogen could lead to a O
O
- vii -
hydrogen burn threatening the containment. Similarly, the late recovery of auxiliary feedwater could lead to a failure of steam generator tubes -from thermal shock and high pressures. I. ate recovery of the emergency core cooling.
'O system could actually accelerate the melting of a core. And finally, late recovery of hydrogen igniters in an ice condenser plant -- after hydrogen had built up - could lead to a large hydrogen burn.
10 o The heating up of the reactor cooling system during a core melt could
. lead to a structural failure of a main coolant pipe or of steam generator tubes. Either case would be a sequence considerably different from those now modeled.
O
. o Steam explosions within the reactor could lead to the separation of either the upper or lower reactor head with consequent damage to the containment or other important safety systems. Such explosions can also occur
.O-While probably incapable of causing the i outside of the reactor vessel.
containment to fail, such external explosions could nonetheless lead to the redistribution of fission products within the containment and, possibly, to other, J 1arger steam explosions. In BWRs, steam explosions outside of the O ressil might cause failure of the containment in Mark II designs.
i o The failure of a discharge line from a safety relief valve (SRV) in a Mark I or Mark II BWR containment combined with a stuck-open SRV could lead to '
O a rapid overpressurization and early failure of the containment. This acci-dent sequence was not included in the source term reassessment.
i o A number of precursors to "interf acing" LOCAS - i.e. accidents at the
.O' -boundary or interf ace between the high pressure reactor coolant system and a low. pressure system outside the containment -- have occurred in BWRs over the i
past decade. These accident sequences could lead to a bypass of the contain-
, ment but were not considered in the source term reassessment.
- O l o While Mark I/II BWRs are rormally operated with an inerted containment (that is, with the oxygen renoved f rom the containment atnosphere), they are
'O
O
- viii -
O permitted to operate de-inerted shortly after startup and shortly before shutdown. Under de-inerted conditions even a small hydrogen burn during a core melt accident could cause containment failure. Accidents with de-inerted O containments have not been included within the source term review.
o Main steam isolation valves (MSIVs) are large, normally open valves in the steam lines of BWRs and are supposed to close and stop the flow of steam O during accidents. If these valves leak during an accident, fission products released from a melting core could escape into the environment. MSIV leakage has not been taken into account in estimating source terms for all BWR accident secuences.
O Neglected External Events In addition, external events could cause or aggravate accidents. While this is well known, no externally initiated accident sequences were included within the NRC reassessment. Some of the O more important sequences are described in Chapter Five of this report and are briefly sunmarized here.
o Earthquakn can damage plant equipnent and lead to a wide variety of O accident sequences with potentially large source terms. For example, earth-quakes may be one of the few credible causes of the so-called " double-ended guillotine break" in one of the large pipes carrying coolant to or f rom the reactor vessel. Aftershocks of a major eartlquake might also be the final O straw for a containment weakened by a major quake and then re-pressuri::ed during or after the resulting core melt accident. Seismic events could cause the wall supporting the polar crane to fail, shearing off the main coolant and steam pipes which pass through the containment. In another sequence, the O reactor vessel could sway back and forth shearing off the main steam lines.
If the vessel then toppled it could strike the wall of the containment causing it to fail. Given the uniqueness of U.S. nuclear plant designs, such seismic analyses of source terms should be carried cut for each plant.
'O o While the crash of large aircraft into containment buildings are rou-tinely considered in PRAs, other crash scenarios have not. The crash of an O
r s
~ ~
h 3i airplane into the auxiliary building, for example, could cause otherwise unan-alyzed ~ failures such as ' the shearing of pipes between the containment and auxiliary buildings or the destruction of imortant safety systems. Other
) crash scenarios include imacts into offsite power lines' and structures or the conduits that carry cooling water to and from the plant.
o Accidents resulting from sabotage have not been considered in PRAs to
) date despite the fact that a number of incidents have occurred at operating-plants. Sabotage, whether internal or external, could lead to extremely serious consequences. Since saboteurs could both cause an accident and y disable the equipment needed to mitigate the accident, this should be considered in the source term review.
o There have been a large nutter of incidents in recent years in which biofouling - the clogging of pipes by the growth of a<patic organisms - has ,
) resulted in the partial or complete unavai hbility of safety systems in nuclear power plants. We pheremeron, which could reault in the loss of containment heat renoval, is not included in the reassessment of source terms.
) o Some aspects of fires at nuclear plants have not been adequately re-viewed for their relevance to source terms. The effects of sucke damage, of operating fire-protection systems, and of the high humidity levels resulting y from fighting fires need to be further analyzed.
o Other factors considered " external" events in causing or contributing to severe accidents include errors in design, construction, or installation, and unusual operating mistakes. For example, valves have been found installed
) backwards or upside down and large voids have been found in the concrete of containments.
Neglected Plant Designs and Design Variations
)
The NRC reassessment (NURB3-0956) developed revised source terms for five operating U.S. plants. The regulatory validity of using only these five
).
d )
-X-O reference plants as a basis for changing the rules applying to all plants depends on whether they fairly represent the entire light water reactor (LhR) population cf about 100 plants. As a result of the lack of standardization 0 anong u.s. reactors, however, there are only broad similarities among them, even within the same plant class. These differences are reflected, in part, in the widely v.ffering calculated probabilities for identical sequences at different plants. Some of the important differences a::eng operating plants O are sumarized next.
Safety and Reactor Coolant Systems The only two PWR designs analyzed in
! the source term reassessment were 3- and 4-loop Westinghouse reactors. No O two-loop Westinghouse or any Babcock and Wilcox (B&W) o r Combustion Engineering (CE) designs were analyzed. Yet there are many safety-significant differ ences a:rong the various pressurized water reactors in operation. In some plants the " hot legs" enter the steam generators f rom above; in others O from below. Moreover the number of loops per generator differs from plant to plant. Similarly, F5W steam generators are once-through while CE and Westing-house are inverted-U shaped. The water volumes in the secondary sides of the steam generators also differ substantially.
O Some PWRs have pilot-operated relief valves (PORVs); others do not. The lack of PCRVs will affect plant response to various accidents. Without PORVs there is no chance of using " feed and bleed" -- a process that utilities hope O will be able to rer:ove decay heat when the steam generators cannot do so.
Even arrong plants of a single vendor there are significant differences. The numbers of steam generators, and hot and cold legs differ. The locations of pressure vessel penetrations vary. A review of U.S. PWRs showed that there O are twenty different auxiliary feedwater systems.
There are many different forms of containments not represented in the NRC's source term review. For example, no large dry steel containments were
-O analyzed. Similarly, ice condenser plants have differing containments, spray configurations, vacuum breakers, and design pressures, yet only one ice condenser plant was reviewed.
!O
_ . . __ . ~_ . . . .- .
'O '
~
- xi'-
O' Boiling water reactors also differ considerably among themselves. Ibr exanple, all of the Mark II containments are unique. And the designs of the pedestal areas where the reactor vessels are supported differ significantly.
_O.
The Mark-III's also have significantly different containments and drywells.
. Lastly, PNRs differ greatly in the designs of the cavities below the 4
pressure vessels. - Some are designed to collect water during an accident;
-o. others are designed to remain dry. These design differences have important implications for a wide variety of accident phenomena including steam I explosions and high pressure melt ejection. ]I g.
- O Limitations of neoretical Models m
,s After a specific accident sequence has been selected for review it is i necessary to employ a series of computer models to estimate the associated lO source term. These models are enormously complex and perforce either neglect or treat in a simplified manner various physical or chemical pheromena. Some of the limitations and the uncertainties the models introduce into source '
term estimation are sumarized below. See Chapter Seven for further discussion.
The codes suffer from three principal limitations. First the user is req 11 red to make many assumptions before the codes can be -run. These assump-t tions take the form of parameters that the user nust initially specify. For
,O example, the temperature at which the core will melt must be initially specified. The user nust also choose one of three options for describing the behavior of the core after melting begins. Current models simply assume that iodine will be present in the reactor coolant system as highly soluble cesium n
V iodide, although recent experiments have shown that radiation fields will j' dissociate cesium iodide. Similarly, the conditions at which the containment will fail must be specified by the user and are not calculated by the models themselves. Sensitivity studies show that the results of several of the
!O. nodels now employed are highly sensitive to the values initially specified.
Source term estimates are thus highly dependent on many initial assumptions on
- the part of the user, some of these assumptions being uncertain over a very iO
O
- xii -
O wide range. That is, the end results differ greatly depending upon the asstaptions used.
O A second general problem with the models relates to their validation: the process by which confidence is gained that the models adeg2ately represent the relevant phenomena. Unfortunately there are no generally agreed upon criteria for validating source term codes. Instead, developers, users, and reviewers O of the codes use their judgment to determine if a particular code is
" adequate." It is generally accepted that most of the source term codes have not been properly validated.
O Lastly, because the codes have not yet been validated it is not possible to quantify the uncertainty in their results. Moreover, the state of the art would have to advance substantially before this would become possible.
O' In short, the codes are still in a developmental stage, require many assumptions to be made by the user, have not been validated ,against experi-ments, and have very large but unquantified uncertainties in their results.
g' Containment Performace During Severe Accidents Containments represent the last major barrier to the escape of radio-active materials during a severe nuclear accident. The timing and manner of O containment failure will have a critical effect on the amount of materials that will enter the envirorment. In this section we summarize the uncertain-ties surrounding the performance of containments during internally initiated
! severe accidents, the only ones considered in the source term review. The uncertainties are discussed in detail in Chapter Eight.
There are many different containment designs in U.S. LWRs. Steel contain-ments are the most comon in BWRs and large dry concrete containments in PWRs.
O There are, though, many variations in their design details. Taking all the differences into account gives a total of 22 different designs.
l l
O
I oi
- xiii -
O
, he - record of containment integrity during normal operations is not especially encouraging. One study found BWRs adequately leak-proof only 77 percent of the time, and PWRs 96 percent.
O Present containments have not been designed to cope with many of the -
severe accidents which are being considered in the source term review. If they are capable of withstanding these sequences it would be largely a result 0 -- of conservatism in their deston. Thus, containments are not designed to with-stand external blast waves, nor missiles from a ruptured turbine, nor the impact . of a large aircraft, unless the plant is located near an airport.
Similarly, they are not required to withstand missiles from the rupture of the O. flywheels that are coupled to reactor coolant pumps, nor the rupture of a major component of the reactor coolant system. .
Over-pressurization of containments can occur during accidents either
- O slowly (statically) or very quickly (dynamically). steam and hydrogen explo-sions could lead to dynamic over-pressurization and these have not been 'ade-l quately addressed in the source term review.
iO There are other phenomena that can threaten containments, if indirectly:
among them, high temperatures, ionizing radiation, and missiles. These too have not been adequately treated in the source term review. We uncertainty l of these effects is directly reflected in the uncertainty of the manner and
- O timing of containment failure, md hence in the corresponding source terms.
In sum, the role of containments in mitigating accidents is crucial, yet there are major inadeg2acies in the analyses that have been completed to date.
- O In the source term reassessment, arbitrary assunptions have been made about containment performance. Experimental wo rk is required to reduce uncertainties, and analyses will have to be performed for each operating plant l
to determine the likely timing and mode of failure for specific severe accident sequences.
lO O
O
- *iv -
lO Neglected Phenomena 'Ihat Can Alter the Nature of Accident Sequences g Severe reactor accidents will give rise to a wide variety of physical and chemical phenomena. This section describes some of the phenomena that have been " neglected" in current source term estimates, and which, if included, could alter the nature of the accident sequence. Until these neglected effe ts can be either properly included or dismissed as insignificant, source O
term estimations cannot be regarded as convincing or complete. See Chapter Nine for further discussion, g Steam Explosions Steam explosions can be caused by the r@id mixing of molten material with water. Many violent explosions of this sort have occurred in various industrial facilities. The worst steam explosion would be one in the reactor vessel in which the head blew off the vessel and penetrated the containment. Depending on their location, steam explosions could rupture O.
the reactor cooling system, the containment, or increase the amount of radioactive material suspended in the containment atmosphere.
g Steam explosions, particularly inside the reactor vessel, are very diffi-cult to mdel and expensive to experiment with. The likelihood of such explosions is very uncertain; however, they cannot be ruled out. Since they have the potential to cause direct pathways for radioactivity into the environment their eventual inclusion in the source term models is essential if reliable estimates for source terms are to be made.
High Pressure Melt Ejection (HPME) HPME is the ejection of molten j material at a high velocity from the reactor vessel. The miten material
,0 would arise f rom a core mell while the reactor coolant system remained at a l
high pressure. As with steam explosiors, HPME is an uncertain but potentially significant event for all types of light water reactors. In the worst of circunstances it could lead to the prompt failure of the containment with a O
large release of radioactivity. It would also provide mechanisms for the suspension of radioactive materials in the containment atmosphere.
- O 1
l
I D
1 _ xv -
0-As indicated above, there is considerable uncertainty over the likelihood and processes involved with HPME and much more theoretical and experimental work is required before it can be f ully integrated ' into source terms
'O' estimates.
Hydrogen Explosions During a severe accident, large annunts of hydrogen can be generated by a number of processes: by chemical reactions between the "y
steam and the fuel cladding, from iron-steam reactions, and f rom core-concrete reactions after the fuel has melted through the reactor . bottom. The collection of this hydrogen in the containment could lead to an explosion that could pose a major threat to the containment.
w Some plants are equipped with hydrogen igniters but these could be inoper-ative during some accidents. PWRs with large dry containments have no such -
Igniters. Even plants with normally inerted atmospheres (certain BWRs) are O susceptible to hydrogen explosions in- their reactor buildings. Source term estimations to date have not addressed the if kelihood and consequences of hydrogen burns. We subject requires'more experimental and theoretical work.
O Vulnerability of Steam Generators Steam generetors in PWRs contain thousands of thin-walled tubes that transfer heat from the high pressure reactor coolant system to the lower pressure steam turbines. During a severe
.O accident there are several phenomena (e.g., heat f rom fission products deposited in the tubes) that could cause these tubes to rapidly heat up and
.perhaps fail. Such a failure would lead to a direct path for radioactivity f rom the core to the environment. The subject requires nuch nere attention in O
the source term reassessment.
Secondary Releases There are two other processes not considered within the source term reassessment that could contribute to large source terms. We
- O first is resuspension of radioactivity into the air from water or dry surf aces following the explosive failure of the containment. We second is melting of some of the fuel in the spent fuel pools. This might occur following water
- O
. _ . , _ . ~ ... . , ,,,,.~, .....,,,,m. , , _ _ - . . , , . , . , . . - . . . -m , . - . . - . . , , _ . - _ . . . - - . -
O I
l
- xvi -
g loss from the pool either f rom a breach of the pool as a result of an explosion or from evaporation following a failure of the pool coolng system.
It appears that a release f rom a spent fuel pool could add significantly to O the source term for a reactor accident. Further effort is required to . assess the importance of this I ienomenon.
f CONCLUSIONS O-While the NE source term reassessment program has led to the development of new computer models, the remaining technical uncertainties preclude abanchnment of current safety requirements. As this present report makes O clear, it is still not possible to make precise, quantitative statements about overall reactor risks.
e The underlying tool of nuclear risk analysis --
probabilistic risk
- O malysis - suffers f rom - a range of shortcomings which, in light of the enormous complexity of nuclear power plants, may never be overcome.
Source terms are a critical element of risk assessment and estimating
!O them accurately is an extremely complex task that is just beginning to be seriously addressed. Even the sinple identification - for each plant - of all th7 important accident sequences, internally as well as externally initi-ated, that should be examined is a daunting task and may not, in principle, be
.O solvable. Modeling the vast mmber of non-equilibrium chemical and physical processes and interactions that constitute these sequences is also beyond I today's capabilities. Indeed it is by no means clear that all of the physical I
phemmena that are important in the development of these accidents have even
'O been indentified.
Protecting public health and safety is the NRC's primary responsibility in regulating nuclear power. In the face of these uncertainties and the
- O relatively primitive state of nuclear risk assessmen t, we find the NRC's proposal to begin relaxing its safety regulations totally unwarranted.
- O
10 0 CHAPTER ONE Source Terms - A Historical Perspective StPNARY
!O: ne primary responsibility of the NRC in the regulation of nuclear power plants is the protection of the public health and safety. Many of the Comission's safety regulations depend! critically on estimates of accident source terms, which are defined as the quantity, timing, and character-istics of radioactive materials released to the environment O following core melt accidents.
Source - terms have been used by - the _ NRC since 1957. he .
i source terms for regulations governing the siting of nuclear plants were published in 1962 in the TID-14844 report.
Source terms derived from this report have been subsequently O used by the NRC in accident evaluation in its Regulatory Guides. In 1975 the NRC released the Reactor Safety Study; the source terms in this report have been used in subsegant probabiliste risk assessments. After - the 'IMI accident the NRC undertook a review of the source term issue to determine whether the source terms it now uses are too large and, as a O result, its regulations needlessly restrictive. We source terms for five plants and a number of accident sequences have been estimated.
In July,1985 the NRC published the results of this review, NUREG-0956 and invited public coment. h is docment is the
- O - subject of the present ocS review.
1.0 Introduction I
The basis for federal regulation of comercial nuclear power plants is
.O As a general rooted in the protection of the "public health and safety".
j matter, the public health and safety is threatened by a nuclear power plant-when a large release of radioactive materials to the enviroment occurs. An estimation of the release magnitude and the chemical and physical characteris-
.O tics of a release of radioactive materials from a nuclear plant for a specific accident sequence is referred to as a source term.
Since 1980, the U.S. Nuclear Regulatory Comission has had underway a O accidents involving reassessment of source terms for severe accidents -
severe core damage which could result in a radiological release to the environment if the containment fails or is bypassed. his lengthy reassess-O
'D -
1-2 ment program has culminated in the release for public coment of NUREG-0956,
-O Reassessment of the Technical Bases for Estimating Source Terms, July 1985
he importance which NRC attaches to the source term issue cannot be O overestimated. Source terms affect many areas of NRC regulation of nuclear power plants, including plant siting, emergency planning, indemnification policy, equipment qualification, and contalment leak rate testing require-ments. For example, one utility has already requested an exemption to O existing regulations which, if granted, would reduce by 96 percent the area 1
for which emergency planning is required. Given the pervasive impact on NRC regulations of passible changes in source term estimates, an independent review of the NRC source term reassessment program is thus timely and O important.
Before proceeding to an analysis of some of the technical issues involved with source term estimation, it is useful to understand the history
.O of source term estimates. Chapter One of this report briefly traces the developnent of source terms from 1957 through 1985.
i
{ 1.1 The Earliest Source Terms: NASH-740 l
- O
- In March 1957, in anticipation of the operation of the first comercial nuclear power plants in the U.S., and as part of the consideration of liabil-ity limits for accidents at nuclear power plants (later finalized in the O " Price-Anderson Act") , the U.S. Atomic Energy Commission (AEC) transmitted to l Congress a report on nuclear power plant accidents. %is report, referred to l as NASH-740 [ BECK, ET AL. , 1957], provided upper bound source term estimates for three " damage states" created by nuclear power plant accidents.
lO
%e three source terms postulated in WASH-740 were:
- i I a. %e contained case - all of the fission products in the core 10 were asssmed to be vaporized and dispersed within the con-tainment shell, but no release to the enviroment was assumed to occur (no containment failure) .
l
- O -
1-3
- b. De volatile release case - all of the -volatile fission
- O products in the core were assumed to be released to the environment at the time of the ' accident (imediate contain - '
ment failure or contalment isolation failure).
- c. Re 50% release case - half of all fission products in the '
core were assmed to be released to the environment (imed-
- O iate containment failure) .
%e contained case does not, strictly speaking, result in a source term per se_ since no release to the enviroment actually occurs. W is case was of
!O . interest because no radiation shielding was assmed to be used, resulting in a
. gama " shine dose" to persons near- the plant due to gama radiation penetrat-ing the steel containment shell.
Q 1.2 The Siting Source Term: Part 100 and TID-14844 In 1962, the AEC first defined reactor siting requirements . in its regulations at 10 C.F.R. Part 100. @e siting regulations were based on an lO AEC report which defined the source term for the " maximum credible accident" required to be considered in the design and siting of nuclear power plants.
! his report, referred to as TID-14844 (DiNUNNO, ET AL.,1962), made what were at the time claimed to be conservative assumptions in defining the regulatory i
O source term. @e report states in pertinent part:
"his technical docment sets forth one method of computing distances and exposures, for ene general class of reactors. In developing this example conservative' assumptions have been O intentiMally employed . . . For pressurized and boiling water reactors, for example, the ' maximum credible accident' has frequently been postulated as the complete loss of the coolant upon complete rupture of a major pipe, with consequent expansion of the coolant as flashing steam, meltdown of the fuel and
. Partial release of the fission product inventory to the O atmosphere of the contaiment building . . . @e amount of each kind of radioactive material present in a reactor system can be estimated fairly closely . . . but the quantity of this material that would be releasd as a result of an accident is unpredictable
. . . In using this method, it is recognized that . . . the results obtained are approximations, sometimes relatively poor o ones, to the result which would be obtained if the effects of all the variables and influencing factors could be recognized and fixed with certainty - an impossibility in the present state of the art . . . (T)he net effect of the assmptions and approx-imations is believed to give more conservative results (greater O
l
~ -, - . - - . . - - . . - - - - . - - - - ~ _ - - - - - - - - _ _ _ _ _
O 1-4 distances) than would be the case if more accurate calculations
'g coulci be made."
. TID-14844 provided that the maximum credible accident (now . referred to as the " design basis accident" or DBA) would result in the release to the ntainment building atmosphere of 100% of the noble gases, 50% of the iodine, O
and it of the remainder of the fission product inventory. Half of the quantity - of iodine was assumed to plate out in the reactor building ~or internal components, and the remaining iodine was assumed to be. available for g release to the environment. For the purposes of estimating releases to the enviroment, the contairunent was assmed to isolate and leak at a constant volumetric leak rate (0.1% per day) .
The TID source term has become widely used by the NRC. Se source term O
incorporated in Regulatory Guides 1.3 and 1.4, which provide the basis for accident evaluation in - utility-submitted Final Safety Analysis Reports (FSARs), was derivsd from the TID source term. It is also used in site g suitability assessments and in establishing safety equipnent environmental qualification standards. Until 1979, the TID source term (and the low population zone established under le C.F.R. Part 100 using the TID source 1
term) formed the basis for offsite radiological emergency planning. It has g also been _used in defining what constitutes radiological sabotage (deliberate acts must result in offsite doses exceeding the Part 100 limits to be officially classified as radiological sabotage).
- g 1.3 The Reactor Safety Study, (NASH-1400)
In 1972, the AEC undertook the first detailed assessment of both the probabilities and consequences of nuclear power plant accidents. Completed
'O and released in October 1975, the Reactor Safety Study (RSS, or, as it is often referred to, WASH-1400 or the "Rasmussen Report") defined source terms for a variety of accident sequences at a pressurized water reactor (Surry Unit
s lO
%e RSS estimated source terms for many specific accident sequences.
Prior to performing accident consequence calculations, however, accident
.O
- O 1-5 L sequences and their associated source terms were grouped or " binned" into O " release categories." In this process, accident sequences with similar characteristics and source terms were grouped into a single release category.
We RSS source term analysis incorporated certain simplifications due O to limitations in the state-of-the-art and the lack of experimental data. We RSS also stated that deposition of materials released from the core could occur prior to release to the containment. Due to the difficulty involved in providing estimates of primary system retention, however, the RSS assmed -
O essentially no deposition in the primary coolant system. Probabilistic risk assessments performed subsequent to the RSS have used either the RSS release categories with minor modifications, or the computer models developed during or shortly after the RSS was published.
O Criticisms of the RSS and its PRA techniques followed quickly upon its publication. In August 1977, the Union of Concerned Scientists published a book-length review of the RSS concluding that its assertions on nuclear risks O could not be trusted (UNION OF CONCERNED SCIENTISTS, 1977). Anong the problems plaguing PRA were the following:
- a. Much of the elementary data on the reliability of plant O Components were incomplete, uncertain, or unavailable;
- b. For most of the RSS analysis, failure of one component was assumed to be independent of failures of other components.
Wat is, "comon mode" failures were largely ignored; O c. RSS generally assmed that current reactor designs were adequate, overlooking possible intrinsic design deficiencies;
- d. We RSS was lax in addressing major problems that contribute to nuclear risks, such as aging and degradation of plant components, earthquakes, sabotage, and terrorism.
O In respanse to these and other criticisms, the NRC established a Risk Assessment Review Group in September 1978. We Review Group reported that while the RSS was a " substantial advance" over previous assessments of reactor O risks, the Review Group could not determine whether the RSS accident probabil-ities were too high or too low (LEWIS, ET AL., 1978). %e Group also questioned RSS's " questionable methodological and statistical procedures."
1 O
O l-6 In-1979 the NRC issued a policy statement retracting its endorsement of the RSS risk estimates: "%e Comission does not regard as reliable the Reactor Safety Study's ntnerical estimate of the overall risk of reactor 4
accident."
O 1.4 .The Three Mile Island Accident, March 1979 On 28 March 1979, the Unit 2 reactor at the %ree Mile Island nuclear M station -(located approximately ten miles southeast of Harrisburg,
. ' Pennsylvania) suffered' the most' serious commercial nuclear power plant i accident in the U.S. to date. %e reactor, a Babcock & Wilcox PWR with a
- large dry contalment, suffered severe core damage and substantial core O . melting dee to an extended loss of coolant and misoperation of the emergency
, . core cooling system [KD!ENY, ET AL.,1979; ROGOVIN, ET AL.,1980] .
Although the WI-2 containment successfully isolated early in the accident sequence, several million curies of noble gases (principally 1
xenon-133) and a minor amount of iodine were released to the environment.
Much has been made of the disparity between the relatively large amounts of
! noble gases released compared with the- small amounts of iodine. Indeed, some industry observers have stated that this " proves" that severe accident source terms had been overestimated in the past.
i
! ne MI-2 accident and its source . term is of sufficient general O interest, and has been so badly misconstrued, that a brief discussion is included here. %e WI-2 accident was a partial core melt accident in which the reactor vessel never failed, and in which containment integrity was L maintained throughout the course of the accident. Up until the SI-2 accident, no such sequence of events had ever been analyzed.
%e fission product transport pathway from the core to the containment l is of particular interest. %e pathway in the MI-2 accident carried fission h.. products from the r: ore throtx3h the reactor coolant piping to the pressurizer.
! From the pressurizer, the released materials were carried through the stuck-
! open relief valve and through a pipe to a water-filled tank located in the t
lO l
O 1-7 bottom of the containment. As a consequence of this water-bounded release O pathway, most of the radioactive materials in the containment atmosphere on the day of the mI-2 accident were in the form of chemically inert noble gases, mostly xenon-133 and krypton-85.1*1 0 This is not expected to be characteristic of many PWR core melt accidents. As observed in a recent report on releases during nuclear accidents by Battelle Columbus Laboratorie. [GIESEKE, ET AL.,1984d]:
O "In the tree Mile Island accident, the release of radionuclides to the containment atmosphere was significantly limited by a large quantity of water in the pathway of release to the environment. Wis is not characteristic of accident sequences g leading to complete core melting of the type selected for the present analysis. In the sequences studied here, fuel heatup would not begin until the water level had dropped below the top of the care, and very hot steam and hydrogen from the melting core would superheat the structures in the pathway to the containment."
O In short, the SI-2 accident was most atypical and, given the pathway of the radioactive releases through water, the releases to the environment should have been expected to be small. nus, the mI accident " proves" O nothing about releases from other kinds of accidents.
1.5 The NRC " Technical Bases" Report, NUREG-0772 O On August 14, 1980, two scientists at Oak Ridge National Iaboratory and one at Los Alamos Scientific Laboratory wrote to the chairman of the NRC questioning the validity of the then-accepted methods for estimatire the release of radioactive iodine under certain accident conditions. In partic-O ular, the scientists stated their belief that iodine would be released from the core as cesium iodide (CsI) and, being much less volatile than molecular iodine, much less would escape into the envirorment (STRATION, PALINAUSKAS &
CAMPBELL, 1980) .
O Other letters to the NRC Commissioners followed, some of which stggested that source terms for severe accidents had been overestimated by one O
'O -
l l-8 j 1
- to two orders of magnitude (i.e., by factors of le to 100) . As a result of
- O these' letters, the NRC Comissioners held a public meeting on November 18, I
. 1980 to discuss the source term issue. Following the meeting, the NRC staff was directed to develop a plan to explore and resolve the source term issues l raised in the meeting and in the letters to the comissioners, iO-In June 1981, the NRC issued NUREG-0772, Technical Bases for Estimating Fission Product Behavior During LNR Accidents (SILBERBERG, ET AL., 1981).
I %is report st.mmarized the state-of-knowledge in 1981 concerning fission l0 product transport estimates and methods. Among the principal conclusions of
! NUREG-0772 were the following:
i
- a. Iodine would be released under most accident conditions as
!O cesitan i dide (CsI) .
- b. About 0.03% of the iodine inventory of the core would be converted to methyl iodide, a volatile form of iodine.
- c. Predicted retention of fission products in the primary system
- in core melt accidents varies from almost none to greater
- O than set under certain conditions.
l
- d. Iodine attenuation in the containment for early contalment i failure would not depend greatly on its form (CsI aerosol versus iodine vapor) . Attenuation in the reactor coolant system Would be much greater for CsI, however.
l e. Some engineered safety features (such as supprassion pools, l containment sprays, and ice beds) should be very effective in j removing fission products under most conditions.
i h In December 1982, the NRC's Office of Nuclear Regulatory Research i formed the Accident Source Term Program Office (ASTPO) to coordinate NRC l source term programs and research. %is office has subsequently published
- i. NUREG-0956 as the culmination of the NRC source term work.
lO i
l 1.6 The Source Term Reassessment: BMI-2104 and NUREU-0956 l
! %e fragmented analyses carriad out for the NUREU-0772 report indicated h that the source terms for some accident sequences might have been overesti-l mated in the past due to the neglect of certain fission product reduction
- mechanisms. In order to obtain revised source term estimates, NRC contracted l
l lo
O l-9 with Battelle Coltanbus Iaboratories (with assistance from Sandia National Iaboratories and Oak Ridge National Laboratories) to estimate source terms for O selected accident sequences at six reference nuclear power plants. %ese analyses were conducted using ' a series of camputer codes to estimate the time-dependent source term for severe accident sequences.
O 1hese analyses were published as BMI-2104 in July 1984. We BMI-2104 analyses were subject to a detailed review by a panel of 15 persons invited by the NRC as well as by invited observers. As the review process went forward, additional needs were identified. As a result, the source term reassessment O program was expanded to take into account some of the review coments.
1.7 End Note for Chapter One O
1.1 About 38,000 curies of iodine were airborne in the contaiment atmosphere on the day of the accident (about 0.06% of the core inventory) (WILSON, ET AL., 1985). Sampling of the TMI-2 containment atmosphere established that about 2,000 curies of iodine were airborne in the containment as O late as three days after the accident (about 0.003% of the core inventory). In addition, another 20,000 curies of iodine (about 0.4% of the core inventory) was found on containment surfaces in June 1979 (PELLETIER, ET AL., 1981].
1.8 References for Chapter One O
BECK, ET AL.,1957 C.K. Beck, F.P. Cowan, K.W. Downes , J. A. Fleck, Jr., J.B.H. Kuper, J.
McLaughlin, I. Singer, and M. Smith, Theoretical Possibilities and Consecuences of Major Accidents in Large Nuclear Power Plants, U.S. Atomic O Energy Comission (Washington, D.C.), WASH-740, March 1957.
DiNUNNO, ET AL., 1962 J.J. DiNunno, F.D. Anderson, R.E. Baker, & R.L. Waterfield, Calculation of Distance Factors for Power and Test Reactor Sites, Division of Licensing
!. and Regulation, U.S. Atomic Energy Commission (Washington, D.C. ) ,
jO TID-14844, second printing, 23 March 1962.
, J. A. Gieseke, P. Cybulskis, R.S. Denning, M.R. Kuhlman, K.W. Iae, and H.
l Chen, Radionuclide Release Under Specific LWR Accident Conditions: Volume _
L V, PWR-Large, Dry Containment Design (Surry Plant Recalculations) ,
!O Battelle Coltsnbus Laboratories (Columbus, chio), BMI-2104, vol. v, draf t, July 1984.
i IO I
4
^
O-
~1-10 i
- J.G. Kemeny, B. Babbitt, P.E. Haggerty, C. Imwis, P.A. Marks, C.B.
'O' Marrett, L. McBride, H.C. McPherson, R.W. Peterson, T.H. Pigford, T.B.
j Taylor, and A.D. Trunk, The Need for Change: The Legacy of TMI, President's Comission on the Accident at %ree Mile Island (Washington, D.C.) , October 1979.
~
'O H.W. Lewis, R.J. Budnitz, H.J.C. Kouts, W.B. Loewenstein, W.D. Rowe, F.
von Hippel, F. Zachariasen, Risk Assessment Review Group Report to the
, U.S. Nuclear Regulatory Comission, U.S. Nuclear Regulatory Commission,
, (Washington, D.C.), NUREG/CR-0400, September,1978.
1'
'O C.A. Pelletier, C.D. Romas, Jr., R.L. Ritzman, & F. Trooper, Iodine-131 j Behavior During the TMI-2 Accident, Science Applications, Inc. (Rockville,
, Naryland), prepared for the Nuclear Safety Analysis Center, NSAC-30, final j report, September 1981.
I RASNUSSEN, ET AL., 1975 O N.C. Rasmussen, et al., Reactor Safety Study: An Assessment of Accident Risks in U.S. Comercial Nuclear Power Plants, U.S. Nuclear Regulatory
. Comission (Washington, D.C.) , WASH-1400 (NUREG-75/014), October 1975.
I ROGOVIN, ET AL.,1980 l M. Rogovin, G.T. Frampton, E.K. Cornell, R.C. DeYoung, R. Budnitz, P.
O Norry, et al., Three Mile Island: A Report to the Comissioners and to the j Public, Special Inquiry Group, U.S. Nuclear Regulatory Commission
- (Washington, D.C.), NUREG/CR-1250, Vols. I and II,1980.
l SILBERBERG, ET AL., 1981 j M. Silberberg, R.R. Sherry, M.A. Cunningham, C.N. Kelber, R.S. Denning,
- O R.P. Wichner, T.S. Kress, R.A. Iorenz, R.M. Elrick, J.T. Bell, R.A.
- Sallach, L.M. Toth, D.O. Campbell, A.P. Malinauskas, J.A. Gieseke, M.R.
- Kuhlman, K.W. Lee, H. Jordan, T.C. Davis, W.F. Pasedag, A.K. Postma, & R.
l Adams, Technical Bases for Estimating Fission Product Behavior Durina LhR Accidents, U.S. Nuclear Regulatory Commission, Battelle Columbus
! Laboratories, Oak Ridge National Laboratories, and Sandia National
!O Iaboratories (Washington, D.C.), NUREG-0772, June 1981.
i SILBERBERG, ET AL., 1985 M. Silberberg, J.A. Mitchell, R.O. Meyer, W.F. Pasedag, C.P. Ryder, C. A.
Peabody, and M.W. Jankowski, Reassessment of the Technical Bases for Estimating Source Terms, Accident Source Term Program Office, Office of
'O Nuclear. Regulatory Research, U.S. Nuclear Regulatory Commission
- (Washington, D.C.), NUREG-0956, draft report for coment, July 1985.
! STRATION, MALINAUSKAS, & CAMPBELL, 1980 i W.R. Stratton, A.P. Malinauskas, & D.O. Campbell, letter dated 14 August i 1980 to NRC Chairman J. Ahearne.
50 i-O
,0 1-11 i UNICN OF CCNCERNED SCIENTIS'IS,1977 )
Henry W. Kendall, et al., %e Risks of Nuclear Power Reactors, A Review of - ,
the NRC Reactor Safety Study, WASH-1400 (NUREG-75/014) , Union of Concerned I
'O Scientists, (Cambridge, MA), August 1977.
4 WIISON, ET AL.,1985 i R. Wilson, K.J. Araj , A.O. Allen, P. Auer, D. Boulware, F. Finlayson, S.
Goren, . C. Ice, L. Lidofsky, A.L. Sessans, M.L. Shoaf, I. Spiewak, & T.
T *D'*ll ' Report to the American Physical Society of the Study Group on O Radionuclide Release from Severe Accidents at Nuclear Power Plants, draft, February 1985, to be published in Reviews of Modern Physics, 1985.
i
)
- O f
10
- O l
!O lO l
l l
!O i
1
!O t
!O i,
.,,+-__.s..e.., _.r_. _ . . . - , . _ . , _ _, , _ . - . - , . , . - .,._y,.,,..ym_ _,,yy.-_. ,-__m.____.. .. _-,-_-_m., ,m._,,.-_
i f'
I CHAPTER TWO Outline of the NRC Source Term Reassessment Program i: St2EARY I~
l A brief history of the NRC source term effort is presented.
) '
Source terms have historically been estimated piecemeal, relying on limited hand- and caputer calculations. We NHC's source term program. has focused on the develognent of caputer simulations of severe accidents and has resulted in publication of a ntsnber of technical reports, including NUREG-0956, a stmmary of NRC's current views on source terms
} and the subject of the present report.
2.0 Introduction We NRC source term reassessment is a multi-year theoretical and
}
experimental effort of the NRC and the national laboratories to impyove the understanding of source terms for use in the NRC's regulatory program. We results of these efforts have been published in a ntsnber of reports issued by
] the national labs and the NRC itself.
We most recent of these reports, NUREG-0956 [SILBERBERG, ET AL.,
1985], stmmarizes the NBC's present views on source term estimation. We
] publication of NUREG-0956 ends the second phase of the NRC source term reassessment progran. We first phase was stamarized in NUREG-0772, an earlier interim report completed in 1981 [SILBERBERG, ET AL., 1981]. In 1986, the NRC plans to publish another report, NUREG-ll50, that will include a revised estimate of risk for the five source term reference plants (Zion,
]
Surry, Sequoyah, Peach Bottczn, and Grand Gulf) .
We purpose of this chapter is to present a brief overview of the
] progran that has led to the publication of NUREG-0956.
2.1 NRC Approach to Source Term Estimation p In the past, source terms have been estimated in a piecaneal fashion using a limited ntsnber of computer models and hand calculations. Many asstanptions were made in both the modeling of accidents and in the choice of
)
D
3; 2-2 !
l i
physical parameters used in the models. We results of such calculations had _
'O -large uncertainties associated with them, and while the resulting source terms were believed to be " conservative" - that is, more pessimistic than actually expected - no one knew for sure that this was always the case, or by how much. ;
O one of the stated goals of the NRC source term reassessment program, as carried out in the BMI-2104 series of reports by Battelle Colmbus Labora-tories (BCL), was the developnent of caputer models to permit the systematic calculation of source terms in a step-by-step fashion. Most of the camputer O models used in the new source term ' calculations are either entirely new (e.g.,
CORSOR, MERGE, VANESA, SPARC, and ICEDF) or substantially revised versions of previously available codes (e.g. , MARCH 2.0, TRAP-MELT 2, and NAUA-4) . A diagran fra NURm-0956 displays how the various caputer codes are used in O the source term analyses - (Figure 2-1) . me deficiencies of the codes are discussed in Chapter Seven. Accmpanying the development of these codes has been an experimental program exploring the various physical and chemical processes that could occur during severe accidents.
O 2.2 Principal Source Term Reassessment Program Documents NUREG-0956, the stmmary docment in the PRC assesment, focuses pri-O marily on the results of the Battelle (blumbus Laboratories calculations as presented in the BMI-2104 report series (GIESEKE, Pr AL.,1984a-e] . (Battelle also prepared a separate volme of BMI-2104 which addressed the principal cm-ments of peer reviewers of earlier draft versions of the BMI-2104 analyses O (GIESEKE, ET AL. ,1985a] .)
The BMI-2104 reports present source terms that were calculated using the codes identified in Figure 2-1. Source terms were estimated for a snall O n mber of accident sequences at five reference plants. ne plants and accident sequences considered in BMI-2104 are listed in Table 2-1. (te calculations for a sixth plant, Limerick -- a General Electric BWF/4 reactor with a Mark II contairment, were inemplete at the time this report was
'O written, and were therefore not available for review.)
O
O 2-3 According to the NRC, the primary purpose of the BMI-2104 analyses was O "to develop and test a systmatic analytical procedure for source term estima-tion using the insights gained from the (1981] NUREG-0772 review of the state of knowledge concerning fission product release and transport." [SILBERBERG, ET AL. , 1985] However, as discussed later in Gapter Seven, the computer O codes used in these calculations have not been validated against experiments.
W us, in no sense have the analytical procedures been " tested."
NUREG-0956 draws on a nmber of source docments for its conclusions. l O A partial listing is provided in Figure 2-2, which portrays the relationships mong the NRC source term reports in the decision-making process on developing a regulatory schee for severe accidents. As set forth in this figure, the principal NRC source term reports are the following: 1 l
O '
A. BMI-2104, the Battelle Colmbus Laboratories source term evalutions for selected accident sequences at Surry, Sequoyah, Zion, Peach Bottom, and Grand Gulf [GIESEKE, ET AL., 1985a-e}, plus the forthecming report on Limerick.
O B. SAND 84-0410, the so-called " QUEST" program report fra Sandia National Laboratories, docmenting uncertainty / sensitivity analyses for three accident sequences (two at Surry and one at Grand Gulf) (LIPINSKI, ET AL., 1985a-c; LIPINSKI, ET AL.,
1985d].
O C. oRNL/1M-8842, the oak Ridge National Laboratory source term emputer code validation report (KRESS,1985] .
D. NUREG-1979, the report of the Contaiment Loads Working Group
.O E. NUREG-1937, the report of the Contalment Performance Wrking Group (BA(IHI, ET AL., 1985].
F. We review of the source term reassessment program by the 7merican Physical Society (WILSON, ET AL.,1985] .
Were are, in addition to these reports, a host of others with partic-
- ular relevance to the NRC source term reassesment program. Were are as well a nmber of source term calculations referenced in NUREG-0956 which were not subjected to peer review. 7mong the more important of the additional O doc ments are:
1 0
__ - ... - . - - . . . - . . . -. _ . . . - - - - . - _ . ~ . . _ - _ _ - _ _ - _
- D 2-4
- A. . NUREG/CR-3449, the Sandia National Laboratories " Severe Accident Uncertainty Analysis" (SAUNA) report (RIVARD, ET
-O. AL., 1984].
B. An informal report to NRC fra Battelle Cblunbus Laboratories docmenting additional source tem calculations for contain-ment isolation failure and contairment leakage (GIESEKE, ET AL., 1984f].
.O.
C. A series of " supplemental" source tenn analyses (GIESEKE, Er ,
i AL., - 1985b], which, in actuality, fonn the backbone for the preliminary risk reestimation for Surry that is included in 01 apter 6 and Appendix D of NUREG-0956.
iO 2.3 The Surry Risk Reevaluation in NUREG-9956 At the outset, the approach to source tenn estimation in the NRC source.
tenn reasessment program was to have involved source tenn estimation on a
- O plant-by-plant and sequence-by-sequence basis, as was described in the very first volune of BMI-2194 (GIESEKE, EP AL.,1983]:
It is to be recognized that this report describes an analytical
+O
- *PP' 'Ch f * **ti"*ti"9 **di """ lid' "'*"*P ** ""d d*P *iti "
- which incorporates physical and chenical processes on a mechan-
)
istic basis. . tis apprcach is being evaluated for use in pre- i dicting fission product source terms for release to the environ-ment on a _ specific case-by-case basis (reactor, . accident se-quence) and when verified would be expected to replace the generi tabular release fractions such as those in Table 6,
'O i Appendix V, WASH 1400 where release fractions are given for broad
! classes of accidents. [enphasis added]
l It is therefore quite unfortunate, and, we believe, technically i g indefensible, to now revert to the practice of defining " release categories" s
or " plant damage bins" in which source tenn estimates for specific accident se-j quences are used to create generalized source terms to represent a nunber of different accident sequences. Wis is precisely what was done in the Reactor
- priate. We use of this technique in Chapter 6 and Appendix D of NUREG-0956 i is also inappropriate. Moreover, as will becme evident from later chapters of this report, many of the accident sequences that pose the greatest risk to g' the public were not included in the source term calculations presented in l
4
O 2-5 BMI-2104. Thus, even if the computer models accurately represented accident O phenanena, the BMI-2104 calculations would not provide an adequate basis for estimating overall risks.
NUREG-0956 as published in draft form contains a preliminary risk assessment for Surry, as well as a risk canparison with WASH-1400, but similar O
risk analyses for the other reference plants (as well as a definitive analysis for Surry) are put off to the future in a separate docunent to be designated NUREG-ll50. NUREG-ll50 is expected to be released for conment'in early 1986.
O O
O 0
1 l
l O
i
'O i
!O i
!O
O 2-6 2.5 References for Chapter Two O BACCHI, ET AL.,1985 1- G. Bagchi, et al., Contaiment Performance Working Group Report, Office of Nuclear Reactor Regulation, U.S. Nuclear Regulatory Commission (Washington, D.C.) , NUREG-1937, draft report for coment, May 1985. ,
! GIESEKE, er AL.,1983 -
- O J.A. Gieseke, P. .Cybulskis, R.S. Denning, M.R. Kuhlman, and K.W. - ke, Radionuclide Release Under Specific LWR Accident Conditions: Volme 1 -
i- PWR-Large Dry Containment Draft Report, Battelle Colmbus Laboratories (Colmbus, Ohio) , BMI-2104, Volme 1, draft report for ccament, July 1983.
- Q J. A. Gieseke, P. Cybulskis, R.S. Denning, M.R. Kuhlman, K.W. We, and H.
[
Chen, Radionuclide Release Under Specific LWR Accident Conditions,
! Battelle Colmbus Laboratories (Colmbus, Ohio) , BMI-2104, draft, vols.
II-VI, July 1984.
- O J.A._ Giesehe, H. Chen, P. Cybulskis, R. Freenan-Kelly, M.R. Kuhlman, and K.W. Lee, Informal Report on Source Term Predictions for Various Containment Failure Mode Assumptions, Battelle Columbus Laboratories (Colmbus, Ohio) , 29 August 1984.
1
! GIESEKE, ET AL., 1985a J.A. Gieseke, P. Cybulskis, R.S. Denning, M.R. Kuhlman, and K.W. ke ,
lQ Radionuclide Release Under Specific LWR Accident Conditions: Volume VII, Response to Peer Review Comments, Battelle Columbus Laboratories i (Colmbus, Ohio) , BMI-2104, Volme VII, draft, February 1985.
!O J.A. Gieseke, R.S. Denning, R.o. Wooten, R. Freeman Kelly, K.W. Ime, and P. Cybulskis, Supplemental Source Term Analyses, Battelle Columbus Laboratories (Colmbus, Ohio) ,19 July 1985.
KRESS, 1985 ,
l T.S. Kress, ca piler, Review of the Status of Validation of the Computer Codes Used in the Severe Accident Source Term Reassessment Study
- O (BMI-2104) , Oak Ridge National Laboratory (Oak Ridge, Tennessee) ,
I ORNL/m-8842, April 1985.
(
, R.J. Lipinski, D.R. Bradley, J.E. Brockmann, J.M. Griesmeyer, C.D. Imigh,
!Q K.K. Murata, D.A. Powers, J.B. Rivard, A.R. Taig, J. Tills, and D.C.
I Williams, Uncertainty in Radionuclide Release Un:ler Specific LWR Accident l Conditions, Sandia National Laboratories (Albuquerque, New Mexico) ,
! SAND 84-0410, Vols. 2-4, February-July 1985.
P i
!O
- l. - - . - _ . , , _ . .
) 2-7 LIPINSKI, ET AL., 1985d R.J. Lipinski, P.K. Mast, D.A. Powers, and J.V. Walker, Uncertainty in Radionuclide Release Under Specific LWR Accident Conditions, Volume I:
] Executive Stmmary, Sandia National Iaboratories (Albuquerque, New Mexico) ,
SAND 84-0410, Vol.1, my 1985.
RASMUSSEN, ET AL., 1975 N.C. Rasmussen, et al., Reactor Safety Study: An Assessment of Accident
] Risks in U.S. Comercial Nuclear Power Plants, U.S. Nuclear Regulatory Comission (Washington, D.C.) , NUREG/75-014 (NASH-1400) , October 1975.
RIVARD, ET AL. ,1984 J.B. Rivard, V.L. Behr, R.G. Easterling, J.M. Griesmeyer, F.E. Haskin, S.W. Hatch, A.M. Kolaczkowski, R.J. Lipinski, M.P. Shennan, A.R. Taig, and g A.J. Wickett, Identification of Severe Accident Uncertainties, National Laboratories (Albuquerque, New Mexico), NUREG/CR-3440 Sandia (SAND 83-1689), September 1984.
SILBERBERG, ET AL., 1981 M. Silberberg, R.R. Sherry, M. A. Cunningham, C.N. Kelber, R.S. Denning, g R.P. Wichner, T.S. Kress, R.A. Lorenz, R.M. Elrick, J.T. Bell, R.A.
Sallach, L.M. 'Ibth, D.O. Campbell, A.P. Malinauskas, J.A. Gieseke, M.R.
Kuhlman, K.W. Lee, H. Jordan, T.C. Davis, W.F. Pasedag, A.K. Postnia, and R. Adams, Technical Bases for Estimating Fission Product Behavior During LWR Accidents, U.S. Nuclear Regulatory Ccmission (Washington, D.C .) ,
NUREG-0772, June 1981.
O SILBERBERG, ET AL.,1985 M. Silberberg, J. A. Mitchell, R.O. Nyer, W.F. Pasedag, C.P. Ryder, C. A.
Peabody, and M.W. Jankowski, Reassessment of the Technical Bases for Estimating Source Terms, Accident Source Term Program Of fice, Office of Nuclear Regulatory Research, U.S. Nuclear Regulatory Commission (Washington, D.C.) , NUREG-0956, draf t report for ccmnent, July 1985.
3 SPEIS, ET AL., 1985 T.P. Speis, et al., Containment Loads Working Group Report, Office of Nuclear Reactor Regulation, U.S. Nuclear Regulatory Commission (Washington, D.C.) , NUREG-1979, draf t,1985.
WILSON, ET AL.,1985 R. Wilson, K.J. Araj , A.O. Allen, P. Auer, D. Boulware, F. Finlayson, S.
Goren, C. Ice, L. Lidofsky, A.L. Sesscms, M.L. Shoaf, I. Spiewak, and T.
Tombrello, Report to the American Physical Society of the Study Group on Radionuclide Release f rom Severe Accidents at Nuclear Power Plants, draft, g February 1985, to be published in Reviews g Modern Physics, 1985.
i
)
l l
O O O O O O O O O O -O TABLE 2-1 Accident sequences and plants analyzed using the BMI-2104 analytical procedure Accident Sequences Analyzed Reactor Containment BMI-2104 Accident
- RSS Plant Type Design Volume Sequence
- Identification **
l'. SURRY PWR Large, Dry 1 and V - Large LOCA (PWR) AB Subatmospheric (Recalculations) - Station Blackout TMLB' Small LOCA SD2 Containment Bypass LOCA V
. Loss of Decay Heat Removal TW c'n
- 3. GRAND GULF BWR MARK III III ATWS TC Loss of Containment Cooling TPI Loss of Makeup Water TQUV Small LOCA (ECCS Failure) 5E2 ,
- 4. SEQUOYAH PWR Ice Condenser IV Small LOCA (Delayed ECCS) 52HF Loss of Core Cooling TML Station Blackout TMLB'
- These descriptive labels are not complete descriptions of LOCA = Loss of Coolant Accident the accident sequences analyzed. Tables 4.2-4.6 provide ATWS = Anticipated Transient additional explanation of the sequences.
- These letters refer to a complete description of the ECCS = rgenc o e Cooling accident sequences as defined in the Reactor Safety Study System (WASH-1400).
O 2-9 FICURE ?. - 1 o Fission Product Thermal Hydraulic Transport g Behavior ;
F------------- 7
! ORIGEN MARCH O
Fission Product Inventory in Fuel 1
' Overall l Behavior of El
- h=
l Reactor Coolant System, Molten Core, and
=1a
'O flI I Containment l
g CORSOR l 3l I l I vl 8 Release from Fuel l
Retained in Fuel l l l s
!I I l I
- O SI I I I I - i =
TRAPMELT 5 i MERGE i
I i 8 I i l Reactor Coolant System .. Detailed Temperatur,e, 3 Transport and e Pressure, and Flow in g Retention B ! Reactor Coolant System l " I
- O l l L____ _ _ _ _ _ _
____a l I i l l l CORCON l O VANESA y
-l m S Release from -- Detailed Core- -
5l
@ $ Core Concrete Melt . Concrete Temperature 3l C
4 l and Interactions al Y1 lO gl , , , ,
j j NAUA, SPARC, ICEDF l
l Containment Transport and Retention l g
l
.O I I L _ _ _ __ __ _ _ _ _ _ ___ _ ;
9 Release of fission products to the environment: Source Term
!O BMI-2104 suite of codes as used in the source term reassessment.
I S ource: NUREC-0956
- O
! Source Term Reassessment Program ,
Relationships I
i ResearCh 8 Regulation Source Term Analysis l l BMI-2104 i I I
i Uncertainty Estimates l
SAND 84-0410 m s
O 1
Status of Validation W u
- ORNIJTM-8842 M
4 o
- Source Term ChanDes in Reassessment Study C Source Term-Based y 8 "'" "*
l Containment Loads s i Working Group n NU R EG-1079 q
i Containment Performance l Working Group l
N U R EG-1037 I.
1
- l 1 [ APS Review Severe Accident Severe Accident i { Risk Rebaselining/
! Risk Reduction [ Regulatory Implementation Program (SARRP) l I
I
-l I
j Relationships of the NRC Source Term Reassessment Program with regulatory implementation.
1 i
.)
I _ _ _ _ ____ __ _ _______ _ _ _ _ _ _ ___-_______
CHAPTER THREE
>J l
Probabilistic Risk Assessment - Source Terms in a Risk Context
SUMMARY
Source tenas are used by the NRC for a nmber of regulatory "i purposes: to establish exclusion areas around nuclear plants; to develop standards of performance that electrical equipment must meet during accidents; and, using the techniques of probabilistic risk assesment (PRA), to provide essential input to the calculation of the offsite consequences of severe nuclear accidents. PRA is the analytic tool presently D used to estimate the overall risks posed by nuclear power plants; source terms are one of the most important elments in performing PRAs.
The probles of calculating accurate source terms form the subject matter of the following chapters of the present O report. Chapter 3 describes some of the generci problems inherent in the use of PRA. Wese probles ultimately sts frcxn the enormous emplexity of nuclear power plants and the consequent myriad ways in which they can fail.
Some of these probles are as follows: the virtual impossi-O bility of identifying all important accident sequences that can occur at particular plants; inadequate data on the failure rates of ccxnponents; mcertainties in predicting operator behavior; the difficulty in taking pocr construction techniques into account; a lack of understanding of the many physical and chmical processes that would take place in U serious accidents.
These and other methodological probles mean that there are very large uncertainties associated with PRA predictions and that overall "bottcxn line" estimates of nuclear risks should not be taken seriously.
3.0 Introduction Probabilistic risk assesment (PRA) involves consideration of both the g probability and the consequences of possible accidents. After the source term for a particular accident sequence is estimated, it is used as an input to the estimation of the consequences of that accident. We consequences are then ecxnbined with the estimated probability of that accident sequence to yield an g estimate of the risk posed by that accident. Source term estimates for spe-cific severe accidents can also be used for a variety of purposes without regard to the probability of the accidents. For example, the NRC's siting O
)
3-2 regulations at 19 C.F.R. Part 100 utilize the so-called " TID source term" to establish exclusion area and low population zone distances for power reactors l
[DiNUNNO, EP AL., 1962]. In most applications, however, source terms will be used in a context requiring explicit consideration of the likelihood of the accident sequences and containment failure modes which give rise to the source
! terms.
i O
l
'Ihe identification of specific accident sequences and calculations of their likelihood and consequences are the principal products of probabilistic risk assessment. Wis chapter briefly discusses PRAs and highlights some of their well-known deficiencies in assessing the risk posed by the operation of ecmnercial nuclear power plants. Wese deficiencies, which rendered the Reactor Safety Study inadequate as a decision-making tool, have not been overcme. As a result, PRA cannot be used as a basis for relaxing safety O
requirements.
3.1 The Probabilistic Risk Assessment Process O
Figure 3-1 depicts the major tasks in a probabilistic risk assessnent (PRA) for a nuclear power plant. A probabilistic risk assessment begins with the postulation of an initiating event - an accident (such as a loss-of-coolant or steam line break accident) , a transient (such as turbine trip or O
loss of offsite electrical power), or an external event (such as a fire or an earthquake). We process continues with the identification of possible failures (including haan actions) involving the safety systens and structures designed to mitigate the initiating event. Wis process, which identifies the O
types of accident sequences which are possible for a particular plant design, is referred to as event tree analysis.
Bef re the probability of each possible failure identified in the O
accident sequence or event tree can be estimated, it is first necessary to define the system success criteria or minimm performance required of each safety system to mitigate the accident. For exanple, the nmber of cooling water peps that must operate or the maximm time available to initiate pmp O
O 3_3 l
E operation is determined by the particular accident postulated. tus , the
. probability of failure of a safety system will vary depending upon the O initiating event being analyzed.
Estimating the probability that a safety systen will fail to satisfy the i systen success criteria involves an attempt to identify all of the ways in ,
- O which the safety systen can fail and to assign a probability to each possible failure. In addition to failures in the safety system itself, all potential failures in essential support systens (such as electrical power, lubrication and cooling systens) must also be considered. An attempt must also be made to
.O identify ccanon-mode failures (single failures affecting multiple systems),
and system interactions (failures in one systen which result in failures in other systens) .
10 We probability of each failure mechanism in the fault tree is assigned based on data fra operating plants, experiments, and/or the judgment of the
! analyst. We probability of a particular failure can vary with the accident being analyzed. For example, 'if an electrical caponent is subnerged in water 40 in one accident sequence but not another, its probabilty of failure may depend on which accident is under consideration. i i
j We need to identify all possible failure mechanisms for every caroponent
!O and to assign an accurate probability to each identified failure mechanisn is an extremely difficult, if not impossible task. Wis is a major drawback of PRA that renders reliance on its results as the sole or primary basis for safety decisions inappropriate.
'O We next consideration is the timing and physical progression of events in the accident sequence, such as core uncovery, fuel melting, and the failure of the reactor vessel. An attempt is made to identify all of the possible (O phenmena that can occur in a severe accident, and to estimate the impact of these phenmena on the release of fission products from the core, the transport of fission products through the reactor coolant systen and the contalment, and the possible mechanisms of contaiment failure. In each of O these areas there are major analytic anl scientific uncertainties which empound the problems of equipnent failure describe above.
O
. l
O 3-4 Accident sequences that are believed to have similar accident progression behavior are often grouped into accident sequence classes or plant damage i O states. his is done to reduce the neber of accident sequences that would otherwise have to be analyzed in detail. However, this grouping or "binning" of accident sequences that appear, without detailed analysis, to be similar can result in the failure to identify accident sequences that pose a O significantly different risk than those with which it is grouped. In fact, since the models currently used to analyze accident progression behavior have been shown to be extremely sensitive to plant-specific design details and to the specific accident sequence analyzed, significant errors can be introduced
.O by grouping accident sequences into plant damage states.
We next step in a PBA is to establish the modes and timing of contaiment failure for each accident sequence or sequence class. Wis n" portion of the PRA analysis involves construction of a contairment event tree which tracks containment phenomena and their resulting loads on the contalment to estimate when and by what mechanims the contalment might fail. We contaiment event treee usually consists of a series of yes/no O questions, and is quantified by using emruter models, hand calculations, separate engineering studies (e.g., ultimate contaiment strength under static and dynanic loads), and judgments by the analysts.
O ne output of this task is a listing of accident sequences (or sequence classes) with their contalment failure modes, and the frequencies of these sequences. @e timing of contalment failure, the mode of containment failure, and the physical phenmena accmpaning the accident sequence all O serve to determine the source term for the accident sequences (or sequence l
classes) . We source tems are then used as input into accident consequence calculations.
O Historically, source terms have not been estimated for each accident sequence / containment failure mode combination. Rather, the individual accident sequence /contalment failure mode cmbinations were grouped into release categories to reduce the n mber of separate accident consequence O calculations which must be perfomed. We frequency of the release categories
!O
.O I 3-5 is determined by adding the frequencies of the accident sequence /contaiment failure mode canbinations which contribute to the release category.
O 3.2 PRA Studies Completed to Date O More than twenty PRAs of comnercial nuclear power plants have been completed covering plants in the U.S. Wese studies were undertaken in a variety of NRC programs as well as at the initiative of a number of utilities.
t e various types of PRAs are briefly discussed below.
O 2e first large-scale PRA of nuclear power plant operation was the 1975 Reactor Safety Study (RSS or NASH-1400) [RASMUSSEN, ET AL. , 1975]. We RSS performed PRA sttriies of two nuclear power plants: Surry Unit 1 (a 3-loop
- O Westinghouse PWR with a large dry subatmospheric contaiment) and Peach Bottom Unit 2 (a General Electric BWR/4 with a steel Mark I contaltstent) .
The RSS was extensively criticized in both draft and final form.
g Critical reviews of the RSS include those by the Anerican Physical Society
[LENIS, ET AL., 1974), the U.S. Envirornnental Protection Agency (EPA, 1975],
the NRC-sponsored " Risk Assessnent Review Group" [ LEWIS, ET AL. , 1978], and the Union of Concerned Scientists [KENDALL, ET AL.,1977) . In January 1979, 9 in the wake of these reviews, the Comnission fonnally stated that it no longer regarded "as reliable the Reactor Safety Study's numerical estimates of the overall risk of reactor accident."
O Shortly after publication of the Reactor Safety Study in 1975, NRC instituted a program called the " Reactor Safety Sttriy Methodology Applications Program" or RSSMAP to extend RSS-style analyses to additional nuclear power plant types. We RSSMAP analyses were abbreviated RSS-style PRAs of limited
- g scope and detail and were an attempt to determine whether important plant-to-plant differences could affect risk.
Following the March 1979 accident at the tree Mile Island Unit 2 iO reactor, the NRC's "Wree Mile Island Action Plan" [NRC, 1980; NRR, 1980) identified the need 'or additional risk assessment studies. We " Interim Reliability Evaluation Program" or IREP was created to identify particularly
- O
O 3-6 high-risk accident sequences at individual plants and to determine regulatory O
initiatives t reduce these high-risk sequences. As with the RSSMAP sttx3ies, the IREP PRAs were not intended to be absolute estimates of risk [ TAYLOR &
MURPHY, 1979].
Finally, operating utilities and nuclear industry groups have sponsored a O
large neber of PRA studies for a variety of purposes. Table 3-1 provides a ,
list of the cctnpleted PRA studies and the estimated core melt frequency where available. Many of these PRAs have had NRC-sponsored reviews; where the O
reviews provided estimates of core melt frequency, these results are also provided.
We core melt probabilities, when viewed in aggregate form, suggest that O the average core melt frequency for U.S. reactors generally may be of the
-4 per reactor year, or about 1 chance in 3,300 per reactor order of 3 x 10 year. See reactors will have higher and lower core melt frequencies than this rough average, but the average implies that there is about a 45% chance f at least ne severe ac ident within the next 20 years, and about a 10%
O chance of two such accidents within 20 years.
3.3 PRA Uncertainties and Sensitivities O
Some of the factors which contribute to uncertainties in PRA results are discussed below. See of these factors affect only the likelihood of accident sequences, while others affect the progression of the accident by introducing new phenmena r altering the time history of the accident sequence. However, O
all argue against placing heavy reliance on any PRA quantitative results.
3.3.1 Aging of Structures and comoonents b
Most of the information in a PRA data base comes from operating experience of deestic nuclear power plants. Much of the data base used for recent PRAs, while it reflects data frcm a steadily increasing ntsnber of O
plants, e nes frm operating experience in the early part of a plant's life.
However, some types of failures can be expected to be related to aging of structures and equipnent. Wese failures are not adequately represented in O
~ - - _. - -- . - _-
O. '
3-7 the existing data base. Sus, estimates of failure probabilities made .using this data base may underestimate failure rates.
O 1b illustrate the nature of the problem, the Indian Poin't Probabili::itic Safety Study (IPPSS), which was empleted in 1982, collected operating data
- O fr m 31 dmesti PWRs covering 131 reactor-years of experience to establish initiating event probabilities. Of these 31 plants, only seven had operating experience of more than five years in the data base, and only two had operating experience of more than ten years. %us, a little over 62 percent f the t tal data base used in the IPPSS for Indian Point . Unit 2 initiating O
event probabilities represented operating experience of five years or less (about one-fourth represented operating experience of three years or less) .
Plant aging can also affect sei nic risk analyses. Camponent fragilities O
are based on design information. A caponent such as a heat exchanger can experience. aging effects which lead to thinning of tube walls. %us , a realistic seismic risk analysis should account for this decrease in structural strength with time. Such factors have not been accounted for in seimic risk O
analyses to date.
3.3.2 Technical Specification Violations and Tenporary Exernptions
- O PRAs are performed on the assumption that the utility will only operate y the plant within the bounds set by the technical specifications, which form part of the plant operating license and set safety limits and limiting
- g conditions for plant operation. Wis assunption can affect the quantitative l results of a PRA, but the magnitude of the effect is usually not detemined.
For exmple, in the Grand Galf PRA, all accident sequence " cut sets" (minim m cmbinations of specific equipnent failures and hman errors resulting in a g given accident sequence) which violated the technical specifications were culled fra the PRA. An NRC-sponsored progra evaluated the impact of this practice and found that it resulted in reducing the probability estimates for eight of the top nine accident sequences identified in the Grand Gulf PRA by factors ranging fra 1.4 to 4.7 (CATHEY, ET AL.,1984) .
i O
l l l
'O 3-8 It can scarcely be disputed that violations of technical specifications occur. An examination of NRC inspection reports ard enforement files shows that such violations are rather cmmonplace. In addition to violations of technical specifications (whether accidental or deliberate), there is also the matter of temporary exmptions to technical specifications granted by the NRC, often on a one-time, " emergency" basis. A review of the NRC's Federal O Register notices indicates that the agency routinely grants " emergency" relief fra technical specifications that would, absent the relief, require plant shutdown. Te assme, as PRAs do, that plant operation is always in accordance with the requirements of the technical specifications is not realistic and O
clearly causes overoptimistic results.
3.3.3 The Effects of Assumed Levels of Operability on PRA Predictions O
In a 1982 conference paper, Dr. Peter Cybulskis of Battelle Colmbus Laboratories exmined the effects of success criteria, and assmed levels of operability, on PRA results. System failures are almost always modeled in PRAs as complete failure, whereas in reality there are a wide variety of O partial syste failure modes that are also possible. We complete failure asseption is one of analytical convenience since it would be very difficult to model and analyze the range of partial failures which might occur. At the see time, however, the assmption of emplete failure (or emplete success)
O affects accident sequence definition, which forms the basis for analysis of plant response to a particular accident.
Dr. Cybulskis pointed to numerous examples of how assmed levels of O operability can affect PRA results. In both PWRs and BhPs, low pressure injection systems will not be able to function in accident sequences which involve high pressure. Mien the reactor vessel fails, however, reactor pressure quickly drops and low pressure injection systems may initiate,
-O dumping large quantities of water through the reactor vessel and onto the hot core debris. Consideration of such systs behavior could lead to different behavior than when such operation is ignored. Wis is particularly signifi-cant in the analysis of PWRs with the so-called " dry cavity" design where the O injection of large amounts of water onto the molten core could significantly affect the nature of the accident.
O
3-9 l
The timing of the switchover of ECCS systems from injection to recirculation in PWR LOCA sequences can be draatically affected by whether only one pump or more than one pmp operate, and whether the contairunent l
sprays also operate at the sane tine. Het only will the timing of key events in the core melt progression be affected by such considerations, but the probability of operator errors during switchover (for plants without autmatic
) switchover) will be affected by the time available for such actions.
Another example, and one which has important source term implications, involves PWRs with large dry contaiments which have both contaiment sprays
) and fan coolers. In mall LOCAs, the combination of the operation of fan coolers and heat absorption by structures in the contaiment may be sufficient to keep the contaiment pressure below the actuation setpoint of the contaiment spray systen. Contaiment pressure will suddenly increase at the time of reactor vessel failure (particularly if the sequence is a high pressure sequence) , but there are (as discussed in Chapter Nine) phenmena which could result in contaiment failure at this tine. 'Ihis could totally negate the impact of containment sprays on the source term, even though in
) theory they were available for operation if manually actuated by plant operators (CYBULSKIS, 1983]. 'Ihese and other exanples are largely ignored in PRAs.
J 3.3.4 Containment Strength and Failure Modes Contaiment strength is a key parameter in both PRAs and in source term estimation. Source terms for specific accident sequences are greatly affected
) by the assuned failure pressure of the contaiment.
Failure pressure estimates can vary significantly, depending on the nature of the analysis. For example, the Zion contaiment was assessed in the
) Zion PRA as failing at 134 psig due to hoop steel yielding in the middle cylindrical section of the contalment. A more recent analysis by Brookhaven National Laboratory assessed the failure of the Zion contalment at 111 psig j due to loss of shear , capacity at the basenat/ cylinder junction (located underground) [SHARMA, WANG & REICH, 1985].
)
1
1
- O l 3-10
- The probability of containment failure assigned in evaluating a j
'O. particular accident sequence may be significantly different depending upon which of these two predictions of contaiment failure pressure is believed. :
For example, in the Battelle source term analysis of the station blackout !
sequence, contairment pressure was calculated to rise to abcut 99 psig.
jO Relying on the Zion PRA eeimate of 134 psig for contalment failure, Battelle assmed that the contairment would not fail in a station blackout sequence.
(GIESEKE, ET AL., 1984f]. However, a peak contairment pressure of 99 psig is much closer to 111 psig contairment failure pressure predicted by Brookhaven.
jd Consequently, there would be a greater likelihood that the contaiment would fail during a station blackout sequence if Brookhaven's estimate is accepted '
than if the Zion PRA contaiment failure pressure is accepted.
O contaiment failure- pressure analysts also differ on the uncertainty of their estimates. - Capare the results set forth in Figure 3-2 [CYBULSKIS, 1982]. We failure pressure curve for Zion is markedly different fra the other curves presented and indicates very little uncertainty about the
!O pressure at which the Zion contairment will fail. te other three analyses-indicate a wider band of pressure over which the probability of contairment failure goes from negligible to certainty.
J
!g In addition to these factors, containment failure assessments are to some extent dependent on the accident sequences and phenomena selected for examination. For exmple, the Grand Gulf PRA mitted contaiment steam ;
explosions, whereas the RSS analysis of Peach Botta and the Limerick PRA iO included such phenmena. Similarly, EWR PRAs that exclude steam generator tube ruptures fra consideration have not evaluated accident sequences that pose a contalment bypass scenario. Steam generator tube failures can occur as initiating events or as consequential failures in other accident sequences
- O and result in a cmpletely different mode of contalment failure and a different source term (see Chapter Four) .
- Another deficiency in the assesment of contairment failure probability
.O is that current estimates are based on the asseption that the structure has 4
been built in conformance with design drawings. Failure to accotrit for the
] possible flaws in fabrication and construction, deterioration due to aging,
- O 4
O 3-11 design errors (RASMUSSEN, ET AL.,1981], or degradation due to external events n or enviromental conditions would result in overestimates of contaiment
" strength.
As the source term reassessment has shown, the sooner a contaiment fails, the larger the source tem. Clearly, a better basis is needed for O predicting such a crucial parameter as contaiment failure pressure.
3.3.5 Influence of Human Actions O
ne Nac source term reassessment program evaluated accident sequences assmirg a " hands off" approach - the operators are assmed to take no action after core melting begins. This is certainly a convenient analytical assmption, but in reality operator actions will be part of almost any O accident sequence.
' PRAs also consider operator actions that can initiate or aggravate accidents, but usually only up to the point of core danage. Consideration of O operator recovery actions is also coumon and this results in significant reductions in the estimated frequency of accidents involving core melting.
The review of the source term issue sponsored by the nuclear industry,
~O (IocoR), considered sme such operator actions, but focused almost entirely on those operator actions which mitigate accidents. Were has been no similar effort to identify those operator actions which could aggravate accident sequences.
O Operator actions and caissions can have a substantive effect on the course of an accident, but this matter has not been addressed adequately in the NRC source term reassessnent program.
.O O
!O
3-12 3.4 References for Chapter Three i
O BERRY, ET AL., 1984 D.L. Berry, N.L. Brisbin, D.D. Carlson, R.G. Easterling, J.W. Hicianan,
- - A.M. Kolaczkowski, G.J. Kolb, D.M. Kunsman, A.D. Swain, W.A. Von
- . Riesenann, R.L. Wodfin, J.W. Reed, and M.W. McCann, Review and Evaluation
- of the Zion Probabilistic Safety Study, Sandia National Laboratories (Albuquerque, New Mexico) , NUREG/CR-3300, Vol.1 (SAND 83-lll8), May 1984.
4 CATHEY, ET AL., 1984 N.G. Cathey, E.A. Krantz, J.P. Poloski, B.J. Shapiro, and W.H. Sullivan, Catalog of PRA Dominant Accident Sequence Information (Draft), EG&G Idaho, ,
Inc. (Idaho Falls, Idaho) , NUREG/CR-3301 (EGG-2259) , draf t, August 1984.
!O- CHELLIAH, ET AL., 1984 E. Chelliah, J.Meyer, J. Carter, A. Acharya, L. Reiter, R. Bari, H.
Ludewig, I. Papazoglou, J. Boccio, M. PtCann, and A. Kafka, Review Insights on the Probabilistic Risk Assessment for the Limerick Generating i Station, Office of Nuclear Reactor Regulation, U.S. Nuclear Regulatory Comnission (Washington, D.C.) , NUREG-1968, August 1984.
- O
. CHOW & DCH,1984 R. Buch, Safety Evaluation Report by the Office of Nuclear E. Chow and Reactor Regulation, Consumers Power Company, Big Rock Point Plant, Docket No. 50-155, Office of Nuclear Reactor Regulation, U.S. Nuclear Regulatory
% Cmenission (Washington, D.C.) ,17 May 1984.
v CYBULSKIS, 1982 P. Cybulskis, "On the Definition of Containment Failure," Battelle Columbus Laboratories (Columbus, Ohio), 1982 in W. A. Sebrell, ed.,
Proceedings of the Workshop on Contaiment Integrity, Sandia National Laboratories (Albuquerque, New Mexico) , NUREG/CP-0033 (SAND 82-1659) , Vol.
- O-I, October 1982.
i
. CYBULSKIS, 1983 P. Cybulskis, " Sensitivity of Degraded Core Cboling Accident Predictions to the Assuned Levels of Operability of Engineered Safety Systems,"
Battelle Columbus Laboratories (Colunbus, Ohio) , in Proceedings of the
- O- -
International Meeting on Thennal Nuclear Reactor Safety, U.S. Nuclear Regulatory Comnission (Washington, D.C.) , NUREG/CP-0027, Vol. 3, February
- 1983.
1
( DiNUNNO, ET AL. ,1962 J.J. DiNunno, F.D. Anderson, R.E. Baker, and R.L. Waterfield, Calculation
- O
=
of Distance Factors for Power and Test Reactor Sites, Division of
- -Licensing and Regulation, U.S. Atomic Energy Commission (Washington, D.C.) , TID-14844, 2nd printing, 23 March 1962.
EPA, 1975 U.S. Envirorsnental Protection Agency, Reactor Safety Study (MSH-1400): A
}O Review of the Draft Report, U.S. Envirorsnental Protection Agency (Washington, D.C.) , August 1975.
i i
iO
. ,__..__ ,__..- ___ _ - ,_ . _ ,, . - . _ _ _ , , . _ . _ - . , , . ~ m _ _ . . _ . , . _ - - _ _ - - _ . _ _ _ _ _ _ _ _ _ _ _ . _ . _ _ _ _ _ _ _ _ _ _ _ ,
- O 3-13 1
GARRICK, ET AL., 1980 B.J. Garrick, et al., paper on Oyster Creek Probabilistic Safety Analysis O (OPSA), 1980, in Proceedings of the American Society / European Nuclear Society Topical Meeting, Thermal Reactor Safety, Volume II, CONF-800403/V-II, 1980.
GIESEKE, ET AL., 1984f J. A. Gieseke, P. Cybulskis, R.S. Denning, M.R. Kuhlman, K.W. Lee, and H.
Cien , Radionuclide Release Under Specific LWR Accident Conditions: Volume O VI, PWR-Large, Dry Containment Design (Zion Plant) , Battelle Colunbus Laboratories (Colunbus, Ohio) , BMI-2104, Vol. VI, draft, July 1984.
- O Nuclear Power Station Probabilistic Risk Assessment (Internal Events and Core Damage Frequency) , Brookhaven National Laboratory (Upton, New York) ,
, NUREG/CR-4050 (BNL-NUREG-51836) , June 1985.
KENDALL, ET AL., 1977 H.W. Kendall, R.B. Hubbard, G.C. Minor, W.M. Bryan, D.G. Bridenbaugh, D.F.
-O Ford, R.A. Finston, J.P. Holdren, T.C. Hollacher, D.R. Inglis, M. Kanins, A.A. Lakner, and J. Primack, The Risks of Nuclear Power Reactors: A Review
~
i of the NRC Reactor Safety Study WASH-1400 (NUREG-75/014), Union of
! Concerned Scientists (Cambridge, Massachusetts) , August 1977.
LEWIS, ET AL., 1974' H.W. Iewis, R.J. Budnitz, A.W. Castleman, D.E. Dorfan, F.C. Finlayson,
'O R.L. Garwin, L.C. Hebel, S.M. Keeny, Jr., R. A. Muller, T.B. Taylor, G.F.
Smoot, and F. von Hippel, Report to the American Physical Society by the Study Group on Light-Water Reactor Safety, 1974, in Reviews of Modern
- Physics, Vol. 47, Supplanent No.1, Sumner 1975.
!O H.W. Lewis, R.J. Budnitz, H.J.C. Kouts, W.B. Iowenstein, W.D. Rowe, F. von Hippel, and F. Zachariasen, Risk Assessment Review Group Report to the U.S. Nuclear Regulatory Comnission, Ad Hoc Risk Assessment Review Group, U.S. Nuclear Regulatory Ccannission (Washington, D.C.) , NUREG/CR-0400, Septenber 1978.
O NRC, 1980 U.S. Nuclear Regulatory Comnission, NRC Action Plan Developed as a Result of the TMI-2 Accident, U.S. Nuclear Regulatory Comnission (Washington, D.C.) , NUREG-0660, Vols.1 and 2, May 1980.
O NRR, 1980 Office of Nuclear Beactor Regulation, Clarification of TMI Action Plan Requirements, U.S. Nuclear Regulatory Comnission (Washington, D.C.) ,
NUREG-0737, November 1980.
NUSCO, 1985 O Northeast Utilities Service Company, Millstone Unit No.1 Probabilistic Safety Study, Northeast Utilities Service Canpany (Hartford, Connecticut) ,
prepared for Northeast Nuclear Energy Company, July 1985. ;
'O
O 3-14 POLOSKI, SULLIVAN & GENTILLON, 1982 J.P. Poloski, W.H. Sullivan, and C.D. Gentillon, A Review of Consumers O Power Company Probabilistic Risk Assessment of the Big Rock Point Plant, EG&G Idaho, Inc. (Idano Falls, Idano) , EGG-EA-5765, April 1982.
RASMUSSEN, ET AL., 1975 N.C. Rasmussen, et al., Reactor Safety Study: An Assessment of Accident Risks in U.S. Comercial Nuclear Power Plants, U.S. Nuclear Regulatory C, Commission (Washington, D.C.) , NUREG-75/014 (KASH-1400) , October 1975.
RASMUSSEN, ET AL., 1981 N.C. Rasmussen, H.J.C. Kouts, S.H. Bush, T.J. Connally, H.G. MacPherson, D. Okrent, L. Squires, and E.L. Zebroski, Reoort of the Reactor Safety Research Review Group, prepared for the President's Nuclear Safety O oversigne Ccumittee (Washington, D.C.) , B. Babbitt, Chaintan, Septenber 1981.
ROWSOME, 1982 Internal NRC menorandun from F. Rowsome to J.F. Meyer,
Subject:
" Damage State Likelihoods for Indian Point," 2 Decenber 1982.
O SHARMA, YANG & REICH, 1985 S. Sharma, Y.K. Yang, and M. Reich, Ultimate Pressure Capacity of Reinforced and Prestressed Concrete Containments, Brookhaven National Laboratory (Upton, New York) , NUREG/CR-4149 (BNL-NUREG-51857) , May 1985.
O TAYLOR & MURPHY,1979 NRC menorandun frau M.A. Taylor and J.A. Murphy to F.H. Rowsome,
Subject:
"IREP - Initial Plant Study," 4 October 1979.
O O
O O
O
O 3-15 TABLE 3-1 Published PRA Studies and Core Melt Frequency Results Core Melt Plant Name Plant Type Frequency Notes
-5 PwR, B&W, large dry cont. 5.0 x 10 h O Arkansas Unit 1 .
Big Rock Point BWR/1, GE, spherical steel cont. 9.8 x 10-4 a
-3 a ,b 1.5 x 10 Browns Ferry Unit 1 BWR/4, GE, Mark I cont. 2.0 x 10-4 h
-4 h Calvert Cliffs Unit 1 PWR, CE, large dry cont. 1.3 x 10 O
Calvert Cliffs Unit 2 PWR, CE, large dry cont. 4.0 x 10-4 i,1
-4 h Crystal River Unit 3 PWR, B&W, large dry cont. 4.0 x 10 Grand Gulf BWR/6, GE, Mark III cont. 4.0 x 10 i
-4 Indian P int Unit 2 twR, West., large dry cent. 4.7 x 10 a O -4 3.5 x 10 a,f
-4 Indian Point Unit 3 PWR, West., large dry cont. 1.9 x 10 a
-4 a ,g 3.5 x 10 Limerick Units 1 & 2 BWR/4, GE, Mark II cent. 2.4 x 10 -5 a O 9.4 x 10-5 1,e Midland Units 1 & 2 PWR, B&W, large dry cont. 3.1 x 10-4 a
-4 h Millstone Unit 1 BWR/3, G, Mark I cont. 3.0 x 10 8.1 x 10 n O -$
Millstone Unit 3 PWR, West. , dry subatmos. cont. 6.1 x 10 a
-4 a ,d 1.9 x 10 Oconee Unit 3 PWR, B&W, large dry cont. 8.0 x 10" i
-4 2.5 x 10 a 10- 4.8 x 10 m
Oyster Creek BWR/3, GE, Mark I cont.
Peach Botton Unit 2 BWP/4, G, Mark I cont. 3.0 x 10 ' j
-4 Seabrook Units 1 & 2 PWR, West., large dry cont. 2.3 x 10 a Sequoyah Unit 1 PWR, West., ice condenser cont. 6.0 x 10 ' i Shorehan BNR/4, GE, Mark II cont. 5.5 x 10-5 a,k
-4 a ,k 1.4 x 10 Surry Unit 1 PWR, West., dry subatmos. cont. 6.0 x 10 j
-5 a Zion Units 1 & 2 PWR, West., large dry cont. 5.2 x 10 3.7 x 10- a,e l+
I l
3-16 TABLE 3 - 1 (continued) ,
Published PRA Studies and Core Melt Frecuency Results O
Notes:
a Includes external events.
b NRC-Sponsored review by EG&G, Idaho, Inc., and NRC staff review results O [POLOSKI, SULLIVAN & GENTILLON,1982; CHOW & DCH,1984] .
c NRC-sponsored review by Sandia National Laboratories [ BERRY, Er AL.,1984) .
d NRC staff review [NRR,1984] .
e NRC-sponsored review by Brookhaven National Laboratory and NRC staff review
f NRC staff review estimated core melt frequency at 1.0 x 10- before fixes; after-fix value shown in table [ROWSOME, 1982).
9 NRC staff review estimatd core melt frequency at 6.8 x 10 -4 before fixes; after-fix value shown in table [ROWSOME,1982] .
h Interim Peliability Evaluation Program (IREP) PRA, sponsored by NRC.
i Reactor Safety Study Methodology Applications Program (RSSMAP) PRA, O sponsored by NRC.
j Reactor Safety Study (RSS) PRA, sponsored by NRC.
k NRC-sponsored review by Brookhaven National Laboratory [ILBERG, ET AL.,
g 1984]; includes internal flooding, but no other external events.
1 Original RSSMAP PRA estimated core melt frequency at 1.5 x 10-3 per reactor year. After auxiliary feedwater systen fix, FRA estimated lower core melt frequency which is cited in the table.
m Based n ANS conference paper [GARRICK, EP AL.,1980] .
'O n Recent utility-sponsored PRA [NUSCO,1985) .
l O Source: Union of Concerned Scientists l
<O i
l O
3-17 FIGURE 3-1 O
O CONT AtNME467 AN ALYS13 SYSTEM ANALYSIS
,r
( i I
s gutut. fait g OEvtLOPMENf l I
t anaLTsts CF ANALT5es Or ACCIDENI* AN ALT SIS 08 ma0s04WCLIOC t hvta 0 m u t es t a L settaat EstEnNAL ~
~ St0gimCE
- Qu a u flFIC a fl04 PacCtS5t$ faansP0af Constout%Cil STSitu O .
u008LemC 4 UNCtafaamtT amatt$15 I
ANALTSt$ 08 ,,,,,,,,,
vva att.a..Lif, OtytLOPut NI AND INItarattafl0M OF atSULf 8 ame Pa0Ct0 gats OEv6LOPutu?
O I l l L8,EL i LtvEt : Lt Et
.. . . . . . . . . . . . . . SCOPt SC0pt SCOPE Pm00UCTS em00uCf3 Pn00uCf8 O
Major Steps in a Probabilistic Risk Assessment (PFA) 10 1
.O O
iO
- 0 3-18 FICURE 3-2 O' . .
RSS Surry 1.0 -
_ - - RSSMAP Oconee
' 7 ' p ,1~~~
RSSMAP Calvert Cliffs / ,8 O _
/ / '
. . . . . . . . . . Zion / / ,
/ / e s
0.8 -
/ / '
_ /
/
,/ l
/ '
-o - / '
E / /
'yw 0.6 -
/ , ,'
w 7 i o' - / s
$' / ,
.O /
n 0.4 -
1 / l
$ ~
/
/ /
/ /
/ s' 0.2 -
/
I
.O -l /
s' l
' I i F ' ' --- L"' '
0 1.5 2.0 2.5 3.0 1.0 ,
RELATIVE PRESSURE
'O (Accident. Pressure / Design Pressure)
Cbntainment Failure "ressure Curves (fram CYBUIERIS,1982)
- O l
- o.
e l
i I
1
, I
(
lO
f CHAPTER FOUR O_
Neglected Accident Sequences and Containment Failure Modes:
Internally Initiated Accidents O
SUMMARY
If an accurate estimate of nuclear risks is to be made, all potentially important accident sequences arvi containment failure modes nust be identi fied and their source terms O eval ua ted . This chapter provides examples of. accident sequences and containment failure modes that were inade-quately considered (or omitted altogether) in the NRC's source term assessment program but which could have large source terms and therefore contribute significantly to overall nuclear risks.
O Accident sequences not considered include reactor transients -
in which the reactor fails to scram; the rupture of steam generator tubes; loss-of-coolant accidents (LOCAs) f rom the failure of in-core instrument tubes or seals on the reactor coolant pumps; reactor vessel f ailure f rom pressurized O thermal shock; late recovery of failed safety systems; steam explosions within or outside of the reactor vessel; so-called
" interfacing LOCAs"; and station blackout -(loss of all AC power).
The failure to include impo rtant sequences such as these O-means that the estimated nuclear risks will necessarily be underestimated.
4.0- Introduction O
NRC's Source Term Reassessment Study estimated the source terms for a limited set of accident sequences for each of five nuclear power plant designs. A comparison of these sequences with the results of probabilistic risk assessment (PRA) studies on the five NRC reference plants reveals that a O As a result, any subsequent large number of accident sequences were igrnred.
risk analysis using PRA would necessarily underestimate the actual risk, perhaps by a considerable extent if the associated source terms are large.
O O
~
f 4-2 l
l I The NRC makes no claim that the source term reassessment prog ram l identified and evaltnted all accident sequences that contribute significantly to risk f rom nuclear power plants. Indeed, the limited number of sequences for which source terms have been recalculated by the NRC is far from complete.
If an accurate estimate of risk is to be made, all potentially important accident sequences ani containment failure modes must he evalua ted. This chapter provides exanples of accident sequences and containment failure modes with potentially large source tems that were not considered in the source term reassessment program. gection 4.1 addresses neglected accident sequences arri containment failure modes for pressurized water reactors. Section 4.2 provides a similar discussion for boiling water reactors.
4.1 Neglected PVR Accident Sequences and Containment Failure Modes b
In this section, we highlight examples of PWR accident sequences and containment failure modes that were not examined in the NRC source term reassessment program. The examples given are meant to be illustrative rather
) than comprehensive.
4.1.1 PdR ATdS Sequences
) Anticipated transients without scram (ATAS) are among the most impo rtant accident sequences for pressurized water reactors. For example, the PRA for Calvert Cliffs Unit 1 coreltried that ATVS was the dominant cause of core melt, accounting for about 1/3 of the overall probability of core melt rpAPE, F.T
) AL., 19841. However, ATdS sequences were disregarded in the NRC source term reassessment program.
An anticipated transient without scram (ATWS) is a failure of the reactor
) to shut down (scram) when required following an anticipa ted event such as turbine trip or loss of feedwater. In an A7WS sequence, the nomal capacity for heat removal from the reactor coolant system (RCS) is severely degraded or lost, but the reactor continues to produce a significant fraction of full
) power output. This imbalance between the core power level and heat renoval capacity causes a large, rapid increase in pressure in the reactor coolant system.
)
4-3 l
In some A'IWS sequences, the RCS pressure is predicted to rise well above normal operating pressure (about 2200 psig), posing a threat to the integrity of the RCS pressure boundary and to the functioning of safety system components. We predicted severity of A'IW3 sequences varies among PWR designs.
In general, Westinghouse PWRs seem to be more resistant to an A'IWS
) accident than do Babcock & Wilcox and Combustion Engineering plants. An analysis of an A'IWS initiated by a loss of feedwater, with the reactor at full power, iMicates that the reactor operator has about 00 seconds to manually scre the reactor to avoid serious plant damage from high RCS pressure FLU, ET AL., 19R31. However, analyses of the response of Babcock & Wilcox PWRs to A'IWS events indicate that operator has only 40 seconds to manually initiate a scram. We time is shorter because the amount of water in the secondary side of the steam generators in ' a - B&W plant is smaller than in a Westinghouse -
plant. Thus, the RCS pressure ircrease begirs sooner.
i A complicating factor for both B&W and Combustion Engineering PWRs is
! that during the first 10-40 seconds of a loss-of-feedwater A'IWS event, plant D conditions resemble those of a normal loss-of-feedwater event with successful scram. In addition, operators may not he at the proper control panels fo r rapid response to randomly occurring A'IWS accidents. NRC estimates that this could extend operator response time by 25-30 seconds IMAT'EN, ET AL.,1%31.
l PRA studies to date indicate that the most likely A' INS sequences are 1
those . in which a full complement of containment heat removal systems remains operable. Under such circumstances, it might be postulated that the source
) term would be reduced by operation of the containment spray system and/or the containment atmosphere fan coolers which could reduce aerosol concentration and delay or prevent containment failure . However, ATW sequences are important in the source term context because they could present unusual, and thus far unanalyzed, accident conditions.
l l
- An ATds sequence with a potentially large source term involves reactor vessel rupture caused by the high RC9 pressure following the scram failure.
Failure of the reactor vessel head bolts could result in significant damage to the contalment spray system or to the containment itself f rom the impact of 3
l
O 4_4 the vessel head. A precursor to such an accident has occurred. Several reactor vessel head bolts at the Fort Calhoun plant were Found to be degraded C and one had broken ICmTRELL, ET AL.,19941. Even if the containment does not fall imediately f rom the impact of the vessel head, it could be weakened to the point that it would f ail at a lower containment pressure.
O Studies of ATM sequences have also suggested that the reactor vessel head might lift in response to the high RCS pressure and subsequently fail to reseat. Should this condition persist throughout core mel ting, a direct pathway for fission product transport from the core to the containment would
" exist.
Another A1WS sequence that could have a large source term even without containment failure involves rupture of one or more steam generator tube O"
failures. The high pressure in the RCS could cause the rupture of the steam This would be particularly true generator tubes IAK9TULEWITZ, ET AL., 10R51.
if the steam generator tubes have been weakened as a result of corrosion, a fairly conron condition in operating PWRs. A tube rupture could dramatically O alter the A'IWS sequence because fission products' could be transported through the ruptured tube and released to the environment via the main steam safety valves or the atmospheric dump valves, n
V 4.1.2 9 team Generator Tube Rupture (SGTR) Accidents A steam generator tube rupture (SGTR) can occur either as an initia ting
' event or as a subsequent failure in other accident sequences such as the AT*4 n
v sequences discussed above. Other accident sequences which can involve steam generator tube failures include pressurized thermal shock sequences, loss-of-coolant acciden ts, main steam line break (MSLB) accidents, and transients resulting in steam generator overfill leading to water in the steam lines n" fBARRE1T & THOM, 1985; AKSTULEWITZ, ET AL. , 1985; CLARK, CLAPP, & BROADWA'IER, 19841
'The rupture of one or more tubes in a steam generator would constitute a O loss-of-coolant accident. The principal concern with such an accident is that the release of radioactive materials can hypass the conta inmen t. Fission a
I lO
J 4-5 products released by core melting could travel through the broken steam generator tube aryl be released to the envirnnment outside the containment building through the min steam atmospheric dunp valves or the steam generator safety valves. In addition, coolant water that is lost through the broken tube would not collect in the containment strup and would not be available for reci rcula tion cooling when the water supply for the iniection phase of n" energency core cooling system (ECCS) operation is exhausted. It is possible tha t the borated water storage tank could be refilled, but the O onee PRA assigned a 50% f ailure probability to this recovery action ISUGNET, ET AL.,
19841 O
'Ihrough 1994, there have been four major steam generator tube ruptures in U.S. plants and one in Europe. All of the domestic accidents have involved single tube failures. The Doel Unit 2 plant in Belgium experienced a failure O' of "several tubes" INRC, 19791. There have also been mre than 100 cases involving steam generator tube leakage INSAC,1984; BROOKS, ET AL.,19941.
Despite this history of problems, tube rupture accidents have been excluded f rom NRC-sponsored plant-specific PRAs. However, the NRC staff has completed a draft generic risk assessment which considered single and r:ultiple tube ruptures, both as initiating events and as consequential failures for other accident sequences. The staff estimtes the overall core melt proba-O bility from such accidents to be 4.7 x in" per reactor year IAxSmLew1Tz, ET AL., 19851. The dominant contributors were assessed to be single tube ruptures with a stuck-open ruin steam safety valve and AWS sequences with consegential tube ruptures.
- O Industry-sponsored PRAs have estimted overall core melt probabilities to R x 10 per reactor f rom tube rupture accidents ranging f rom 1.3 x 10 year INORTHEAST UTILITIES,1993; HURRARD, ET AL., NA4; GARRICK, ET AL., 199't; O~ PASNY & CONm,19821. These studies typically claim that the contribution of such sequences to overall core melt probability is about one percent or somewhat mre. However, a draft review of the Millstone Unit 1 PRA by Lawrence Livermore National Laboratory concluded that steam generator tube O rupture was the largest single contributor to core melt probability for that plant fGARCIA, ET AL.,19841.
O
J 4-6 None of the recent source term reviews (NRC, IDCOR, and the American Physical Society) have considered tube rupture segjences. In view of the q
V estimated probabilities and the possibility of containment bypass, there is no justification for the exclusion of tube rupture accidents f rom the NFC source term reassessment program.
O 4.1.3 In-Core Instrument nahe LOCAs EMR reactor vessels are penetrated by a number of tubes containing in-core noni toring instruments. Most PWRs have such penetrations in the O bottom head, although some Combustion Engineering PWRS have top-head penetrations IANDERSON, ET AL.,19R31. Only the Millstone Unit 3 PRA, of all recent PRAs, explicitly considered LOCAs involving rupture of in-core instrument tubes.
O The Millstone analysis was limited to single tube ruptures INORTHEAST UTI LITIES , 19811. A draft review of the Millstone Unit 3 PRA by Lawrence Livermore National Laboratory identified, but did not analyze, the possibility O of cultiple instrument tube ruptures rGARCIA, ET AL.,14R41.
There have been precursor incidents involving leakage f rom in-core instrument tubes, but no actual failure of the tube or penetration has yet O occurred rNSAC, 19941. For Millstone Unit 3, in-core instrument tube LOCAs were found to be important in that they were the only sequence leading to a i flooded reactor cavity at the time of reactor vessel failure (Millstone Unit 3 employs a design which inhibits reactor cavity flooding by means of curbing O around the cavity area -- the so-called " dry cavity" design) .
The Millstone Unit 3 PRA assigned instrument tube IDCAs to a plant damage state very similar to that described in WASH-1400 for a small-b reak LOCA l
O accompanied by containment spray system failure. According to WASH-1400, this results (i f not corrected) in containment failure due to overpressure. After the containment fails, the ECCS recirculation pumps fall because they are not capable of pumping saturated fluid. The lack of continued ECCS operation O results in core melt.
- O
~
4 4-7 J
i The Millstone Unit 3 PRA estimated the probability of the damage state resulting from an instrument tube IDCA at 3 x in- per reactor year INORTHEAST UTILITIES, 19831, but a recent NRC staf f/ Lawrence Livermore National Iaboratory review of the PRA estimates the probability at 2 x 10~
per reactor year ISHERON,19841. 'thic accident coquence has important source term implications because the containment fails before the core melts. If a n
V high pressure melt ejection (see Chapter Nine) or a steam explosion (either in-vessel such that the reactor vessel fails, or ex-vessel) occurs under such condi tions , the resulting source term could be quite large.
- It is possible, as noted in the 0:onee PRA, that operator actions could be taken to O prevent containnent failure (i.e. , the containment could be vented), but such actions generally are not addressed in emergency procedures for PWRs, nor have they been credited in PRA analyses.
O In-core instrument tube LOCAs should be considered as a special case of LTA to ascertain if they present unique conditions not considered with other similarly-sized LOCAs. Their inclusion within the NRC source term reassess-nent program would be particularly immrtant for plants with the so-called
.O ary cavity- design in view of the containment dypass scenario discussed above.
4.1.4 Pressurized 'Ihermal Rhock Accident Sequences
- O Some types of accidents result, at least initially, in overcooling of the reactor coolant system, rather than inadequate heat removal. In such accident sequences, the reactor vessel may cool down, followed by repressurization of the reactor coolant system. If this occurs late in the plant's 1tfe af ter the O reactor vessel has become embrittled, the reactor veesel itself could fail.
Such accident seglences are referred to collectively as pressurized thermal shock (PE) sequences. There are no safety systems designed to cope with failure of the pressure vessel.
lC Considerable research is in progress to assess the overall risk signifi-
)
cance of PE accident segaences. A key issue in the source term context is the contribution of PTS sequences to the likelihood of reactor vessel f ailure.
O Reactor vessel failure can occur due to several causes, but PE might he the dominant cause of PWR reactor vessel rupture for some plants (mostly older O
h 4-8 PdRs which my be more susceptible to PTS as a result of increased neutron irradiation of their vessels and the presence of trace contaminants in welds which enhance their susceptibility to c rack ing) . Recently, it has been recognized that PE seq 2ences could have other outcomes, including steam In generator tube ruptures or an interf acing LOCA f BARRETT r. THOM, 19R51.
l sum, it is not currently possible to assess with confidence the overall risk significance of PTS accident sequences due to the large amunt of unfinished technical work in the area.
l 4.1.5 RCP Seal LOCA Sequences THRs use two to four reactor coolant ptraps (RCPs) to circulate water through the reactor coolant system (RCS). These pumps are equipped with culti-stage seals to limit leakage of primary coolant around the pump shaf t.
The seals are cooled b/ water flow f rom a separate system.
Since the seals have a limited lifespan, leakage or failure of the seals during reactor operation is not uncomnon and constitutes a small loss-of-coolant accident (LOC A) . More than 200 RCP seal leak incidents occurred between 1967 and 1990 IMURLEY, 19911, and there have been at least nine major RCP seal failures in domestic PWRs since.Tuly 1969.#*I
) In addition to occurring as initiating events (small IDCAs), RCP seal failures can occur as a result of other accident sequences (such as transients involving station blackout or failure of the component cooling water system).
NASH-1400 estimated the probability of all small LOCAs in the size range of tiowever, actual PWR
) RCP seal failures at 1.0 x in per reactor year.
operating experience suggests that the probability of RCP seal failure LOCAs alone is about 2x 10~ per reactor year, or 20 times mre likely than all small LOCAs as estimated in WASH-1400. It has been estimated that the effect
) of including RCP seal failures in the WASH-1400 Surry analysis would be to increase overall core melt probability by a factor of about ten IMURLEY,19811.
If the reactor fails to scram following an TCP seal INCA, this would
) result in the equivalent of an ATAS sequence with a praxisting LOCA. Such a seq 2ence may hwe a probability on the order of 6 x 10- per reactor year. As i
b
v O 4-9 noted in the Oconee PRA, scram failures following a LOCA initiator should not present a separate threat to the RCS pressure boundary since the pressure n
v spike following the scram failure starts from a lower initial pressure (ECCS i initiation pressure versus mrmal operating pressure) . Thus, the principal effect, if any, of scram f ailures following a LOCA would lie in the area of timing of events in accident progression and fission product transport.
O On the basis of the probability of RCP seal LOCAs as accident sequences in their own right, as well as their participation in other sequences (ATE and loss of seal cooling sequences), RCP seal failure accidents and sequence O branches should clearly be included in the source term reassessment program.
4.1.6 Reactor Vessel Rupture Sequences i
'O Reactor vessel rupture as an initiating event has been treated in very few PRA studies to date. Yet, such seg2ences are important for several reasons. First, emergency core cooling systems are generally believed to be inef fective for controlling vessel rupture accidents. Second, missiles O generated by vessel f ailure could rupture the containment.4 *' Depending uoon the loca tion and magnitude of the rupture, the fission product transpo rt pathway from the core to the containment at:msphere may be quite short.
O Although the probability of reactor vessel rupture was estimated at about r
1 x 10 ' per reactor year in WASH-140P, other estimtes of reactor vessel f ailure probability are significantly greater. Fo r example, Bush has collected vessel failure probability estim tes from a va riety of sources O ranging f rom t x in" to 4 x ind per reactor year rausH, 19941. Such estimates provide an indication of the uncertainty surrounding reactor vessel rupture as an initiating event for core melt accidents. In view of these widely disparate probability estimtes, it is essential to include reactor O vessel rupture segiences in the source term reassessment orogram.
4 .1.~I Containment Isolation Failures and Pre-Existing Leaks O Only one of the Put acciaent seg2ences examinea in the sou rce term reassessment (a la rge break LOCA at Surry Unit 1) was evaluated for a
'O
4-10 containment isolation f ailure. A later report to the NRC estimated source i
l terms for several additional accident segaences involving containment leakage
)- and isolation failures for each of the NRC reference plants IGIESEXE, ET AL.,
' 1984fl. We results of these analyses, which are provided in Table 4-1 at the end of this chapter, illustrate that the containment failure mde can have a significant ef fect on the source term for some accident sequences.
Pre-exi sting leakage received very scant treatment in the NRC report on AL., 19A51. Although the Containment containment performance rB/GCHI, ET Performance Working Group report acknowledged that operational incidents involving large pre-existing leakage rates had occurred, such incidents were disnissed almost out of hand and were not addressed in the NRC repo rt.
Rather, the maximum containment leakage assumed was that allowed by the i plant's license (in the range of 0.1 to 0.5% of the containment volume per day).
Battelle Pacific Northwest Laboratory (PNL) has ' reviewed Licensee Event Report and Abnormal Occurrence Report abstracts covering the period f rom April
) 1965 through May 1983, and found approximately 2,000 involving failures to Since some repo rts isola te and/or leakage above the allowable ra tes . .
describe ru1tiple incidents, the total number of separate incidents was found to number about 1,000 IPELTO, AME9, & GALLUCCI,10A51.
)
About 70% of all containment isolation failures were found to be caused by valve failures resulting in leakage. Excessive leakage through large purge isolation valves was a prevalent failure mde for PWRs. RWR containment
) isolation problems were found to be dominated by MSIV leakage and problems with vacuum breaker valves, containnent vent valves, and t raversing in-core probe valves. The remaining M% of the containment isolation failures involved penetration f ailures, with personnel air lock problems accounting for Personnel air lock problems '.amina te the P'.A
)- about 9/10ths of these failures.
operating history, with runy such problems due to human er ro r ar.d occurring with the reactor at or near full power.
) PNL identified several incidents (involving both RNRs and PWRs) in which valve failures led to potentially large leakage rates. In addition, PNL found
)
3 4-11 two instances where the containment was directly penetrated by events such as inadvertent drilling (these events were not reported in LERs, but were noted C in Integrated Leak Rate Test reports).
perhaps the most notable long-term breach of containment integrity occurred at the Palisades nuclear plant. Three-inch diameter valves l O_ associated with the containment purge system were lef t open (in a locked-open l position) for 17 mnths. During this period of time, the plant operated at power for 476 days INRC, 19801. NRC has assessed the probability of an undetected breach in containment integrity to be about 1.1 x 10~ per reactor year IEMRIT, ET AL. , 19851. This is at least an order of magnitude more
- m V
o f requ ent than has typically been estimated in PRA studies for similar containment isolation f ailures. The NRC report noted that a breach of 3
containment integrity "is almost always found during containment integrated n leak rate tests", performed on average about once every 1.5 years.
U A study conducted for American Nuclear Insurers found a total of 19 incidents involving long-term losses of containment integrity (i.e., more than
<C two weeks duration) th rough 1979. The containment leak rate in these incidents ranged as high as a factor of mre than 400 times the allowable leak 3
ra te. A total of n incidents involved leakage rates a factor of 20 or more times the allowable (four with leak rates greater than 100 times the
- O allowable) rWEINSTEIN , 19801.
3 '
The Insurers' study concluded that the " availability of leakage integrity" through 1979 was 0.918 For U.S. reactors generally, but that this
- C represented an upper bound and that the actual result would be about 0.850
~ because: (a) all f ailure data available were not reviewed, (b) not all failure da ta for the period was available (testing was still to be done for some units), and (c) many f ailures were proMbly of longer duration than estimated fO but insufficient information was available to say so with certainty. In other l words, containment integrity is maintained only 95 percent of the time.
A mre recent study estimated that containnent integrity is maintained C about 71 percent of the time, with results of ~15 percent for PWRs and 66 percent for BWRs. These results are dominated, however, by very small leak I
lo l
4-12 1
[. areas -(a.60 scpare inches or less) . The probability of violating technical specification containment leak rate requirements by a factor of up to la times O the allcuable leak rate was assessed at 10 percent. It should be noted that a majority of the opera ting experience leading to this result involve the failure or leakage through one barrier of a dual barrier arrangement. Thus, while technical specifications were violated, these failures did not O necessarily result in a direct pathaay from the containnent IPELTO, AMES &
G ALUJCCI, 19851. In addition, nost of these f ailures would have involved .
leakage into a secondary st ructu r e, usually an auxi lia ry building or safeguards building.
O' More importantly, however, very large leak areas of 28 square inches were found to have a probability of 0.01 to 0.031. This is considerably in excess of nost PRA probability predictions for containment isolation failures IPELTO, O Ames s cAca:ccI,198sl. The source term implications of large leak areas with such a high conditional probability are significant. Such results cannot be casually dismissed. Containment. Isolation failure may totally dominate the source term for accident sequences in which containment integrity would O otherwise be preaicted to be maintained for many hours after core melt and vessel failure.
4.1.R tate Recovery of Failed Safety Systems O
The modeling in PRAs of the recovery of disabled safety systems is typically limited to accident phases prior to severe core darage. 9uch modeling has led to reductions in estimated core melt probabilities in recent O PRAs. Although usually disregarded in PRAs, safety equipment could be recov-ered during or even af ter core melt. Some of these recoveries could have a significant impact on source terms. 9everal examples are discussed below.
4.1.8.1 Late Recovery of Containment Sprays and Possible Hydrogen Burns O
Containment sprays fail in a number of important accident sequences. In some of these, recovery of the spray systems by the operators is possible and O could lead to a hydrogen explosion. Perhans the best example is station blackout with late recovery of AC power (af ter core melt).
O
) - 4-13 Tn PWR saquences with containment spray failure resulting f rom loss of AC power, hydrogen burning in the containment is believed to be suppressed by the
'~' large amount of steam which effectively inerts the containment. Under such condi tions, large g2antities of hydrogen and carbon monoxide (from core-concrete interactions) could accumulate in the containment a tmsphere. If containment sprays are recovered late in such an accident sequence, the steam
~O will be condensed and the containment pressure lowered. A large hydrogen burn (or perhaps a hydrogen detonation) could occur under these ci rcumstances, threatening the containment (even for a very strong containment such as Zion)
IHASKIN, ET AL., 199H . At this time none of the NRC's source term models O treat the detonation of combustible gases. This is an important shortcoming in the source term reassessment program.
4.1.R.2 Late Recovery of Auxiliary Feedwater
.O In accident sequences with failure of main and auxiliary feedsater, the steam generator would dry out within an hour. Rabcock & Wilcox plants would dry out much sooner due to the smaller secondary side water inventory. Once n.
" the steam generators dry out, the tubes would heat up to the primary side temperature. 'Ihe late recovery of auxiliary feedwater could then lead to steam generator tube f ailures f rom a combination of thermal shock, a high pressure dif f erential across the tubes, and the possibility of pre-existing O tube defects. This could chang' sn initial accident into a containment bypass sequence with fission product ransport through the failed steam generator INEIGAND, ET AL. , 19841. The degree of fission product attenuation along this release path would depend upon whether, after its recovery, auxiliary n
" fee <taater flow was ruintained to the failed steam generator (s). However, some current emergency procedures instruct the operator to terminate all feedsater
[
flow to a failed steam generator.
4.1.8.3 Late Recovery of Bnergency Core Cooling Late addition of emergency coolant to a hot, steam-starved reactor core could result in what is referred to as " accelerated meltdown" IRIVARD, ET AL.,
O 19941. This issue is explored in Chapter Nine, but is mentioned here because the occurrence of this phenomenon would represent a significant departure f rom the mre typical accident progression as mdeled in PRAs.
O
9v 4.1.8.4 tate Recovery of Hydrogen Igniter System Hydrogen igniter systems are used in PWR ice condenser plants to control hydrogen burns and reduce their effects on the containment and safety systems during severe accidents. No ice condenser plant PRA to date has addressed the possible consequences of hydrogen igniters on accident sequences; the only existing PRA for an ice condenser plant (gequoyah) was conducted before the hydrogen igniters were required at the plant.
It is unrealistic to assume that the hydrogen igniters would always be Q' ' available for sequences which do not involve AC power failure. We are not aware of any probabilistic analysis of the reliability of the igniter system.
Questions abound:
Are there any DC dependencies for the hydrogen igniters?
What is the probability that the operators will fail to start the manually-O initiated system? What is the impact of delayed operation of the igniters?
What is the failure probability of the igniter system?
Late recovery or operation of the hydrogen igniter system could pose serious risks for ice condenser plants during severe accidents involving station blackout as well as other accident sequences. Late recovery of this system, after the accurulation of large amounts of hydrogen and carbon mnoxide, could lead to a very large burn or detonation (which the NRC source O term codes are incapable of mdeling).
Due to the importance of the hydrogen igniters fo r ice condenser PARS, reliability studies should consider both hardware failures of the system as O well as operator error in f ailing to initiate operation of the system. The sensitivity of the dominant accident seglences to the timing of igniter operation should also be examined. We believe that it is quite likely that ice condenser plant sou rce terms will be shown by such analyses to be f dependent on the timing of operation or the f ailure of the hydrogen igniter
, system.
I 4.1.9 Heating of RCS Components
}o. heatup of During the course of the NRC source tenn reassessment program, the reacto r coolant system during severe accidents was identif ied as a O
O' 4-15 potentially important issue. One peer reviewer (T. @eofanous) hel( the view
'h*' *11 '"" *" Id*"' **9"*""** * "ld ""*"'"*llY ***"I' i"
- l'"9* ' C* ** 'h*
O hot leg pipe failed due to loss of structural strength resulting from heating durirg the accident sequence (SERG, 1985]. @is would drastically alter fission product transport pathways for most sequences should it prove to be a urate, thus limitirg the retention of fission products in the reactor O
coolant systen for some sequences in Wiich current predictions show large retention factors.
It is also_,pla_usible that ste am_ genera to r_ tube _s_co.uld_fa i_1__due__to_a v canbination of heatirg fran the circulation of hot gases, fission product
--- - deposition in the steau generator tubes,- high pressure in the reactor coolant systen, and pre-existing tube defects. @is could charge some sequences into
" "'*i"*"' DYP*** "'*"*'I "' "ith th* 'I*Si " P' d" ' '*I'*** d*P'" *"* "
O the timirg of the tube failure and the status of auxiliary feedwater and main feedster at the time of the tube failure [RIVARD, ET AL. , 1984]. We NRC source tenu reassessnent program has not given adequate consideration to di'* t h**ti"9 f '** ' ' 1*"' 'Y"'** " "P "*"
O 4.1.10 In-Vessel Steam Explosions Stean explosions are discussed in some detail in 01 apter Nine. In-vessel O steau explosions could result in a nunber of different outcones for PWR accident sequences. Although there is considerable controversy regarding steau explosions, the argunent is more over the magnitu3e of the explosions than whether they can occur (SERG,1985] .
l - Perhaps the most extreme sequence branch associated with steam explosions is the so-called " alpha" mode failure of the contaiment in which a steam explosion causes the failure of the upper head of the reactor vessel which O then travels upward and hits the contairment dome, causing its failure. A recent review of the conditional probability of this contairunent failure mode contained estimates ranging fran zero (physically impossible) to 0.1 (one chance in ten) [SERG, 1985). We alpha mode contairunent failure, should it
,O
,' occur, would almost certainly be accanpanied by a very large source tenu. It I should be enphasized that the vessel head need not actually penetrate the l l ,
!O
3- 4-16 containment dome to be significant. If suf ficient structural damage is done to the containment to lower its effective failure pressure, this would also be important, as would any ef fects on the operability of containment spra/s and fan coolers.
It has been pointed out that the presence of the overhead crane might If the deflect the vessel head and/or prevent its impact on the containment.
vessel head missile is projected to hit the crane, however, the possibility must be acknowledged that the crane wall (or containment wall if the crane is attached directly to the containment) could be damaged, leading to isolation failure.
A similar failure mode, involving collapse of the crane wall due to
'4 PRA INORTHEAST an earthquake, was postulated in the Millstone Unit UTILITIE9,1983; NORTHEAST UTILITIES,19941.
O Another sequence branch accompanying in-vessel steam explosions is the IRERMAN, massive failure of the reactor vessel bottom head at high pressure 9WENSON & WICKETT, 19941.
Such a failure rx>de could lead to a very energetic reaction in the reactor cavity. In-vessel steam explosions may also affect v the integrity of other RCS components such as the hot leg pipes and the steam generator tubes. Such failures have not been considered in the NRC source term reassessment program.
O 4.1.11 Containment Steam Explosions Although many authorities agree that steam explosions outside the reactor they vessel can occur, these explosions are disregarded on the grounds that While this may be true, such steam cannot cause failure of a PWR containment.
be ignored because of their explosions are, nonetheless impc>rtant and cannot possible impacts on fission product release to the containment.
9 Containment steam explosions could occur if there is water in the reactor
' cavity during a core melt accident. They could occur initially at the time of I vessel failure, as well as much later when core material that did not i
initially melt, melts and falls from the vessel into the reactor cavi ty.
,)
Were is the possibili ty that the resulting late "small" steam explosion could generate suf ficient pressure on water-covered core debris to cause the melt to
g
- 4-17 he pushed down under the explosion site and rise strongly at other locations.
Mixing of the melt and water under these conditions could give rise to a nuch J 1arger steam explosion IRIVARD, ET AL. ,19A41.
Steam explosions outside of the reactor vessel are important to under-standing Cission product transport because they represent a mechanism for v The aerosol particle size increasing the release of non-volatile species.
distribution, a significant factor in aerosol modeling, could also be affected.
The NRC source term codes are not capable of modeling steam explosions, O either inside or outside the reactor vessel.
Due to the apparent la rge likelihood of steam explosions outside the vessel, this is a significant deficiency in the source term codes. A complete analysis of this phenomenon would be required to demonstrate that the existing source term estimates for
.o PWR sequences do not seriously underestimate the possible release of fission products.
4.2 Neglected RWR Accident Sequences and Containment Failure Modes 0
This section of Chapter Four highlights examples of boiling water reactor il (BWR) accident sequences and containment failure nodes that were not examined in the NRC source term reassessment program.
-O 4.2.1 Mark I/II Wetwell Airspace SRV Discharge Line Rupture All BWRs are equipped with safety / relief valves (SRVs) to limit reactor O vessel pressure (see Figure 4-1) . The valves are located on the main steam lines inside the drywell. In the event of a pressure rise in the reactor vessel, the one or note SRVs will open and steam will be discharged to the water-filled suppression pool via pipes referred to as SRV discharge lines.
O In the General Electric Mark I and Mark II BWR containment designs, the discharge lines pass through the air space above the suppression pool water surface. If a discharge line should break in the wetwell air space following O a stuck-open SRV in that discharge line, the steam would bypass the suppression pool and rapidly pressurize the containment (unless operators reduce pressure by activating the containment sprays) .
O
O 4-19 l
A 1992 Brookhaven Na tional Laboratory (BNL) analysis considered such scenarios for Mark I and Mark II containments. BNL estimated the probability O of the SRV discharge line break to be 7.4 x 10-5 per demand, with a range f rom 1 x 10^ to 1 x 10- per demand (depending upon a variety of assumptions including the presence of corrosion) IECONOMOS, ET AL., 19R2; LEHNER &
ECONOMOS, 19911.
O Using their best estimate for a discharge line break, BNL analyzed several accident scenarios. 7he only one of interest involved a transient followed by a stuck-open SRV, a discharge line break, and failure of a Assuming 100% bypass of steam, it was estinated that a
Mark I containment could f ail within 10 minues and a Mark II containment within 14 minutes. Reducing the steam bypass fraction to 56% extended the estima ted containment failure times to 30 minutes for the Mark I and 35 m" minutes for the Mark II. BNL found that containment failure led to loss of coolant inventory and core melting. Since the containment would f ail shortly af ter the accident begins, the core would melt relatively quickly due to the high decay heat level in the fuel IECONOMOS, ET AL, 19921 O
The probability of this sequence occurring was estimated by RNL to be
-6 This is well within the probability range about 1 x 10 per reactor year.
for dominant accident sequences for BWR Mark I and Mark II plants analyzed in O PRAr to date. BNL recommended that this accident sequence be included in all Mark I and Mark II plant PRAs.
- The omission of this accident sequence with a potentially large source term may be a significant deficiency in the NRC source term reassessment program.
O 4.2.2 EMR Interfacing LOCA Sequences An interf acing LOCA is an accident involving failure of the " interface" O or boundary between the high pressure reactor coolant systen and a low pressure system such as the Low Pressure Coolant Injection system, an emergency core cooling system. Such an accident would involve the sudden discharge of reactor coolant at operating pressure and temperature outside the Q'" primary containment and would also likely disable one or more of the systems required to mitigate the accident. It is an important exaanple of a O
4-19 g
" containment bypass" accident which could lead to fission product release to the atmosphere without an actual failure of the containment.
O A recent case study report by the NRC's Office for the Analysis and Evaluation of Operational Data (AEOD) examined historical BWR interfacing LOCA i precursors involving ECCS systems. Eight events between 1975 and 1994, each f g entailing the failure of a testable isolation check valve on the iniection line of a low-pressure emergency core cooling system, were identifled and evaluated.4* Five of these events involved an additional failure of the second and final isolation barrier -- the inadvertyent opening of a normally n
closed motor-operated valve between the RCS and the emergency core cooling u Four of these five events occurred during power operation, in each system.
case leading to an actual overpressurization of the ECCS. Fo rtunately, no actual pipe failures occurred as a result of these incidents. The AET report concluded, however, that there is no assurance that this pattern will continue V,
due to the possibility of a pre-existing flaw in the piping. This path to an interfacing IDCA, involving human erro r, has not been considered in PRAs, which typically have considered only hardware failures of mto r-operated valves in such analyses.
V, Each of these operational events is considered a precursor to an interfacing LOCA and collectively they indicate a trend with serious safety significance -- that the likliehood of such an accident is a hundred or mre O,
times higher than previously estimated. The AEOD report estimated the overall probability of this BWR interf acing LOCA to lie in the range of 7 x If to ?
x 10~ per reactor year fLAM, 19451.
O Oak Ridge National Laboratory examined a second type of interfacing LOCA for the Browns Ferry Mark I design, this one involving a break in the scram discharge volume. The scram discharge volume is a receiver volume for water displaced from the control rod drive mechanisms during a reactor scram. If a break occurs in the scram discharge volur:e following a scram and the scram signal cannot be cleared, the break is nonisolable. No proMbility estimate was provided for this sequence by Oak Ridge, but other estimtes indicate considerable uncertainty regarding the probability of this accident sequence. _s
,O An AEOD report estimated the probability of this sequence to be about 1 x la O
) 4-20 per reactor year, but with a large uncertainty band extending from 1 x 10 to 1 x 10 -9 per reactor year IRURIN,19811. General Electric, designer of Browns O Ferry plant, estimated the probability of an analogous sequence in its GE9SAR
-9 per standard plant design (similar to Grand Gulf) to.he less than 2 x 10 reactor year fGE,19R11.
O Four variations of the scenario were examined by Oak Ridge, with core melt occurring in each. Estimated reactor vessel failure times ranged from 10.1 to 11.5 hours5.787037e-5 days <br />0.00139 hours <br />8.267196e-6 weeks <br />1.9025e-6 months <br />. Containment failure was predicted to occur due to overheating of electrical penetrations ICONDON, ET AL.,19421.
O There is no basis for excluding BWR interf acing LOCA sequences from the NRC source term reassessment program. Such sequences are significant in view of the historical trend of precursor incidents involving overpressurization of O ECC systems in BWRs and the fact that some of these seauences involve a bypass of the primary containment.
4.2.3 Mark I/II IMR Sequences With De-Inerted containment O
D2 ring normal operations, boiling water reactors with Mark I and Mark TI containments have their containments purged of oxygen and replaced with a nitrogen atmosphere to reduce the possibility of fi es or hydrogen burns (or O detonations) during severe accidents. These containments are considered to be "inerted", and, as a result, phenomena associated with hydrogen burning or detonation receive scant attention in nnst PRAs of Mark I and Mark II boiling water reactors.
O Durir.g the initial phase of reactor startup f rom any outage in which the containment has been de-inerted, Mark I and Mark II reactors are permitted to be operated for up to 24 hours2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br /> under de-inerted conditions. Similarly, for 24 O hours prior to a planned shutdown, the containment may be de-inerted iSECY-80-107A, 19801.
The NRC recently granted a license amendment for Millstone Unit I that O
w old permit the containment to be deinerted for up to 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br /> to conduct a drywell entry at power for perfonning inspections, equipment adiustments, and O l
b 4-21 maintenance ISHEA,19851. We 'have not attempted to ascertain whether similar amendments have been requested by other utilities. Further, there is no clear h indication of how of ten the authority to de-inert the containment while at power will be utilized. Even if this authority is used only once per year, however, this will effectively double the probability of the containment being de-inerted at the time of an accident. Thus, for Millstone Unit 1, the 3 The possibility of an ;
containment could be de-inerted 2% of the time or more. l accident under de-inerted conditions was apparently not considered in the Millstone Unit 1 PRA; this could be a significant omission.
I 3 However, the PRA for Limerick estimated that the containment for that reactor would be de-inerted while the reactor was at power about 1% of the time. %e PRA concluded that if an accident occurred under de-inerted conditions, a hydrogen burn would cause an overpressurization failure of the i.e., involving 9 containment with a direct pathway to the envi ronment -
f ailure of the drywell, thus bypassing the suppression pool. The Limerick PRA estimated there would be a 10% chance of the fission product release approximting that derived in WASH-1400 for a steam explosion which causes a direct breach of the drywell IPECO, 19921. This indicates that accidents
- O occurring with the containment de-inerted should be included in the source term reassessment program.
O Under de-inerted conditions, burning involving even a smil amount of hydrogen would be sufficient to cause containment failure in either the Mark I or Mark II BWR containment designs. It has been estimated that burning the
! hydrogen resulting f rom 13-16% of the fuel cladding reacting with water would O result in the failure of a Mark I containment, while 15-20% would suf fice for a Mark II containment (CAMP, ET AL., 14911. Accident sequences which under inerted conditions would be predicted to be relatively benign would, with the containment de-inerted, become Far more severe.
- O Asstming that the probabilities for the more probable BWR accident sequences (in the range of 10- to 16- per reactor year) remain unchanged during de-inerted conditions, a 1% chance of their occurring during this O vulnera ble time represents a significant sequence branch with important implications for source term estimates. L't is, therefore, important that the
'O
N -
=4-22 q,
' 'NRC-source term reassessment program consider accidents for Mark I and Mark II.
' BWRs under de-inerted conditions.
O - 4.2.4' BWR Reactor Vessel Rupture As' . Indicated in ' the discussion of vessel rupture for PWRs - (Section 4.1.5), reactor . vessel rupture sequences could have important imlications for O. ~ the source term issue. This is particularly true for BWRs due to the relative vulnerability 'of some containment designs to mechanical damage resulting f rom vessel rupture. _
O 'The PRA for Limerick INUS,19831 concluded that a reactor vessel . rupture for a Mark II design could cause an imediate containment failure. .The PRA assigned 1M of the random reactor vessel upture sequences :to a large release
'*fl* ti"9 'hi" **'lY "'*i"**"' f*il"'*
- '*9 'Y (l*r9' * "' * *)
O possibility. It is important that similar assessments be carried out for Mark I and . Mark III designs, and that BWR . reactor vessel rupture segnences he included within the NRC source term reassessment program.
O 4.2.5 Main Steam Isolation Valve (MSIV) Isakage e
Main steam isolation valves (MSIVs) are large, normally open valves in' the steam lines f BWRs. They permit steam to flow f rom the reactor vessel to O
the turbine-generator to produce electricity. The M9IVs are designed to close and stop the flow of steam during accidents. If the M9tVs fail to close or experience significant leakage during a severe accident, fission products can
- g. be released to the environment.
Many. BWRs are equipped with an MSIV leakage control system. However, it is by no means certain that such systems would fun: tion -adequately to prevent
- g. a large fission product release to the environment in the event of MSIV failure or leakage. The problem is of special concern for the BWR Mark TII containment design for which analyses predict considerable fission product
. retention in the suppression pool that effectively surrounds the drywe ll.
MsIV le kage ola Itmit the effectiveness of suppression pools in retaining O
fission product during severe accidents in Mark III RWRs INURE-coQ, 19941.
O.
O 4-23 t
Data collected by the NRC in 19R2 indicate a relatively high probability of MSIV leakage, considerably in excess of Technical specification limits (and a
V also in excess of MSIV leakage control system design speci f ica tions) .
Moreover, the MSIV leakage control system is not always available during a severe accident. Such systems are estimated to fail about 5% of the time on demand. Compounding the problem, the PRA for Grand Gulf indicates that the O steam and waste treatment system (the only possible backup to the MSIV leakage control system) would not be available about 26% of the time for severe accidents (due to loss of AC power to the system). If neither system works, leakage would occur essentially directly to the environment.
O 4.2.6 Steam Explosions As discussed above, steam explosions could occur both inside the vessel O and, for scenarios where water is present below the reactor vessel in the pedestal region, outside the vessel as well. Explosions outside the vessel pose a particularly serious problem for some Bpolling water reactors. For example, the shoreham PRA r redicted that the 24-inch diameter downcomers (used O in the Mark II design to direct steam into the suppression pool) would conduct much of the miten core material into the suppression pool. A recent analysis
' carried out for the NRC Containment toads Working Group suggests that, for the Shoreham design, steam explosions resulting from ex-vessel interactions of This is considerably O molten core debris and water could fail the containment.
mre threatening than the behavior portrayed in the Shoreham PRA, which assumed that all of the melt would be quenched harmlessly in the suppression pool. Ex-vessel steam explosions might also cause containment failure in O plants such as WPPSS Unit 2 that utilize a BWR Mark IT containment with a f ree-standing containment wetwell wall ICORRADINI, ET AL.,19841.
In-vessel steam explosions might cause the so-called " alpha" mode of O containment failure (direct breach by failure of the upper head of the reactor vessel and penetration of the primary containment by this misslie). As with the pas, such a failure mde could be accompanied by a very large source tecn.
O.
O
O 4-24 L
4 . 2 . ~1 toss of Decay Heat Removal Sequence Branches d
]
A report by Oak Ridge National IAboratory examined in detail the loss of The decay heat removal accident sequence for the Browns Ferry Unit I reactor.
residual heat removal function in the Browns Ferry plant can Fall in two principal ways. First, if the residual heat remval (RHR) system itself O fails, not only is heat removal lost, but the ability to use the RHR system to It has circulate the suppression pool water to promote mixing is also lost.
been suggested that this could result in localized heating of the suppression
- pool and subsequent containment f ailure at an earlier time than would O otherwise be predicted. Second, the heat removal capability could be lost if the service water system fails (service water is used as a heat trans fer medium via the RHR heat exchangers) . This causes the failure of decay heat
- removal, but does not affect the ability of the operators to use the RHR
~ This could delay containment O system to circulate the suppression pool water. fails failure beyond the time period in which it would occur if the RHR itself TCOOK, ET AL.,19811.
O A potentially very significant variant of this accident sequence was identifled in the ORNL study. Upon containment failure, a large IDCA could be
. the created, severing the ECCS injection lines during the blowdown of containment. If this failure occurs, fission products released during this Although O phase of the accident sequence would bypass the suppression pool.
containment venting procedures exist that could avoid containment failure,
- thus avoiding core melt, sources differ on the ef fect this would have on the probability of this accident sequence.
O 4.2.8 Station Blackout Sequence Branches, BWR/4 Mark I Design Another ORNL report in the " Severe Accident Sequence Analysis" MASA)
Six variations were examined O program examined the station blackout sequence.
involving variations in the use of HPCI and RCIC, as well as the possibility of a stuck-open SRV ICOOK, ET AL. ,19811.
In two cases examined, containment failure by wetwell failure due to O Such steam jet impingement and condensation oscillations was postulated.
O
4-25 C I phenomena would be caused bf thermal stratification of the suppression pool.
Containment failure under these circumstances was postulated to occur within two minutes after vessel failure.
J In other cases, containment failure by drywell electrical penetration failure was postulated. Such failures were estimated to occur at about 50-90 minutes af ter vessel failure. Fission product release through failed O'- electrical penetrations would bypass the suppression pool.
4.2.9 targe LOCA Sequence Branches, BNR/4 Mark I Design Q In yet a third ORNL report in the SASA series, a large [nCA sequence was examined. 'Jsing the WASH-1400 containment failure pressure of 175 psig, i containment failure times for a variety of cases ranged f rom 17 to 1R3 minutes, with no containment failure predicted in one case fYUE & COLE,19821.
Such variations in containment failure timing need to be examined in the source term context. .
4.2.10 BWR/4 Mark I A'IWS Sequence Branches O
I ORNL also examined ATWS sequences for the Browns Ferry Unit I reactor.
Under a variety of assumptions concerning operator actions (or the lack thereof), containment failure times were estimated to range f rom '47 minutes to V about 12 hours1.388889e-4 days <br />0.00333 hours <br />1.984127e-5 weeks <br />4.566e-6 months <br /> IHARRINGTON & HOOGE, 19841. These are very large variations which could have substantial impacts on the estimated source term for A'I*AS sequences at BWR/4 Mark I plants.
O l
l l
lO
d%
J 4-26 4.3 End Notes for Chapter Four s 4.1 The NRC SGm risk analysis estimates that tubes with defects more than J.. 75% through-wall can rupture in main steam line break (MSLB) sequences ,
and that tubes with defects greater than PR% through-wall can rupture On this basis, during normal operation IAMS1ULIMITZ, ET AL. , 14R51.
cultiple MTRS cannot be considered to be unlikely. An incident at San Onof re Unit 1 involved stress-corrosion cracking in 460 steam generator
. tubes (the plant is a Westinghouse PWR with a dry steel spherical O containment). Over 60% of the affected tubes had defects exceeding 95%
th rough-wa ll . . - This cleary points out the potential for nultiple SGTR sequences. Various sources have estimayed the probgility of nultiple to 6 x in per reactor year SGTRs as ini tiating events at 6 x 10 I AKS1ULEWIT7., ET AL. ,1985; PASNY & CONED,19811.
O 4.2 In-vessel steam explosions would only have to breach the vessel in order to have a -significant inpact on source terms since the containment is already failed in this sequence. An uncertainty analysis of PWR steam explosions suggests that in-vessel steam explosions may, with a relatively high probability, cause failure of the reactor vessel bottom head iBERMAN, SWENSON & WICKETP,19941.
O RCP seal LOCAs have occurred at Haddam Neck, 0 onee Unit 2, H.B.
4.3 Robinson Unit 7, Indian Point Unit 2, Salem Unit 1, Arkansas Nuclear One Unit 1, and Sequoyah Unit 1.
4.4 The Surry analysis in WASH-1400 concluded that the presence of the polar O crane structure above the vessel would preclude missile penetration of the containment. Thus, the Surry PRA modeled reactor vessel rupture fRA9MUSSEN, sequences as occurring with an initially intact containment ET AL., 19751. Whether this conclusion would hold for other designsFor is a matter that needs to be resolved on a plant-specific basis.
example, it seems plausible that in the Sequoyah plant, in-vessel steam O explosions would only need to breach the barrier between the lower compartment and the upper compartment to be significant in the source term context. This is due to the very low failure pressure of the ice condenser containment, and the f act that breaching the barrier between the lower and upper compartments would lead to a bypass of the ice condenser.
.O BNL noted that as of October 1982, no published BWR plant PRA had 4.5 considered the SRVDL break sequence. BNL reconnended more research on SRVDL failure rates, SRV failure rates, operator reliability under very high stress conditions, and the inpact of the SRVDL break environment on the operability of safety systems to better define the SRVDL break O sequence.
d.6 The eight, precursor events occurred at Vernunt Yankee, Cooper, LaSalle Unit 1, Pilgrim, Hatch Unit ?, and Browns Ferry Unit 1.
Events at l
Vermont Yankee, Cooper, Pilgrim, and Browns Ferry Unit 1 involved actual
! overpressurization of an ECCS system while the reactor was at O- essentially full power ILAM, 19951.
I LO l
l L
O 27 4.4 References for Chapter Four AKSTULEWITZ, ET AL.,'1QR5 y A. Akstulewitz, S. Bryan, A. Buslik, E. Butcher, L. Frank, .T. Micthell, M. Hawkins, G.
E.
holahan, T. Ippolito, T. Marsh, R. Martin, C. McCracken, Serbu, .T . St rosnide r, ' B.
Murphy, P. No rlan, L. Phillips, R. Riggs, R.
K. Wichman, NRC Integrated Program for the Resolution of Turovlin, &
f Unresolved Safety Issues A-3, A-4, and A-5 Regarding f Nuclear Steam Re etGenerator r Regulation,Tube ivisi n e Licensing, eelce Integrity, NUR m-0844, draft O U.S. Nuclear Regulatory Commission (Washington, D.C.),
report for coment, April 19R5.
ANDERSON, ET AL. ,1983 Roesn, & G.R.
G.B. Reddy, 9.
L.E. Anderson, D ..T . Ayres, D.L.
Brown,
'Ihomas, Effects of A Hypothetical Core Melt Accident on a P4R Vessel with g". (Palo Alto, Top-Entry Instruments, Electric Power Research Institute (Bethesda ,
California), prepared for IDCOR, Atomic Industrial Forum Maryland), IDCOR Technical Report 15.2A, . Tune 19R1 BN3 CHI , ET AL. , 19R5 Butler, .T. Shapaker, W.
G. Bagchi, P. Niyogi, V. Noonan, K. Darczewski, Pratt, T. Bridges, W.
'g .T . Huang, R. Palla, C. Hofmayer, B. Miller, W.T.
Sebrell, & L. Greimann, Containment Performance Working Group Report, Nuclear Regulatory Comission Office of Nuclear Reactor Regulation, U.9 (Washington, D.C.), NURE-1037, draf t report for comment, May 1985.
RARRE'IT & THOM, 1995 q"
R..T. B3rrett & E.D. Thom, " Consequence Evaluation for Pressurized Thermal D.C.), in Shock," U.9. Nuclear Regulatory Commission (Wa shi ng ton ,
Proceedings of the U.S. Nuclear Regulatory Commission Nuclear RegulatoryTwelf th Water Comission Reactor Safety Information Meeting, U.S.
(WashLngton, D.C.) , NURKi/CP-0058, .Tanuary 1985.
l 20' BFRMAN, SWFJJSON & WICKE'IT,1984 Wickett, An Uncertainty Study of PdR M. Aerman, D.V. Swenson, and A..T.
Steam Explosions, Sandia National Laboratory ( Albuquerque, New Mexico),
NURm/CR-3369 (SAND 93-1418) , May 1984.
BROOKS , ET AL. , 1984 p A.C. Brooks, M.G . K . Evans, G. Forster, P. Guymer, R. Karimi, & P.O.
O'Reilly, Preventive Methods to Arrest Sequences of Events Prior to Core Ihmage, NUS Corporation (Rockville, Maryland), prepared for IDCOR, Atomic Industrial Forum (Bethesda, Maryland), IDCOR Technical Report 9.1, '
Revision 1, September 1984.
N n
BUSH, 1984S.H. Bush, "An overview of Pipe Breaks f rom the Perspective of Operating Experience," Review and 9ynthesis Associates, in Proceedings of the U.S.
Nuclear Requiatory Commission Eleventh Water Reactor Safety Research (dashington, U.S. Nuclear Regulatory Comi ssion Information Meeting, D.C.), NURE/CP-0048, .Tanuary 1984 y
O
g V 4-2R CAMP, ET AL., 1983 A.L. Camp, J.C Cumings, M.P. Sherman, C.F. Kuplec, R .J . Healy, J.S.
Caplan, J.R. Sandhop, & J.H. Saunders, Light Water Reactor Hydrogen C, Manual, Sandia National Laboratories (Albuquerque, New Mexi co) ,
NURm/CR-2726 (SAND 82-ll37), August 1993.
CLARK, CLAPP & BROADWATER,1994 F.H. Clark, N.E. Clapp, & R. Broadwater, possible Modes of Steam Generator Overfill Resulting from Control System Malfunctions at the n Oconee-1 Nuclear Plant, Oak Ridge National Laboratory (Oak Ridge, V
Tennessee), NURED/CR .E92 (ORNL/TM-4061), July 1994.
CONDON, ET AL., 1992 W.A. Condon, S.R. Greene, R.M. Harrington, S. A. Hodge, & D.O. Yue, SBLOCA Outside Containment at Browns Ferry Unit One -- Accident Sequence O Ana lys i s_, Oak Ridge National Laborato ry (Oak Ridge, Tennessee) ,
NURW/CR-2672, Vol.1 (ORNL/TM- R119/V1) , N0vember 1992.
COOK, ET AL., 1981 D.H. Cook, S.R. G reene, R.M. Harrington, S.A. Hodge, and D.D. Yue, Station Blackout at Browns Ferry Unit One - Accident Sequence Analysis, O Oak Ridge National Laboratory (Oak Ridge, Tennessee), NURD3/CR-218 2 (ORNL/NURFU/TM-455/V1), November 1991.
COOK, ET AL., 1983 D.H. Conk, S.R. G reene, R.M. Harrington, and S.A. Hodge, Ioss of DHR Oak Sequences at Browns Ferry Unit One - Accident Sequence Analysis, O Ridge National Labo rato ry (Oak Ridge, Tennessee) , NURED/CR-2973 (ORNL/TM-8532), May 1993.
1984 CORRADINI, ET AL. , i, C.C. Chu, K.Y. Huh, G. A. Moses, J. Norkus, M.D. Oh, and M.L. Corradin T. Welzbacker, Analysis of Ex-Vessel Fuel-Coolant Interactions in an UVR,
.O Nuclear Engineering Department, University of Wisconsin (Madison, Wisconsin), prepared for the NRC Containment toads Working Group, May 1994.
COTTRELL, ET AL., 1994 W.B. Cottrell, J.W. Minarick, P.N. Austin, E.W. Hagen, and J.D.
Davis,
'O Precursors to potential Severe Core Damage Accidents: 1990-1991, A Status Tennessee),
Report, oak atoge National Labo ra to ry (Oak R i c1g e ,
NURED/CR-3591 (ORNL/NSIC-217/V1), July 1994.
ECONOMOS, ET AL., 1982 C. Econocras, J. Lehner, J. Ranlet, & G. Maise, Postulated SRV Line Break in the Westwell Airspace of Mark I and Mark II Containments, Brookhaven O National Laboratory (Upton, New York), BNL-NUREU-11946, October 1942.
EMRIT, ET AL., 1985 R. Emrit, W. Minners, H. VanderMolen, R. Colmar, D. Thatcher, J. Pittman, W. Milstead, R. Riggs, G. Sege, P. Matthews, & L. Riani, A Prioritization O of ceneric Safety Issues, Office of Nucinar Reactor Regulation, u.S.
Nuclear Regulatory comission (Washington, D.C. ) , NURFn-a931, Supplement No. ?, January 1985.
O
w 4-29 j
GARCIA, ET AL., 1984 A.A. Garcia, D.L. Bernreuter, T.E. McKone, P.D. Smi th, P.J. Amico, J.W.
Reed, M.W. McCann, Jr., P.R. Davis, & G. Apostolakis, A Review of the l
O Millstone-3 Probabilistic Safety Study, Lawrence Livermore National l Laboratory (Livermore, california), dratt report,13 May 1994.
GARRICK, ET AL., 1983 B.J. Garrick, K.N. Fleming, D.W. Stillwell, A. Torri, T.E. Potter, D.C.
Bley, A. Mosleh, M. Kazarians, H.P. Perla, J.G. Stampelos, D.R. Ruttemer, et al., Seabrook Station Probabilistic Safety Assessment, O D.C. Iden, Pickard, Lowe & Garrick, Inc. (Irvine, Ca11tornia), prepared tor Public Service Company of New Hampshire and Yankee Atomic Electric Company, PLG-0300, December 1991.
GE, 1981
.O General Electric Company, GE Evaluation in Response to NRC Request Regarding BNR Scram Discharge system Pipe Breaks, General Electric Company (San Jose, California), NEDE-24142 (81NED261), April 1481.
.t GIESEKE, ET AL., 1984f J.A. Gieseke, H. Chen, P. Cybulskis, R. Freeman-Kelly, M.R. Kuhlman, &
'O K.W. Lee, Informal Report on Source Term Predictions for Various Containment Failure Assumptions, Battelle Columbus Laboratories (Columbus, Ohio), 29 August 1994 HARRINGTON & HODGE,1984 R.M. Harrington and 9.A. Hodge, AWS at Browns Ferry Unit One - Accident
'O Sequence Analysis, Oak Ridge National L3boratory (Oak Ridge, Tennessee),
NUHUVCH- 34 m (UHdL/TM-8902) , July 1984
~
HASKIN, ET AL., 1991 F.E. Haskin, W.B. Murfin, J.B. Rivard, and J.L. Darby, Analysis of a Hypothetical Core Meltdown Accident Initiated by a Ioss of Offsite Power for the Zion 1 Pressurized Water Reactor, Sandia National Laboratories O (AlbuquerqJe, New Mexico), NUREG/CR-1988 (SAND 91-0903) , November 1941.
HUBBARD, ET AL., 1984 F.R. Hubbard III, J.W. Stetkar, A. Torri, T.E. Dotter, H.F. Perla , B.J.
Garrick,J.P. Kindinger, H.9. Tsai, T.E. Hollowell, M.T. Nordin, & R.F.
Midland Nuclear Plant Probabilistic Risk Assessment, Pickard, Schofer,
'O Lowe & Garrick, Inc. (Irvine, California), prepared for Consumers Power Company, May 19R4
- LAM, 1995 P. Lam, Preliminary Case Study Report
- Overpressurization of Emergency Core Cooling Systems in Boiling Water Reactors, Office for the Analysis
-Q and Evaluation of Operational Da ta , U.S. Nuclear Regulatory Commission (Washington, D.C.), February 1995 LEHNER & ECONOMOS, 1991 J.R. Lehner & C. Econonos, Evaluation of SRV Pipe Failure Rates Via
- O Probabilistic_ Mechanical Design, Brookhaven National laboratory (Upton, New York), BNL-NURm- Vb5H , March 19Al.
lO
t .
4-30 g
LU, ET AL., 1983 Levine, Analysis of M.S. Lu, W.G. Shier, D.J. Diamond, & M.M.
Loss-of-Feedwater Transients Without Scram for a Westinghouse PWR, Brookhaven National Laboratory (Upton, New York), BNL-NUREI;-31673, July g~
1993.
MA'ITSON, ET AL. , 1983 R.J . Mattson, G.M. Holahan, L. Crocker, M.W. Hodges, S.L. Israel, W.G.
Kennedy, G. Lanik, W.D. Lanning, J.G. Partlow, W..T. Raymond, C.E. Rossi, Meltemes, E. .To rdan , J.
g' P.C. Shemanski, H. Silver, M. Ernst, C.J.
Olshinski, R. Sta rostecki, P. Baranowski, J.T. Beard, C.G. Graves, R.
Kendall, K.G. Murphy, D.W. Pyatt, D.L. Serig, I. Villaiva, & W.E. 011u, Generic Implications of the AWS Events at the Salem Nuclear Power Plant, NURFri-1000, Vol.
Nuclear Regulatory Commission (Washington, D.C.),
U.S.
1, April 1983.
m
MURLEY, 1981 Menorandum f rom T.E. Murley to D.G. Eisenhut, " Reactor Coolant Pump Seal Failure," U.S. Nuclear Regulatory Comission (Washington, D.C.), 27 March 1981.
N RWEAST U W TIES, 1983 O Northeast Utilities, Millstone Unit 3 Probabilistic Safety Study, Northeast Utilities (Berlin, Connecticut), August 1993.
NOR'IEEAST UTILITIES,1994 Northeast Utilities Service Company, Millstone Unit 1 Probabilistic Safety Study,' Amendment N .?, N ttheast Utilities Service Company O' (Berlin, Connecticut) , 7 April L984 NRC, 1979 Of fice of Inspection and Enforcement, " Steam Generator Tube Ruptures at Two PWR Plants," U.S. Nuclear Regulatory Comission CVashington, D.C.) ,
IE Information Notice No. 79-27, 16 November 1979.
,g NRC, 1940 Of fice of the Executive Legal Director, "NRC Staf f's Answers to Consumers Power Company's 'First Round of Interrogatories and Request fo r the Production of Documents'," U.S. Nuclear Regulatory Commission cVashingt n, .C.), ket N . sn >ss-Civil Penalty,11 March 1999.
O NSAC, 1994 Generic Safety Issue Tracking and Nuclear Evaluation Safety Summary Analysis Cente Nuc Description, r , lea r darety Analysis center (eato Alto, California), NSAC-9), July 1984.
O NUREG-0979, 19R4 Office of Nuclear Reactor Regulation, Safety Evaluation Report related to the final design approval of the GESSAR-II RVR/6 Nuclear Island Design, Nuclear Regulatory Commission (Washington, D.C.), NURE73-0979, U.S.
Supplement No. ?, November 1994.
O O
c 4-31 NUS, 19R1 NUS Corporation, Severe Accident Risk Assessment: Limerick Generating Station, NUS Corporation (Rockville, Maryland), prepared for Philadelphia
) Electric Conpany, NUS Report No. 4161, April 19R3.
PASNY & CONSOLIDATED EDTSON, 194?
Power Authority of the State of New York and Consolidated Edison Company of New York, Inc., Indian Point Probabilistic Safety Study, Power Authority of the State of New York and Consolidated Edison Company of New
) York, Inc. (New York, New York), March 1991.
PAYNE, ET AL., 1984 Interim Reliability A.C. Payne, principal investigator, et al.,
Evaluation Program: Analysis of the Calvert Cliffs Unit 1 Nuclear Power Plant, Sandia National Laboratories (Albuquerque, New Mexico),
) NUREU/CR-3511/1 of 2 (SANDR3-20R6/1 of 2), March 1034 PECO, 1982 Philadelphia Electric Company, Probabilistic Risk Assessment, Limerick (Philadelphia, Generating Station, Philadelphia Electric Company Pennsylvania), Rev. 5, September 1982.
) PELTO, AMES & GALLUCCI,1995 Gallucci, Reliability Analysis of P . .T . Pelto, K.R. Ames & R.H.
Containment Isolation Systems, Pacific Northwest Laboratory (Richland, Washington) , NUREU/CR-427.0 (PNL-5432) , June 1985.
) N.C. Rasmussen, et al., Reactor Safety Study: An Assessment of Accident Risks in U.S. Comercial Nuclear Power Plants, U.S. Nuclear Regulatory Commission (Washington, D.C.), NUREU-75/014 (NASH-1400), October 1975.
RIVARD, ET AL., 19R4 J.B. Rivard, V.L. Beh r, R.G . Easterling, .T.M. Griesmeyer, F.E. Haskin,
] M.P. Sherman, A.R. Taig, 9.W. Ihtch, A.M. Kolaczkowski, R ..T . Lipinski, and A..T. Wickett, Identification of Severe Accident Uncertainties, Sandia
( Albuqu erque , New Mexico), NUREG/CR-1440 Na tional Labo rato ry (SAND 93-1689) , September 1984.
) RUBIN, 1991 S. Rubin, Safety Concerns Associated With Pipe Breaks in the BNR Scram System, Office for the Analysis and Evaluation of Operational ruta, U.S.
Nuclear Regulatory Commission (Washington, D.C. ) , NUREU-0785, draft report for coment, May 1991.
SECY-80-107A, 1990
] Memo randum from H.R. Denton to the NRC Commissioners,
Subject:
" Additional Information Re: Proposed Interim Hydrogen Control D.C.),
Requirements," U.S. Nuclear Regulatory Commission (Washington, SECY-80-107A, with attachments, 22 April 1980 D
4-32 .
O-SERG, 1985 A Review of the Current Understanding of Steam Explosion Review G roup, the Potential for Containment Failure From In-Vessel Steam Explosions, Office of Nuclear Regulatog Research and Office of Nuclear Reactor g Regulation, U.S. Nuclear Regulatory Commission (Washington , D.C.),
N'JRm-lll6, June 1985.
SHEA, 1985 J.J. Shea, safety Evaluation by the Office of Nuclear Reactor Regulation Supporting Amendment No.102 to Provisional Operating License No. DPR-21, g' Northeast Nuclear Energy Company, Millstone Nuclear Power Station, Unit.
No. 1, Docket No. 50-245, Office of Nuclear Reactor Regulation, U.S.
Nuclear Regulatog Commission (Washington, D.C.), 5 June 1995.
SHERON,.1984 Memorandum f rom B.N. Sheron to L.G. Hulman, " Millstone-3 DES: Containment U Failure m trix and Source Terms," U.9. Nuclear Regulatory Comission (Washington, D.C.) , 14 May 1984, enclosing NRC report, Containment 10 Failure Matrix and Radiological Source Tenn for the Millstone-3 DES, May 1984.
StK; NET, ET AL., 1994 O W.R. Sognet, project manager, et al., Oconee PRA: A Probabilistic Risk Assessment of Oconee Unit 1, Nuclear Safety Analysis Center '(Palo Alto, Calif ornia) and Duke Power Company (Charlotte, North Carolina), NSAC-60, June 1994.
WEIGAND, ET AL. ,1984 0 c.c. Weigana, M.D. Bergeron, T.E. ale $was, A.L. Camp, M.J. Clauser, R.K.
Cole, Jr., M.L. Corradini, P.N. Demie, S.E. Dingman, F.E. Haskin, S.N.
Kempka, R. Knight, R .J . Lipinski, K.K. Murata, M. Pilch, A.C. Ratzel, Rexroth, Rivard, and C.J. Shaffer, R.T. Reese, P.E. J.B.
'Ihermal-Hydraulic Process Modeling in Risk Analysis: An Assessment of the Relevent Systems, Structures, and Phenomena, Sandia National Iaboratories O (Ainuquerque, new Mexico) , NUREL/CR-39% (SAND 94-1219), August 1084.
WEINSTEIN, 1990 M.B. Weinstein, " Primary Containment Leakage -Integrity: Availability and Review of Failure Experience," Nuclear Safety, Vol. 71, No. 5, Nuclear Safety Infonnation Center, Oak Ridge National Laboratory (Oak Ridge, O Tennessee), September-October 1990, pages 619-612.
WE & EOLE,1992 D.O. Yue and T.E. Cole, BWR/4 Mark I Accident Sequences Assessment, Oak Ridge National Labo ra to ries (Oak Ridge, Tennessee), NUREG/CR-79 7 9 (ORNL/TM-914R), November 1992.
l 2
i
- O 4
0
4-33 TABLE 4-1 Estimated Source Terms for Containment Leakage Scenarios
- Containment Release Fractions Accident Failure Te Plant Sequence Mode CsI CsOH _
Station BMI-2104 1.9 x 10~ 1.4 x 10~ 7.8 x 10 Zion
)i Blackout Basemat (TMLB') Melt thru Medium Leak 7.1 x 10-5 7.2 x 10-5 5,5 x ig -3
~
9.0 x 10" 9.1 x 10~ 1.6 x 10
) Large IAak 7.0 x 10" 7.1 x 10
-3 Isolation 2.2 x la Failure
-2 -2 -1 Surry Station BMI-2104 4.6 x 10 3.9 x 10 .),1 x gg
} Blackout Overpressure (W LB') ~
1.9 x 14 ~3
~
Large Leak 1.1 x 10 1.2 x 10
~
Isolation 2.2 x 10 1.3 x 10~ 1.1 x 10~
Failure
]
~4 -4 ~3 Sequoyah Station BMI-2104 3.9 x 10 4.5 x 10 Lax 13 Blackout Overpressure (W LB')
BMI-2104 1.7 x 10-2 2.1 x 10~ 1.4 x 10 -2
} H2 Burn
~3 -2 Isolation 4.0 x 10~ 4.3 x 10 1.7 x 10 Failure Grand SE EMI-2104 7*"
- 10 G*I
- 10 2*4
- IU"'
2 Gulf No pool bypass SE Is 1 ti n 3.9 x 10 1.5 x 10 ~3 2.0 x 10-2 2
Failure **
) BMI-2104 1.0 x 10'I 4.1 x 10
-2 2.5 x 10
-1 PeacN TC Bottom Drywell Overpressure
~1 -2 -1 3t*** Isolation 1.0 x 10 4.1 x 10 7,5 x gg Failure I
) .
, m.
4-34 g
TABLE 4 - 1 (continued)
Estimated Source Terms for Containment Leakage Scenarios
- O N(7TES:
- Caution mst be employed when interpreting these results. Only release fractions for cesium iodide, cesium hydroxide, and tellurium were provided. There are clearly other radionuclide groups of interest in terms of of fsite health ef fects and land contamination, including It should also be kept in mind that the I strontium and the lanthanides.
results are sequence and plant specific, and depend upon the design features at the plant and the containment failure pressure and node assumed in the analyses. The leak areas were described in the Battelle repo rt (GIESEKE, ET AL. , 1984fl in terms of ef fective leak area versus containment pressure. Isolation failure was postulated to involve a O 6-inch diameter hole effectively leading directly to the environment.
- 'Ihe containment isolation failure would appear significantly larger in comparison with the " normal" sequence if General Electric Company's suppression pool decontamination factors were used in the analysis rather than the NRC's SPARC code.
g
- Containment isolation failure in this case is unable to provide sufficient pressure relief, and the containment still Fails before core Thus, the source terms are melt due to internal steam overpressure.
unchanged.
D D
D D
e O
^>,,
4-35 FICURE 4-1 l
O 0%L-0WG 82-5187 E 70
- w. , , :,m
).a ; ; A*. .
- lI*c
>S. . ,..,
O ,i f'a I I t'.
0 .O
'h .
- *1
. h
~
/
C E>
, REACTOR E .
- n. 3 J r, (i,
o MAIN STEAMLINE .
SAFETY l
C l
l p.* '
RELIEF VALVE
- C h
e il .
- "h,j -
aa '-
O c ')
li
-. q, 1
- u ) *.
( 0 i 3.*. ... -cmm : .f~ d
.s
.e O e
q v
I
- WETWELL * . -
m*p -. -."
~7~ p q *3*C y m em d
- L-.*l[ , v.>d^', ., OL L -
f: l< ,s p; c :.
)* c , r J= .'y OOWNCOMER
- b. .
.' N<
9' -
I '
Og
>0 >*d tse
' ~ * ' ^ ~ ^^^ ^
^ ' ' ' ^
SAFETY RELIEF >
- VALVE 'yl
- v -
DISCHARGE
'9d a ~ * -
D .. C DEVICE 't IOUENCHER) ; ,
j , ,
h*i
,, , ,, ,, o pSUPPORT
... o j ,j -
y COLUMN f O , R -
- - - ' ~~-- {!,
r-
- $** 1 .
lo Diagram of a BWR Mark II Containm2nt Shcwing Safety Pelief Valve and Safety Belief Valve Discharge Lines
!O l
O v
l - . . ._ ._
. . - . ~. ~ _ - -- -- . - . - .
(D .
CHAPTER FIVE n.
V Neglected Externally Initiated Accident Sequences SUM %RY Am ng the ac taent sequen es n t n w incluaea in the s urce O term reassessment are those externally initiated.. %is chapter describes _ some of these seg2ences: seismic events, including various possible co@lications such as the imaet of af tershocks on containments; airplane crashes into struc-tures other . than . the containment ~ (e .g. auxiliary building, ffsite Power equipment, cooling water conduits); sabotage; i'O biofouling by aquatic organisms; indirect ef fects of fires, such as - smoke damage; and design, construction, and opera-tional errors.
4 External events are important ' to the soure term reassesment, because, like internally initiated acciaents, they can lead
- O. -
to large releases of radioactive. materials to the environ-ment. .%e failure to include them in the NRC's review is a serious shortcoming.
4 M General Considerations for External Event Accident Sequences ii
- O Amnng the external events evaluated by recent PRAs are earthquakes, in-plant flooding, in-plant fires, turbine missiles, aircraf t crashes, loss of ffsite p wer aue to vehicle acciaents, extreme weather conattions (lightning,
- O hurricanes, tornadoes, and wind-driven missiles), and near-site accidents 4 involving toxic or flammable materials. W ese analyses have found such accident sequences to be important contributors to core melt f requency and/or
!O t " "S*9"*" ** (**9 *^'lY ' t liti* *"d 1***"" "*""*' " **"* th* ""C
- source term reassessment program has disregardea all externally initiated f accidents. This chapter discusses external event accident sequences that can
[ have a significant effect on source terms.5'I IO A typical impact of external events such as earthquakes or hurricanes is an area-wide loss of offsite power. For loss of of fsite power resulting f rom random component failures, it is likely that offsite AC power will be re vered in a few hours. For area-wide AC power losses caused by external
-O events, it is nach more likely that offsite power will not be recovered for 1
lO
5-2
[
.many hours, .perhaps days. The extended loss of offsite AC _ power increases greatly the probability that there will be inadequate power to run the plant's O_ safety systems.
If extended loss of offsite power were the only effect of external events
. on nuclear.. power plants, PRA analysis would be fairly straightforward since
[O; the only contribution to risk would be the fractional contribution of external events to extended loss of ' offsite power events. ' Not all' external events are that simple, however. Some external events can initiate an accident sequence, r cause the failure of n11tiple and (otherwise) independent sets of sa ety O equipment, delay safety system recovery, affeet human reliability, and/or
. impair offsite emergency response.
- External events are important in the source term context because, like O internally initiatea accidents, they can lead to large releases of radioactive materials to the environment.
5.1. Seismic Accident Sequences O
In 1975, the Reactor Safety Study (RSS) briefly considered seismic
? accident sequences, and concluded that they were unimportant - to risk. The core melt probability from seismic sequences was estimated in the RSS to be about 5 x 10 per reactor year IRASMUSSEN, ET AL. ,19751.$*' This conclusion g AL., 19791.
was subsequently criticized IHSIEH & OKRENT, 1476; LEWIS, ET Later PRAs have found that earthquake-initiated accident sequences can be ir:portant contributors to risk because of their contribution to both the O probability of core melting and significant offsite releases. Thus, the RSS is now' believed to have been incorrect in its conclusions regarding seismic risk fBUDNITZ,19831.
Estimates of core melt probability caused by an earthquake range between
'O -6 per reactor year, well within the range of dominant 2 x 10 and 1 x 10 accident sequences from internal causes. For example, seismically-induced loss of offsite power is expected to occur at a relatively mild g round acceleration value of abou; 0.20g (due to f racture of ceramic insulators),
'O only slightly larger than the safe shutdown earthquake (99E) for most nuclear
O st power plants in the U.S. In addition, recent studies of the potential for causing a so-called " double-ended guillotine break" of a primary coolant pipe O indicate that seismic events nuy be the only credible means for causing such a failure. These studies estimate widely varying p robabi li ties for an
-I to seismically induced double-ended guillotine break, ranging f rom 't.1 x la 6 x la per reactor-year iRAVINDRA, ET AL., 19R5a; RAVINDRA, ET AL., 19R5bl.
O Seismic PRAs to date have not considered several potentially important factors. First, human reliability estimates have not generally included the possible contribution of earthquake-related stress od personnel performance.
O It seems unlikely that operating personnel would be unaffected, particularly if the earthquake occurs in what is commonly regarded to be a seismically q2iet area of the U.S.
- O second, seismic PRAS to date have not considered the impact of af tershocks. Af tershocks follow the main quake, decreasing exponentially in likelihood af ter the main quake. Thus, af tershocks are most likely to occur A'tershocks occurring in the crucial few hours af ter the initial quake.
O during time periods when operator response to seismically-initiated accident sequences is required could critically affect operator performance IA7. ARM, ET AL., 19841.
O More significantly, however, the impact of af tershocks on the likelihood of containment failure has not been analyzed. There is concern about the capability of even the strongest large dry PWR containment to withstand the combined loading introduced by an aftershock occurring when the containment is t, WESLEY,
.O already pressurized f rom a severe accident IREED, 1982; CAMPBELL If a containment is likely 1991; WESLEY, ET AL.,1990; RIVARD, ET AL. , 19941.
to fail under such loading, seismic sequences which might otherwise he analyzed as being comoaratively benign might actually result in a substantial source term. A recent seismic analysis of the Maine Yankee P'4 containment O The study indicates the possible importance of such combined loads.
considered the ability of the containment to withstand an earthquake with and without the presence of the design-basis LOCA load (i.e., 55 psi internal i
f O Pressure) . Without the design-basis pressure, the containment was expected to be able to withstand a 2.lg ground acceleration. With the design-basis
'O l
) 5_4 a
pressure, the containment was expected to be able to withstand only 1.2g, considerable reduction fHASHIMarO, ET AL.,19841.
Some seismic PRAs have uncovered seismically-induced modes of containment failure that could apparently result in rather large source terms. An analysis of the Zion plant postulated that at a large ground acceleration value, soil failure under the rim of the containment basemat could lead to relative motions between the containment and the auxiliary building. This is predicted to shear all pipes running between the two buildings IBOHN, ET AL.,
1984). Such failures my open a fission product release pathway from the containment to the atmosphere.
The seismic risk analysis of the Millstone Unit 3 plant discovered a In this case, the potentially significant direct containment failure mode.
O wall supporting the polar crane is postulated to collapse, causing the shearing of main coolant and steam pipes which cass through the wall and the failure of the containment. Estimates of the probability of this failure range f rom 2 ,: In- to 4 x 10 per reactor year TNORTHEA9T UTILITTES, 1494; RAVINDRA, ET AL. ,1984; SHERON,1984; NRR,19841.
We Limerick PRA also identified seismically-induced containment failure modes. In the Mark II design, the reactor vessel is supported on a tall n At a large 9 pedestal structure and held in place by stabilizer brackets.
ground acceleration value, the stabilizer brackets are predicted to fail, permitting the vessel to sway during the earthquake and shear off the main steam lines. This creates a very large LOCA (far larger in equivalent l
diameter than any considered in plant safety analyses). We vessel is then h predicted to impact the primary containment, causing failure of the l
containment due to mechanical damage. The PRA estimated a mean probabili ty l
for such an accident sequence at is x 101 per reactor year and ranked it ninth l we O in overall likeithood for acciaents f rom all causes for Limertek.
uncertainty of these estimates was quite la rge; the probabili ty estimates per reactor year fNU9, 14911. Much of this ranged f rom 2 x 10 to 5 x la uncertainty derives from the subjective probabilities assigned to earthquakes h criginating in various source zones and the frequencies assigned to the ground Auch uncertainties are acceleration levels associated with each source zone.
O
.. 9 O 5-5 common to all seismic PRAs, and it is unlikely that the understanding of these parameters will change markedly in the near future.
O
%e Limerick analysis also identified a second seismic sequence with a potentially large source term, involving the failure of shear walls in the reactor enclosure and control structure. This failure would sever the residual heat removal system heat exchanger. pipes, causing the suopression pool to be partially drained to the level of the residual heat removal suction lines. We fission products released prior to reactor vessel failure would pass through less suppression pool water and the radioactive materials released after vessel fatture would bypass the suppression pool. Tbts O
sequence was estimated to have a mean probability of about 1. x i d per reactor year, again with very large uncertainty - the probability estimates ranged f rom 2 x 10 -15 to 6 x Id per reactor year. It was ranked 9th in O likelihood for accidents f rom all causes rNUS,1993l .
It is beyond reasonable dispute that all seismic risk analyses are subject to large uncertainties. We uncertainties derive f rom a lack of This lack of experience
- O experience with large acceleration earthquakes.
affects both the ability to estimate the acceleration level at which key systems and structures will fail, and the ability to estimate the likelihood of such ground acceleration levels. Judgments of various experts examining O the same equipment and earthquake source zones differ dramatically. Yet these judgments form the sole basis for seismic risk estimates.
Despite the inherent uncertainties in seismic risk analyses, the results n-V f rom plant specific analyses such as those discussed above for Millstone Unit 3 and Limerick demonstrate that the effects of earthquakes nust be considered.
We failure to consider the risk contribution f rom earthquakes is an important deficiency in the NRC source term reassessment program. A risk profile that l
O does not account for seismic accident sequences does not provide a valid basis j
for regulatory changes to accident source term estimates.
5.2 Accidents Resulting from Aircraft Crash ]
l
- O
' NRC licensing procedures in the area of site suitability involve a review of ai rcraf t crash likelihood. An aircraf t crash is considered a credible O
O 5-6 event if the probability exceeds 1 x 10 ^ to 1x 10- per year.
It is f requently argued that aircraft crash accidents can generally be ignored in O safety analyses of nuclear power plants. A few plant-specific PRAs which have considered aircraf t crach hazards have tended to support this position.
An Argonne National Laboratory (ANL) report on aircraf t hazards raised O qaestions about the adequacy of past assessments. W e report notes that crash probabilities calculated for many sites are of ten near to or marginally within the NRC guidelines, although the analyses may be conservative, i.e., they may overestimate crash probability IKCTP, ET AL.,19821.
O ANL suggests that some aspects of aircraft hazards analysis have received scant attention. For example, crash of large aircraft into the containment has been routinely considered, but other crash scenarios have not been considered. The crash of an airplane into the auxiliary building, for (O
example, could cause otherwise unanalyzed failures such as shearing pipes between the containment and auxiliary buildings or destruction of impo rtant safety systems.
.O offsite power and intake structures and conduits for cooling and service In addition, Fire and
,- water systems may be vulnerable to aircraf t crash.
in explosion hazards associated with aircraf t crash are seldom treated aircraft crash hazards analyses.
lO Double-shelled containments may be particularly vulnerable to aircraft I impact. ANL suggests that it may be possible to deposit a large quantity of fuel in the nnulus between the shield building and the primary containment.
- O Volatilization and subsequent ignition of the resulting fuel-air mixture would cause an explosion that inposes a severe load on the primary containment, and this load could effectively be superimposed on the impact load.
'O The source term reassessment program should evaluate the proMbility of aircraft crashes at each plant site. If the proMhlity exceeds in~ per year, the plant-specific accident sequences and containment failure nodes should be ev lu ted f r their affect on the associated source terms.
iO I
3 5-7 5.1 Sabotage-Initiated Accidents a
" Accidents initiated by deliberate acts of sabotage have not been One reason may be the difficulty of considered in PRA studies to date.
estimating the probability that a particular means of sabotage will be attempted and succeed.
O Events at Incidents involving attempted sabotage have clearly occurred.
Beaver Valley Unit 1 (involving the emergency core cooling system), Nine Mile Point Unit 1 (involving the diesel generators) , and Salem Unit 2 (also n.
" involving diesel generators) were clearly the result of deliberate insider NRC refers to these acts against vital plant equipment IPALLADINO, 14Ril.
Regardless of what they are called, such deliberate acts as " vandalism."
actions clearly have the potential for causing the unavailability of key n
" safety systems at a time of the perpetrators' choice.
The defense-in-depth concept of reactor design, involving the use of cultiple physical tnrriers to the release of radioactivity to the environment, O' is often cited as the basis for the assumed difficulty of successful sabotage.
In essence, though, defense-in-depth boils down to containment integrity once It is
' a saboteur is successful to the extent that a core melt is produced.
clear f rom the source term reassessment work to date that containments diffe O widely in their abiltty to survive severe ' accidents.
"Ihe difficulty of causing a core melt accident via sabotage has been vastly overstated, particularly for insiders (persons with authorized access One need only examine the O to vital areas in a nuclear power plant) . '
historical record concerning the precursors to severe core damage which occurred inadvertently to appreciate that the dif ficulty of causing such l'n accidents by deliberate acts has been f ar overblown, particularly for persons
" with intimate knowledge of plant operation (and perhaps security procedures as well).
There are The sabotage threat is not limited to insiders, however.
nV external sabotage actions that could pose the threat of a large radiological release f rom a nuclear power plant. One recent example is that of vehicle O
5-3
}
bonbs. An internal NRC review in 1984 concluded that vehicle bombs are a feasible form of attack against a nuclear power plant IMCCORKLE, IM41. NRC evaluations of the threat posed by vehicle bombs have already led to the
]
conclusion that " unacceptable damage to vital reactor systems could occur from a relatively small charge at close distances and also f rom larger but still reasonable size charges at large setback distances (greater than the protected aret for mst plants) ." IREHM, 19841
]
It seems relatively clear f rom the source term reassessment work to date that any accident which occurs with the containment already failed produces a Thus, sabotage that both successfully initiates a core g large source term.
melt accident and thwarts containment integrity will result in a large radiological release.
I l
l 5.4 Biofouling - Implications for Containment Heat Removal 3 i In the context of severe reactor accidents, biofouling refers to the clogging of pipes in nuclear power plants due to the growth of aquatic and marine organisms (such as Asiatic clams or blue mussels). W ere have been a 3 large number of incidents in recent years in which biofouling has resulted in partial or complete unavailability of safety systems in nuclear power plants, particularly containment heat removal systems.
D One particularly significant incident involving biofouling occurred at the Brunswick nucleat power plant. This incident resulted in the loss of the e
residual heat remval (R51R) system for Units 1 and 2. Such an event was
~3 j calculated to have a conditional core melt frequency of 7 x 10 troTTRELL, ET l
AL., 19941 1
The Arkansas Nuclear One Unit 2 facility has also experienced severe hiofouling . In 9eptember 1980, the plant was shut down after the containment 3 fan coolers were unable to meet Technical Specificiation limits for minimn water flow through the f an coolers. Af ter shutdown, additional fouling was observed in the seal water coolers for both containment spray ptrips, the seal water cooler for one safety iniection pump, and in the pump bearing and seal coolers for the high pressure safety injection service water supply lines.
l l
5-9 9 .
Examination of the adjacent Unit I facility revealed that two of four containment fan coolers in that plant were also clogged with clam shells
, (IMBRO & GIANNELLI,1982) .
Experience to date with biofooling suggests that there is a potential for causing or aggravating accidents due to multiple " external events." For
- example, consider the possible set of accidents caused b/ earthquakes or fires in combination with a loss of containment heat removal due to bioFouling. As an AEOD report on biofouling observed, "The nature of bivalve fouling in piping systems is such that it may go unnoticed, or not severely degrade system performance, until the system is called upon to function following an i nci dent." TIMBRO & GINNELLI,19921 Many nuclear power plants have experienced biofouling problems. Other
~ plants affected by biofouling include Ralem Unit 2, Oyster Creek, Millstone Unit 1, Sequoyah Unit 1, San Onof re Unit 1, and North Anna Units 1 and 2.
%ese incidents all occurred f rom 19A3 to mid-19AT f.TORDAN, 19911. In lo41, 21 of 4R nuclear power plant sites in the U.9. repo rted the existence of biofouling organisms. We utilities reported a wide variety of biofouling organisms, including clams, freshwater sponges, mussels, oysters, barnacles, and tubeworms. Other plants reported problems with mud and stit, with some claiming that this was a mre serious problem than biofouling riuBRo &
m G T A'NELLI , 199 21.
While no containment failure mdes unique to biofouling have been identi-fled, it is possible that biofouling could result in containment heat remval 3
being unavailable in accident sequences for which it would othersise be unaffected. Biofouling could be the dominant cause of the loss of containment heat remval for some plants. Such losses have important implications fo r containment failure in severe accidents.
5.5 Indirect Ef fects of Fires Fires have been examined in detail in recent PMs, and in some cases have been found to be among the dominant accident sequences. However, fire risk analyses to date have been incomplete in thel r consideration of Fire-related phenomena.
n V 5-10 Snoke-related damage is seldom considered, nor has the possible damage to safety egalpment f rom actuation of fire-protection systems been examined in O fire risk assessments. Other possible comon-mode aspects of Fires have not been treated at all (such as the effects of high humidity resulting from fires and attempts to suppress fires) . In addition, control room fires are of ten suppressed in probahtlity in fire risk analyses as a result of the assumed O success of operators working in remote shutdown locations. It has been pointed out, however, that spurious signals could be a problem in fires, and that the ability of operators to cope with such spurious signals would be limited in the alternate shutdown stations f0RAO,10A51.
- O 5.6 Design, Construction, and Operational Errors Source term assessments to date have not considered a number of types of O unusual events which could be construed to be " external events" since the accident sequences or failure modes that could result would be f rom causes
" external" to the operation of the plant. The impacts of design, construc-tion, and unusual operational errors, for example, have generally not been O addressed, yet such errors do occur. For example, valves have been installed backwards or upside down, mistakes are made in seismic design, and volds have been found in containment concrete. A common deficiency is that the plant's drawings do not acurately represent the "as-butit" plant.
l
!o The history of the nuclear industry is replete with examples of problems caused by errors in the design, construction, and operation of nuclear power plants. For example, the torus at the Hatch Unit 7 BM1 Mark I containment was j
O found in early 1994 to be crackea as a result of exposure to temperatures well below freezing resulting from the release of cold nitrogen gas from the containment inerting system IMASSAno, 19941. Should a severe accident have occurred under these conditions, the containment would have been failed f rom the outset. Another example of the importance of human errors is that the lO Palisades plant was operated for a year and a half with containment isolation valves in the purge system locked in the open position fFERIT, F:T A(,., 14951.
Any severe accident during this time period would have involved a very large
'O leaka9e.
J l
O I
O s-11 There are several instances in which holes have been drilled in containments which were left unsealed until discovered by periodic integrated O leak rate testing (ItRT) IPELTO & COUNTS,1984) . Such holes could have had a variety of effects, depending on location and size, but represent additional examples of pre-existing containment failures simply waiting for an accident to happen.
O Humn errors can lead to unusual loss of coolant accidents. In perhaps the most bizarre exanple, at Point Beach in 1992 a worker from Unit I was designated to begin draining the reactor coolant system in Unit ? which was O shut down. Instead, the operator mistakenly initiated the procedure on unit 1 which was operating at 77% of full power. The operator failed to recognize that he was in the wrong containment, despite obvious differences in the layout of the the two units, and despite the fact that the main reactor O coolant pumps were operating (causing quite a difference in expected background noise compared to a plant which is shut down) ITRAGER, 19941.
The strength of the containment could be significantly reduced by errors during design or construction. There are examoles involving debris in Q
concrete IMADSEN, 19821, voids in drywell concrete in a Mark III BWR containment IPELKE & KNOP, 19831, and degraded tendons in Pit prestressed containments IIGNATCNIS & BRYANT, 19^ 2; ifUrnINS , 19851. Such errors are not accounted for in estimates of containment strength, O
flistory shows that a lack of containment integri ty is quite comon IWEINSEIN, 1980; WEINSEIN, 19R21. Thus, the source term reassessment O Program should have considered the effect of pre-existing containment leakage paths on the source term for each accident sequence analyzed, particularly for those sequences where containment fat ture is predicted not to occur.
O O
O
n 5-12 B
5.7 End Notes for Chapter Five PRAs C 5.1 Actually, the term " external events" is something of a misnomer.
routinely analyze accidents initiated by a loss of offsite power as an
" internal" event, yet it is clearly an " external" event. By the same token, in-plant floods and fires, which are clearly " internal" events, are analyzed in PRAs as " external" events.
V 5.2 The RSS seismic analysis was later found to have a numerical error that, if corrected, might have changed the perspective of that study on the relative significance of seismically-initiated accident sequences. If the error is corrected, the seismic core melt frequency would have been about 2 x 10-6 per reactor year IMOIENI, APORTOLAKIS & CUMMINGS, 1980),
well within the range of dominant accident sequences considered in the O RSS analysis.
%e NRC has an active research program in this area. Preliminary
> 5.3 results reported in 1993 give no basis for a position that fJENKINS human
& GOLLER, performance would be unaffected by seismic events
' 19831. We Limerick PRA INUS, 19931 arbitrarily accounted for such lC seismically-related human performance degradation by increasing human error prohibilities, but failed to consider the impacts of aftershocks on human performance IAZARM, ET AL.,19941.
lO 0
- O lO 10 l
O g.13
'5.8 References for Chapter Five Q
A2 ARM, ET AL., 1994 M.A. Azarm, R.A. Bari, .7.L. Boccio, N. Manan, I.A. o pazoglou, a C. Ruger, Kafka, A Review of the Limerick K. Shiu, .7. Reed, M. McCann, & A. Review of Core-Melt Generating Station Severe Accident Risk Assessment: NUREG/cR-M9 3 Frequency, Brookhaven National Laboratory (Upton, New York),
(BNL-NURm-51711) , .7uly 1984.
M.P. Bohn, L.C. Shieh, .7.E. Wells, L.C. Cover, D.L. Bernreuter, .7.C. Chen, Lappa, O'Connell, & D.A.
.7. 7. Johnson, S.E. Bumpus, R.W. Mensing, W.7.
Application of the SSMRP Methodology to the Seismic Risk at-the Zion Nuclear Power Plant, Lawrence Livermore National laboratory (Livermore,
.g California), NUREG/CR-3429 (UCRL-53433), .Tanuary 1984.
BUDNITZ, 1981 R.7. Budnitz, " Lessons Learned f rom PRA Analysis: External Events Analysis," R ture Resources Associates, Inc. (Berkeley, California), in Pr eedings f the U.S. Nuclear Regulatory Commission Eleventh Water O Reactor Safety Research Information Meeting, U.S. Nuclear Regulatory commission Washington, D.C.), NURG/CP-3048, . January 1994.
CAMPBELL R.D. Campbe1 & WESLEY & D.A. [ Wesley, 1991 Dotential 4elsmic Structural Failure Modes Seismic Safety Margins g Associated With the Zion Nuclear Power Plant:
Research Program, Engineering Decision Analysis Company, Inc., prepared for Lawrence Livermore National Laboratory (Livernore, California),
NUREG/CR-1704 (UCRL-15140), March 1991.
"*8' C ttll' '7 **
- Mi""ri"k' "*"- ^"Sti"' 8'"- ""3""' ' 3 '-
1990-1991, A Status' O Precursors to Potential Severe Core Damage Accidents:
Report, Oak Ridge Na tional Labo rato ry (Oak Ridge, Tennessee) ,
NUREU/CR-1591, Vol.1 (ORNL/NSIC-717/V1), Tuly 1944 DRAO, 1995 "1"" ^"^1Y8'" ^"d P"'" ti "8 ' C^t*' **ti " '
"**"t"' "**Y "i"i"' " f 0 Issues from a Risk Perspective, Office of Nuclear Regulatory Research, D.C.), NURm-1115, March u .8 . Nuclear Hegulatory cornission Nashington, 1995.
EMRIT, ET AL., 1989 Dittman, R. Emrit, W. Minners, M. VanderMolen, 9. Colmar, D. Thatcher, .7.
W. Mllstead, R. Riggs, fl. Sege, P. Matthews, & L. Riani, A Prioritization 0.9.
of Generic Safety issuen, of fice of Nuclear Reactor Regulation, NURm-M11, Supplement D.C.),
nuclear Regulatory comission Nashington, No. 2,.7anuary 1985.
O O
1 5-14 O
)
HASHIMOTO, ET AL., 1984 Wesley, Conservative P.S. Hashimoto, R.P. Kennedy, R.B. Narver, & D.A.
Seismic Capacities of the Maine Yankee Reactor Containment Including and O Excluding Design Incident Pressure, Structu ral Mechanics Associates (New}M rt Beach, Calltornia) , ND/MA-16001.n l, preDared for Maine Yankee Atomic Power Company, December 1984.
H9tEff & OKRENT, 1976 T.M. lisleh & D. Okrent, Some Probabilistic Aspects of the Seismic Risk of Nuclear Reactors, University of California at Los Angeles (Los Anqeles, O Ca11 t o rnia) , UCLA-ENG 76113, December 1976, cited in " Analysis of PAR's without seismically Qualified Auxiliary Fee &ater Systems," IJ.9 Nuclear Regulatory Commission (Washington, D.C. ) , 8 August 1990, attachment to memrandum f rom R.J. Mattson to D.G. Eisenhut, 9 August 1990 O HUDGINS, 1995 C. Hudgins, "Farley Tendon Problem Blamed on Nater as NRC Mulls Generic
IGNATONIS & BRYANT,1992
^. tonatonis & J.C. Bryant, " Degraded Tendons in the Shield Wall," is.s.
0 Nuclear Regulatory commission, Region II office of Inspection and Enforcement (Atlanta, Georgia), PNO-92-51, 29 April 1982.
IMBRO & 'iIAWELLI,1992 E.V. Imbro & J.M. Giannelli, Report on Service Water System Flow Blockages by Bivalve Mo11uses at Arkansas nuclear One and Rrunswick, Office for the Q Analysis anu Evaluation ot operational Da ta , IJ .s . Nuclear Regulatory Comission (Washington, D.C.), AE'V Report No. AEOD/C202, February 1992.
.TENKINS & GOLLER,1991 J.P. .Tenkins & M.R. Goller, " Reactor Operator Performance Under the Stress of a seismic Event," office of Nuclear Regulatory Research, is.9. Noctear O Regulatory Commission (Washington, D.C.), presentation to the Seismic Risk and Heavy Industrial Pacilities Conference, San Francisco, California, 12 May 1991
.TORDAN, 1991 Q E.L. Jordan, "Conmon-Mode Valve Failures Degrade Surry's Recirculation 9 pray Subsystem," Office of Inspection and Enfo rcement, IJ .9 . Nuclear Regulatory Comission (Nashington, D.C.), IE Information Motice No. R3-44, 11 July 1991.
C.A. x t, H.C. Lin, J.n.
O Evaluation of Aircraf t Crash Hazards Analyses for Nuclear Power Plants, Argonne National Laboratory (Argonne, Illinois), NUR EG/CR- 29 59 (ANL-CT-91-12), June 1992 O
O
I 5-15 g
LEWIS, ET AL., 1978 II.W. Lewis, R.7. Budnitz, H.7.C. Kouts W.B. I.owenstein, W.O. Rowe, F. von Ilippel, & F. Zachariasen, Risk Assessment Review Group Report to the U.S.
Nuclear Regulatory comission, Ad soc nisk Assessment Review Group, U.s.
O Nuclear Regulatory Comission 04ashington, D.C.), NUR m/CR-64M , September 1 1978 MCCORKLE, 1994 Memo rs.idum f rom G.W. McCorkle to R.F. Burnett, "Sem!-Annual Design Basis treat Review No. 4," Of fice of Nuclear Materials Safety and Safeguards, O U.S. Nuclear Regulatory Commission 04ashington, D.C.), 19 7anuary 1994 MAOSEN, 1992 G.L. Madsen, " Deb ri s in Floor Concrete," U.9. Nuclear Regulatory Commission, Region IV Office of Inspection and Enforcement (Da llas ,
Texas), PNo-Iv-R2-39, 21 October 1982.
O
- A99ARO, 1984 S.A. Massaro, editor, "Through Wall Crask in Vent Header of BWR Containment Torus," in Power Reactor Events, Office for the Analysis and Evaluation of Operational Data, U.S. Nuclear Regulatory Commission 04ashington, D.C.), NUREG/BR-0351, Vol. 6, No. 1,.7dne 1994.
g MOIENI, APOS%IAKIS, & CUMMINGS,1980 P. Moloni, G. Apostolakis, & G.E. Cunmings, interim Report on Systematic Errors in Nuclear Power Plants Seismic Safety Margins Research Program, (Livermore, California),
Lawrence Livermore National Laboratory NuRea/CR-1722 (uCRL-1s274), october 19sa.
O NORDIEAST UTILITIES, 1984 Utilities, M111 stone-3 Probabilistic Safety _ Study, Northeast Northeast Utilities (Berlin, Connecticuu , menrinent no. 4, 4 apru Ma.
NRR, 1994 0 office of Nuclear neactor Regulation, Pinal Environmental Statement related to the operation of Millstone Nuclear Station, Unit No. 1, U.9 D.C.), NUREI3-1064, December Nuclear Regulatory Commission (Washington, 19A4.
i NUS, 1993 O NuS Corporation, Severe Accident nisk Assessnenti Limerick cenerattna 1161, Station, NUS Corporation (Rockville, Ma ryland ) , MUS Report No.
prepared for Philadelphia Electric Company, April 1991, t par.t.ADINO, 1991 Markey, 7 Letter f rom NRC Chairman N.7. Palladino to Con]ressmn E.7.
O rebruary 1o91.
PELME & KNOP, 1991 P. nelke & R. Knop, " News Media Interest in Drywell Volds, Former Employee Allegations," U.9. Nuclear Regulatory Comission, Raqion rtt~! ofOccomber flee of >
Innpeetion and Enf o rcement (Chicago, Illinois), PNO-41-191, ,
O.. 1991 i t
O
5-19 0 .
PELTO & COUN W , 1984 P.J. Pelto, & C.A. Counts, " Reliability Analysis of Containment Isolation Systems," Pacific Northwest Laboratory (Richland, Washington), presentedVirginia, Arlington, g
at the Second Workshop on Containment Integrity, V 11-15 June IW4.
RASMUSSEN, ET AL., 1975 al., Reactor Safety Study: An Assessment of Accident N.C. Rasnjssen, et Risks in U.S. Comercial Nuclear Power Plants, 0.9. Nuclear Regulatory Commission (Washington, D.C.), NURm 75/014 (WASH-1400), October 1975.
RAVINDRA, ET AL.,1994 M.K. Ravindra, R.H. Sues, R.D. Kennedy, & D.A. Wesley, A Program to to Determine the Capacility of the Millstone 1 Nuclear Power Plant Withstand Seismic Excitation Above the Design SSE, 9tructural Mechanics Associates (Newport Beach, California) , NE/9MA ?.6401.01-R?, prepared for O Northeast Utilities, November 1984.
RAVINDRA, ET AL., 19ASA Banon, Probability of_
M.K. Ravindra, R.D. Campbell, R.D. Kennedy, & H.
Pipe Failure in the Reactor Coolant 1400 of Westinghouse IMR Plants, Vol.
Structurai 3r Guillotine Break Indirectly Induced by Earthquakes, O Mechantes Associates (Newport Beach, calt rornia) , prepared for tawrence Livermore National Laboratory (Livermore, California), NURm/CR-3660, Vol.
3 (UCID-19998, Vol. 3), February 1985.
RAVINDRA, ET AL.,1985B Ranon, probability of H.
M.K. Ravindra, R.D. Campbell, R.D. Kennedy, &
O Pipe Failure in the Reactor Coolant Loops of Combustion Engineering RVR Plants, Vol. 3: Double-Ended Guillotine Break Indirectly Induced by Earthquakes, Structural Mechanics Associates (Newport Reach, California),
Laboratory (Livermore, prepareu tor Lawrence Livermore National California), NURFXi/CR-1341, Vol.1 (UCRL-51501, Vol.1), January 1095.
O nEEn, 1982 Mountain View, Letter f rom J.'V. Reed (Jack R. Benjamin & Associates, Inc.:
California) to W.A. von Riesemann (Sandia National LAborato r ies t Albuquerque, New Mexico), " Proposed Research Program to improve the solsmic Analysis Procedures for Drohibilistic Rink Assessment of Nuclear Power Plants," 2 April 1992.
REHM, 1994 Mennrandon f rom T.A. Rehm toNuclear the NRC Commissioners, Regulatory Com ission" Nashington, Weekly Information Report," Enclosure E, 0.9 D.C.), 27 April 1994.
O RtvARn, ET AL., 1984 G riesmnyer, F.E. Haskin, J.B. Rivard, V.L. Reh r , R .G . Easter 1tng, J.M.M.P. Sheman, A.R. Talg, &
S.W. Hitch, A.M. Kolaczkowski, R.J. Lipinski, A,J. Wickett, Identification of Severe Aceldent Undertaintien, 9andia National tatoratories 7
(Albuquerque, New Mexico), NtlR EG/CR-1440 (SAND 91-1689) , September 1994.
O
5-17
.D I
SHERON, 1984 Memorandum f rom B.W. Sheron to L.G. Hulman, " Millstone-1 DES: Containment Failure Matrix and Source Terms," with attachments, U.9 Nuclear jg Regulatory Conunission (Washington, D.C.),14 May 1984.
TRAGER,1994 E.A. Trager, Human Error in Events Involving Wrong Unit of Wrong Train:
i Special Study Report, Ottice tor the Analysis and t;valua tion ot
~
D.C.),
i Operational Data, U.S. Nuclear Regulatory Consnission (Nashington, lg AE00 Report No. AROD/S401, January 19Rd.
WEINSTEIN, 1980 M.B. Weinstein, " Primary Containment Leakage Integrity: Availability and 5, l
Review of Failure Experience," Nuclear _ Safety, Vol. 71, No.
September-October 1980 O WEINSTEIN,1992 M.B. Weinstein, " Integrity Failure Experiences With Reactor Containments,"
American Nuclear Insurers (Farmington, Connecticut), in Proceedings of the Workshop on Containment Integrity, Sandia National Laboratories (Albuquerque, New Mexico), N1rROG/CP-0013 (SAND 62-1659), October 1992.
O WESLEY, ET AL., 1980 D.A. Wesley, R.D. Campbell, P.S. Hashimoto, & G .9. Hardy, Conditional Probabilities of Seismic Induced Failures for Structures and Components for the Zion Nuclear Generating Station, Structural Mechanics Associates e (Nwport Beach, California), prepared for Pickard, Lowe & Garrick, inc.,
n SMA-12901.02, October 19A0.
y lO l
f i Q
l l
0 0
O i
CHAPTER SIX Neglected Plant Designs and Desiqn Variations SUMMARf
%e NRC's source term reassessment program falls to take into account the vast differences among plant designs. These
] differences are significant: neither source terms ror overall risk assessment for one plant can be assuned to be applicable to any other. We concept of analyzing only a fw accidents at a handful of " reference plants" and extending the results to all other plants is fundamentally flawed.
Some of the differences that should be taken into account in pressurized water reactors include: the number of coolant loops, the location and height of coolant pipes, the design of steam generators, the volume of water in secondary loops.
the lack of pilot-operated relief valves, or other engineere:1 O safety features, on some plants, and diff erfng containment designs. In boiling water reactors, design differences that should be examined are: containment designs, including differences in the pedestal area and containment cooling systems.
n other factors that should be considered include: construction materials, operato r training p rog ram, and emergency procedures.
6.0 Intcoduction
'O I
Both the U.9 Nuclear Regulatory Comission f9ILRERBFM , ET AL., 10951 and the nuclear industry's IDCOR program f fMTANA, ET AL., 10941, used
" reference plants" in their assessments of source terms. The regulatory C justification for using revised source terms would depend significantly on whether the ref erence plants chosen for detailed analysis fairly represent all domestic light water reactors. Factors that need to be comidored include not only the possible accident seglences but the different modes and ro ta tive O likelihood of containmnt failure, and differences in design and performance of safety systems.
As of September 1995, there were 95 nucinar power plants licensed for lO operation in the united states. Aoproximtely as-la adat tional plants may So
'i l
l 0
L
6-2 g.
itcensed in the next decade. The lack of standardization armrq these plants means th9t there are only broad similarities among them, even within the same plant class and, indeed, available probabilistic risk assessments show widely O dif fering f reglencies for identical accident segjences at different plants.
This chapter examines the degree to which the reference plants represent the other rusclear power plants in operation and under construction.
O This discussion is illustrative; we have not attempted to search for and .
document every technical difference between the reference plents and the remainder of the LWR population. There are many such differences that could significantly aff ect source term estimates.
6.1 Differences in Nafety System and Reactor Coolant System Design We only PWRs included in NBC's source term review were Westinghouse O plants, including the three-loop Surry design and the four-loop designs at Sequoyah and Zion. No two-loop Westinghouse plants were examined, although there are a number in operation. Nor were there any B&W or CE plants anal-yzed. There are similar shortcomings in the soleiction of FNRs; there are O styngricant piont_to-plant differences anorg nwRs which are not rmresented in the sou rce term reassessment. This section explores some of the desiqn dif ferences amorg 11.S. light water reactors as they might affect estimates of source terms.
O 6.1.1 Dif ferences in WR RCS Geometry There are several differences among R&W, CE, and Westinghouse PWRs which O may have important source term imiteations. For examle, in a&W olants the lorg hot leg pipes (which carry coolant away f rom the reactor vessel to the steam generators) rise above the steam generators and enter them f rom the top, whereas in CE and Westinghouse PWRs the hot leg pipes run short, horizontal O distances between the reactor vessel and the steam generators, and enter the steam generators f rom the bottom (see Figure 6-1) . in addition, all 05W plants and most CE plants have two hot leg loops and four cold leg loops with two steam generators. This is not a configuration used in Westinghouse O plants, all of which Mve one steam generator per loop with the number of v
O
I s-3
- D loops varying f rom two to four (see Figure 6-2) . These differences in RCS geometry could . be igortant to fission product retention in the RrS, decay .
! heating in the hot leg piping (which has been postulated in extrane cases to O lead to hot leg failure), as well as possibly causing differences in response to pressure surges originating in the reactor vessel (f rom, for exa@le, l
in-vessel steam explosions) .
l h There are other igortant differences as well. For exagle, B&W reactors are equipped with an entirely different type of steam generator than the ;
l Westinghouse and CE designs. The latter two PWR designs eglof inverted U-tube steam generators of varying designs, whereas the R&W PWRs employ a
-Q orwe-through steam generator (CyrSG) design (see Figure 6-3) . This difference in design could lead to differences in fission product deposition in the steam generators and to different responses to AM sequences.
The water volume in the secondary side of the R&W system is also mch f l smiler than in either the CE or Westinghouse designs. This could lead to a mch mre rapid response to temerature and pressure changes, particularly in l accident sequences involving primry-to-secondary heat imbalance (such as in
.O comlete-loss-of-feedwater, station blackout, and Am sequences),
i 6.1.2 CE Plants Without DORVs l
O sone cortustion Engineering owns lack pilot-operated relief valves (PORV) ,
on their pressurizers. We lack of PC7tVs will affect the response of such plants to a variety of accident scquences because primry system presure reitef can only occur through the safety valves, which are set at a higher
'O presure than are PORVs. Amng the accident sequences, and their associated source terms, potentially affected by the lack of PORVs are comlete loss of feertsa tar, smil LOCAs, station blackout, AM, and steam generator tube ruptu res.
In addition, the lack of PORVs (cortined with a low shutoff head f or the !
i high pressure iniection pumpn) precludes the use of so-called " Feed md bleed" j cooling (une of high pressure injection and the PORV to form a continuous (
O cooling cycle when feetwater is lost). This prohibly means that CE Ms !
[
=
l l
0
V l 4-4 without PORVs will have a higher f rquency of core melt's due to complete loss l of feedwater than plants with PORV9 where feed and bleed cooling might he possible. However, in no plant has " feed and bleed" been shown to be I
effective in core cooling.
a Transient-induced t.0CAs involving PORVs are relatively likely accident sequences at plants with PORVs. Such sequences are, of course, not possible at plants lacking these valves. ,
6.1.3 Differences in RCS Geometry Among CE Plants
) There are many significant dif ferences in the configuration of CE reactors tMt could affect the evolution of accidents. Most CE PWRs, for example, are of the two hot-leg /f ou r-cold leg design with two steam generato rs. Maine Yankee, however, is of a three loop design (three hot legs and three cold legs) with three steam generators. Similarly, while cost PWRs have in-core instrument tubes which penetrate the bottom head of the reactor vessel, some CE PWas have top-entry in-core instrument tubes (e.g. , Calvert Cliffs Units 1 and 2 and Fort Calhoun) (BME, 19781 (see Figure 4-4). Maine
) Yankee has bottom-entry instruments. The procence of bottom-entry instrument tubes could affect the mode of reactor vessel failure in core melt accidents.
Recent probabilistic risk assessmnt studies for plants with such instrument
' tubes hive identified a localized vessel failure modo involving failure of
) partial penetration welds around the bottom-entry in-core instrument tubes, followed b/ ejection of the tubes and release of the molten core debris through the hole. Combined with the different RCS geometry alluded to above, Maine Yankee must ha considered an anomaly amng CE PWR designs. On this i
) basis, a separate source term analysis should be undertaken for Maine Yankee.
Since the phemmenon of high pressure mit eiection (HPME) is dependent on a localized vessel f ailure modo, CE plants with top-entry in-core
) instrument tubes may b) less likely to experience IIPW than other reactors.
'io far as w> have been able to determine, there Mvo been no analyses of the phemmena involved when a reactor vessel falls via complete failure of the botton head at high pressure. Such an analysis is necessary, in our view, in I o rder to adeglately understand reactor cavity phemmena in CE PN9s with top-entry in-core instruments.
j 6-5 6.1.4 Differences in RCS Geometry Among B&W Plants Most B&W PWRs are of the so-called lowered-loop design. The navis-Besse
[
PWR, however, has a " raised-loop design" (see Figure 6-5) . This difference could af fect fission product transport within the reactor coolant system (RCS) and accident sequence event timing for some types of accidents (i.e., those In addition, O involving natural circulation of coolant th rough the RCS).
Davis-Besse has a large dry steel containment with a secondary concrete shield building. No such plant has been analyzed in the source term reassessment program.
O An analysis of the potential for different modes of containment failure is particularly imp)rtant for Ibvis-Besse, especially modes involving isola-tion f ailure. An example of such a failure node would involve the vacuum-relief valves at Davis-Besse which are not present on large, dry, concrete PWR O These containments (except the subatm3 spheric containments such as 9urry).
dif f erences make rhvis-Besse significantly dif f erent f rom other R&W plants and necesitates a separate source term study.
O 6.1.5 Differences in Reliability of Safety Systems
'There are a large number of other variations in design acnng U.9. light water react rs. For example, twenty diff erent aux 111ary feedsater system O configurations have been identified by the Accident 9equence 7. valuation Program (ASEP) . Service-water systems were found to be essentially plant-the specific. Without accounting for the design of service water systecn, g preliminary ASEP assessment identified 79 PWR plant grourn (out of 72 PMts examined) and 15 BWR plant groups (out of 11 BWRs examined) . As the NRC's ASEP program technical monitor ohnerved, "This is a dramatic evidence of the lack of standardized designs in the U.9." fEtXI, 19941 O Some of the design differences can drastically aff ect the oroh1hility of the loss of important safety systems. For example, some B&W PWRs have a commn pipe leading f rom the refueling water storage tank to the ECm and ntainmnt spra/ ptr,ps. There is a mnually-operated valve in this pipe.
If
!O
! this valve id left closed and an accident occurs, the low-pressure and high-r lO l
m 6 e; V
pressure injection and the containment spray pmps will quickly fail as a result of the pumps' drawing suction on a closed valve. The PRA for the g Oconee plant found that this resulted in a relatively high probability of LOCA sequences involving both failure of ECCS injection and containment spray injection IMOIB, ET AL., 19911.
s -called " interfacing toCA" or " Event V" sequences have been recognized O ever since WASH-1403 as important PNR secpences because they lead to a loss of coolant outside the containment and a bypass of the containment for fission product transport. A review of safety system alignments at a limited number f p ants e ndu tea dv ak aldge nati nal tab ratory identified is different
'O
" Event V" configurations at the 44 PWRs surveyed f HEDDLESON,19831 (see Figure 6-6). Of course, each " Event V" configuration will have a dif ferent probability, thus the risk significance of this accident secpence is different
! f r each coneiguration.
O 1
6.1.6 Other Design Variations Were are a host of other design varia ttom. Among them are the O
- fallowing examples
4 A. Unlike mst ice condenser PWRs, D.C. Cook has no passive i
i upper-head iniection system.
40 l
B. %e Oconee reactors have no min steam isolation valves ILAAKSONEN f. SHERON, 19921.
i
- C. San Onof re Unit 1 can use high pressure injection to provide i
feedsater to the steam generators via a cross-connect between
- O the systems iHEnoLEs0N,19931.
D. At Peach Bottom, the safety valves discharge to the drywell rather than directly to the suppression pool IPECO,19821.
] E. 0/ ster Creek and Millstone Unit I hwe no jet pumps fLARSON,
- O 1994).
F. Nine Mile Point Unit 1 and Oyster Creek have separate dedica ted containment spray systens which can spray the dryw11, the suppression pool, and the containment dcunconers (
finaNER, 19921.
O lO
6-7 G. Turbine bypass capability of some plants is very limited.
For exanple, the turbine bypass capability of St. Incie and Palisades is only 5% of full f1w ILOBNER, 19921. Such 3
plants are nuch mre likely to lift the secondary side relief v valves in mny types of accidents.
6.2 Differences in Containment and Containment ESF Design c The three PWR containments evaluated in the NRC source term J reassessment program were of the large dry concrete subatmospheric, large dry concrete atmospheric, and steel ice condenser designs. There are other PWR containment designs which do not fall within this umbrella. In addition, there are design variations a ong RWR containments that were not studied.
g
- This section provides a discussion of variations in containment design and in -
containment engineered safety features (ESFs) that could have source term impit cations.
l lO variations in Ice condenser containments and Containment ESF Designs 6.2.1 With the exception of D.C. Cook, ice condenser plants all have steel containments with secondary containments and annulus filtration systems.
Units 1 and 2 of 0.C. Cook employ steel-lined, reinforced-concrete ice condenser containments but have no secondary containment (and, thus, no l annulus filtration system) . The containment response to static and dynamic loads and the containment failure mde, loca tion, and pressu:e could all be expected to be different for 0.C. Cook than for the other ice condenser
! containnents.
Th ice condenser containments have varying design pressures. The O containments at the 9equoyah and 0.C. Cook plants have a design pressure of 12 psig; other ice condenser containments have a design pressure of B psig.
These design pressure differences could be exnected to show up as well in containment failure pressure estimtes.
Very little detail concerning the placement or reliability of hydrogen igniters is available for the ice condonner containments. Ilydrogen igniters are needed for these plants to help cor. trol the pressures achieved odring hydrogen burns in nevere accidents. Differences in the placement of hydrogen ;
l l
l lO
. - = . . -
64
)
igniters could affect the progression of accident sequences by affecting the timing and magnitude of hydrogen burns. Since the hydrogen igniter systems are manually initiated, differences in their reliability should be expected.
Failure of the hydrogen igniters to operate in non-station blackout sequences would represent a potentially significant accident sequence for ice conrtenser ,
plants.
) Containment spray configurattom also vary. 0.C. Cook has sprays in ,
both the upper and lower compartments IIMP.C, 19921. Other ice condenser PWRs l
have sprays only in the upper co@artment. The impact of spray operation in j the lower compartment at 0.C. Cook during a variety of severe accidents should be analyzed since such operation could have an igact on the likelihood of hydrogen burns.
Catawha has no vacuun breakers; it relies instead on a " Containment
)
l pressure Control system" to prevent operation of the containment sprays when the containment pressure is less than 0.25 psig fDPC, 19821. Other ice l
condenser containments rely on vacuum breaker devices to prevent suh3tmos-pheric pressures in the containmnt. Such vacuum breakers rey represent a containment bypass path for fission product release to the environment; to l date this problem has not been addressed.
i l
The containment spra/ return systems (by which spray water is returned
. f rom the upper compartment of the containment to the sump in the lower compartmnt) vary among the ice condenser plants. Catawha's spray water return system consists of six 9-inch diameter drain lines fDPC,19911, whlie 9cquoyah's spray water return system consists of two 14-inch diameter drains ICARL90N, ET AI.. , 19911. It is possible that the differences in the number of l
drain holes co 1d affect the probability of the holes being lef t plugged af ter a refueling outage.
) 6.2.2 targe Dry Steel Containments
%ere are a number of PWRs with largo dry steel containmnts. No plants with such containments were included in either the NRC or IDCOR source term evaluattom. PWRs with large dry stool containments also have concrete
6-9 O
secondary containments and annulus filtration systems fKotACZK0WGKI, ET AL. ,
19831. %ese plants also have a vacuum relief system which represents a large containment isolation failure source not present on most other PWRs. The g Dicie Unit 1, a plant with a large dry containment failure pressure for 9t.
steel containment, has been estimated at 95 psig b/ Ames Laboratory TGREIMNN, FANOUS & BLUHM, 19841. This is relatively low compared with most large dry
" "'^1""*"'"-
0 6.2.3 BVR Mark I Containment Design Differences Am ng BWRs with Mark I containments, all have steel containments except O the two-unit Brunswick plant. Brunswick Units 1 and 2 are of a steel-lined, reinforced-concrete construction. This appears to be uniqJe in the Mark I containment class, and has been unexplored in the severe accident arena. It is not known, therefore, what the modes, locations, and pressure of contain-g' ment failure will be for the Brunswick reactors. There may be differences in seismic response of the Brunswick containments compared to other Ret Mark I l plants which could present unique failure mdes.
'O 6.2.4 RVR Mark II Containment Desiqn Differences Each of the Mark II containmnts uttitzes a unique design. 14hile mst f the Mark Its have stool-lined concrete containments, the WPPSS Unit 2 plant O cmploys an elliptical steel containment with a concrete floor inside the steel rhell ISERKI7.,19931. %e Limerick and LaSalle Mark II containments have flat bottoms. To the best of our knowledge, the containment failure characteris-tics of the WPPSS Unit 2 Mark II containment are unknown .
In addition, there are wide and imortant variations amng the Mark Its in the design of the pedestal area of the containment; these variations could have signifi ant impacts on containment behavior. For examle, the Rhoreham O design has water below the diarhragm floor (bilow the reactor vessel in the pedestal region), and four large downcomer pipes directly belw the reactor vessel. We Shoreham PRA concluded that molton core debris would travel through the dwncomers upon vessel fatture and be quenched in the suppression O pool water f SAI,19R41.
O
)_ 6-10 In contrast, while they also have water in 'the pedestal region, the WPPSS Unit 2 and Limerick plants have no downcomers in the pedestal area
) ISERKIZ, 1983; PECD, 1982) . Lastly, the LaSalle design has a solid concrete plug in the pedestal region below the diaphraepn floor IGREENE, 1994; CECO, 19841. It seems obvious that Mark IIs with concrete plugs in the pedestal will have significantly different accident seg2ences f rom those with water
) under a diaphragm floor. For example, core-concrete interactions should be drastically diff erent.
6.2.5 Mark III Containment Design Differences D
Were are at least two major containment design groups in the Mark III containment class. The Grand Gulf class plants have steel-lined reinforced concrete containments with no secondary containment while the GESSAR-II class p plants have steel primary containments with concrete secondary containments fGESSAR-III . The differences in failure pressure and failure location have l not been factored into the NRC source term assessment.
)
L Even armng "similar" plants there are significant design differences.
For example, Grand Gulf has a containment spray system while River Bend does not. Rather than having the containment sprays which are present at Grand quif, River Bend has containment air coolers which serve to limit post-D accident containment pressure. We relative reliability of air coolers versus c sprays is not well known and should be examined. There would also be reason to expect differences in performance regarding Cission product transport; such dif ferences would, in general, be expected to be nost significant for D sequences involving some degree of suppression pool bypass to the outer containment air space. With a steel containment and air coolers instead of sprays, River Bend represents a significant departure from the Grand Gulf I design and should be examined in its own right.
O 6.1.6 Variations in THR Reactor Cavity twsiqns Phenomena af fecting containment perfonnance are discussed in Chapter 9 3 of this report. We impacts of such phenomena as ax-vessel steam explosions and high pressure melt ejection are dependent upon the detalis of the reactor cavity geometry.
O
6-11 O
'there are considerable variations in reactor cavity designs. These differences are addressed in part in an NRC report on containment emergency q
stynp performance ISERKIZ, 1983). Suffice it to say that even among the NRC reference plants there are variations. zion and Sequoyah share some similarities, but at Zion the cavity opens up into the upper part of the containment, whereas at Sequoyah the cavity opens up into a dead-ended portion of the lower compartment at this ice condenser containment. At Surry, the
,3
" reactor cavity is very diff erent. Exanples of various PWR reactor cavity configurations are provided as Figure 6-7 Another area of difference among PWR reactor cavities is the so-called
" dry cavity" design. Most PWR reactor cavities are aligned so that water f rom the containment sprays (and from many RCS break locations) would drain to the reactor cavity. Some PWRs, however, are designed with barriers surrounding the cavity so that the cavity remins dry. The dry cavity design has certain O implications for core-concrete interactions, high pressure melt ejection, and the rate of containment pressurization due to steam. The dif ferences in severe accident response between " wet cavity" and " dry cavity" designs deserve a more complete treatment than they have been accorded thus far.
O Another difference in reactor cavity geometry arises in some Babcock &
Wilcox PWRs. The Arkansas Nuclear One Unit 1 plant has the containment spray sump in proximity to the reactor cavity. In the PFM for this plant it is O noted that this may lead to f ailure of containmnt sprn/s in recirculation following reactor vessel failure due to Fouling of the spray pumps with metal chips caused by the interaction of core debris with water in the reactor cavi ty f MOLR , ET At.. , 19921.
Thus, the f ailure of containment spra/s in recirculation (intnediataly following vessel failure, a time of high aerosol loading in the containment) could be much mre itkely at such plants. It would therefore seem appropriate to examine the impact of such containment spray rect reulation failure on important post-vessel f ailure phenomena such as ex-vessel steam explosions,
! high pressure melt ejection, and containment prennurt:ation f rom large t
L hydrogen horns at vessel failure.
g i
i 6-12 O
i 6.1 Conclusions I l
We NRC's source term reassessment has pubilshed results for only five O
reactor designs. As a result of the lack of standardization among the 95 licensed U.S. nuclear plants there are only broad similarities acnng them and If and when reliable, generic conclusions on source terms cannot be drawn.
O accurate moaels for source term analysis are developed, they will have to be applied separately to each operating plant to determine applicable source terms for risk analysis.
O The differences among plants can be seen in virtually every aspect of reactor design. Among the factors that need to be considered in individual source term analyses are the following:
O o %e numbers of loops and steam generators in P.VRs, o The length and height of coolant pipes, o We design of steam generators, o The availability of pilot-operated relief valves (FORVs),
O o We location in "the pressure vessel of instrument tube penetrations, o The design of containments, including the presence of secondary containments, annulus filtration systems, maximum design pressures,
,.O and the location and reliability of hydrogen igniters, o te design, availability, and reliability of various reactor systems including auxiliary feedwater systems, service-water systems, jet -
pumps, containment sprays, and main steam isolation valves,
.O o Turbine bypass capability, o Other containment differences such as the design of the pedestal region, the availability of doancomer pipes, the presence, or lack thereof, of water in the pedestal area, and the design of reactor l
cavity designs, v A comprehensive examination of plant-to-plant design and operational dif f erences must be made and they must be taken into account in any source term reassessnent.
'O 1
i i O.
6-13 O
6.4 References for Chapter Six 0 RG&E, 1978 Baltimore Gas & Electric Cortpany, Calvert Cliffs Nuclear Power Plant, Units 1 and 2, Final Safety Analysis Report, Baltimore Gas & Electric Company (Baltimore, Maryland), January 1971, revised 11 December 1978.
CARLSON, ET AL., 1991 D.D. Carlson, W.R. Cramond, 7.W. Hickman S.V. Asselin, and P. Cybulskis, O Reactor Safety Study Methodology Appl [ cations Program: Sequoyah El Ps'R Power Plant, Sandia National Laboratories (Albuquerque, New Mexico),
NURD3/CR-1659, Vol.1 (SANDP0-lR97/1 of 4), April 1991.
CECO, 1984 Commnwealth Edison Company, LaSalle County Station Updated Final Safety M, Q Analysis Report, Cormonwealth Edison company (cnicago, Illinois), Rev.
April 1984.
DPC, 1982 Duke Power Company, Catawba Nuclear Station Final Safety Analysis Report, Duke Power Company (Charlotte, North Carolina), Amendment 27, Revision 5, O 11 June 1992.
Etr3, 1934 2.1.1, U.S.
T. Eng, " Plant Catego rization," NRC Severe Accident Issue Nuclear Regulatory Commission (Washington, D.C.), draft 52, 30 April 1484.
n* EnNTANA, ET AL., 1984 H. A. Mitchell, S.V. Asselin, E.P.
M.H. Fontana, A.R. Buhl, E.L. Fuller, St roupe, J.C. Carter, K. A. Meyer, and R. Satterfield, Nuclear Power Plant Technology Response to Severe Accidents: IDCOR Technical Summary Report, for Energy Corporation (Knoxville, Tennessee), prepared for the Atomic Industrial Forum's IDCoR Program, November 19a4 O
GESSAR-II General Electric Company, BWR/A Standard Plant Probabilistic Risk Assessment, General Electric Company (San Jose, calttornta), Rev. 2, December 1982 (proprietary; some portions released via Freedom of O Inforration Act) .
GREENE, 1984 l S.R. G reene, Realistic Simulation of Severe Accidents in WRs - Computer Modeling Requirements, oak H1dge National La bo ra to ry (UaK Kidge, Tennessee), NUREG/CR-2940 (ORNI/TM-8517) , April 1984.
O GREIMANN, FANouS & BLUnM, 1984 L. Greimann, F. Fanous, and D. Bluhm, Final Report, Containment Analysis (Ames, Iowa),
Techniques: A State-of-the-Art Summary, Ames Iaboratory prepared tor Sandia National Laboratories ( Albuquerque, New Mexico) ,
NUREU/CR-3653 (SANDR3-7461), March 1994.
- O O
6-14 O
?
HEDDLESON, 1983 F.A. Heddleson, Suntnary Report on a Survey of Light-Water-Reactor Safety Systems, Oak Ridge Na tiona l Labo rato ry (Oak Ridge, Tenneccee),
NURE/CR-2069, Revised (ORNI/NSIC-191/Rl), April 1983.
- O IMEC, 1982
! Indiana & Michigan Electric Company, Final Safety Analysis Report Update, Donald C. Cook Nuclear Plant, Indiana & Michigan Electric Gompany, July 1982.
O KOIAC2K04 SKI, ET AL. ,1983 A.M. Kolaczkowski, P.B. Bleiweis, M.T. Drouin, and W.L. Ferrell, Interim Report on Accident Sequence Likelihood Reassessment (Accident Sequence Evaluation Program), Sandia National Laboratories (Al buquerque , New Mexico), draft, February 1993.
O KOIB, ET AL., 1981 G.J. Kolb, S.W. Match, P. Cybulskis, and R.O. Wooten, Reactor Safety Study Methodology Applications Program: Oconee 53 PdR Power Plant, Sandia National Laboratories (Albuquerque, New Mexico), NURG/CR-1659/2 of 4 (SAND 80-lR97/2 of 4), January 1991.
Q' KOLB, ET AL. ,1992 Bell, N.L. Brisbin, D.O. Carlson, S.W.
G.J. Kolb, D.M. Kunsman, B.J.
Hatch, D.P. Miller, B.J. Roscoe, D.W. Stack, R.B. Worrell, J. Robertson, Murphy, R.O. Wooten, S.H. McAhren, W.L. Ferrell, W.J. Galyean, and J.A.
Interim Reliability Evaluation Program: Analysis of the Arkansas Nuclear One -- Unit 1 Nuclear Power Plant, Sandia National Laborato ries Q (Albuquergae, NeW Mex1Co) , NUdW/CH-z787 (SANDR2-097R) , June 1992.
IAAKSONEN & SHERON, 1982 J.T. Laaksonen and B.W. Sheron, Evaluation of PWR Response to Main Steamline Break With Concurrent Steam Generator Tube Rupture and U.S. Nuclear Regulatory Commission (Washington, D.C.),
Small-Break LOCA, g
' NURW-WW , December 1982.
LARSON, 1984 (Idaho Falls, J.R. Larson, System Analysis Handbook, EG&G Idaho, Inc.
Idaho), NURFD/CR-4041 (EGG-2354), December 14R4.
n V LOBNER, 1982 P. Lohner, Nuclear Power Plant Damage Control Measures and Design Changes, Science Applications, Inc. (LaJolla, California), prepared for Sandia New Mexico), NURDG/CR-2585 National Laboratories (Albuquerque, (SAND 82 7011), May 1992.
O' PECO, 1982 Philadelphia Electric Company, Peach Bottom Atomic Power Station Updated Final Safety Analysis Report, Philadelphia Electric Company (Philadelphia, Pennsy>vania), July tw 2 O
O
6-15 g
SAI, 1994 Science Applications, Inc., Probabilistic Risk Assessment, Shoreham Nuclear Power Station, Inng Island Lighting Company, Final Report, Science Appliations, Inc. (Palo Alto, California), prepared for Long Island v Lighting company, 24 June 1984.
SERKIZ, 1983 A.W. Serkiz, Task Manager, Containment Emergency Sump Performance:
Technical Findings Related to Unresolved Safety Issue A-41, U.S. Nuclear Regulatory Cormiission (Washington, D.C.), NURm-0897, Revision 1, draft, O. December 19R3.
SILBERBERG, ET AL., 1985 M. Silberberg, J.A. Mitchell, R.O. Meyer, W.F. Dasedag, C.D. Ryder, C.A.
Peabody, and M.W. Jankowski, Reassessment of the Technical Bases for Estimating Source Terms, Accident Source Term Program Of fice, Office of d, Nuclear Regulatory Research, U.S. Nuclear Regula to ry Commission (Washington, D.C.), NURm-0956, draf t report for coment, July 1995.
O O
O O
O
6-16 i
k FIGURE 6 -'1 l
Hot. Leg Geometry Comparison -
Babcock & Wilcox Versus Westinghouse PWRs i
eeQ* LCG Pa8 e%G sit su.'Of-ot%fe ,
- 13) ' co%*actaC3 ~
D.%rsiw E $ . ..:. , y ,:. . . .
< rLW 5'.vC?u at _j l
s, M. .
I
- '*d [e
., a W
(
6
~,
,. ey ,
7: .
t s
.. ,.,e . c . .. ...- .... .
- o., . ; -
l . c.
~
-Mes q)) ;0 ^-===f )
)
g,f q'
{ tts
,.C.~' o ts '
) E UZ :.
Us'.i"i.n:#^.;::.. p- * .i;"" " '" " -
te(".vaitc 3 sgTS t .* )
m .c . 3. vr t s. 6 COLD LFG
c f.h p v
... _ . _ . . . . . . ,_ _ ._, y m .
O g ,
ar...
vs' yl c, fd
%d'
- s e J'y'JN i i
4 V' ,
J y;
) ~y, s __
[ Ai i jf f
= aw t. s d-W
~~ ww.
]
) ., ,
y"w,u, t!st!C4
. . .
- i . 3 7., , F o rt Lact it p a r t it
6-17 FICURE 6-2 Reactor Coolant System Geometry Comparison --
Combustion Engineering Versus Westinghouse PWRs O
us
%J
'N._ j sftau 515au C(=taatOn Gtutmaton 40 ? (
40 t
/v '
- r,. :::.. v
"-4' O 'rt Q}
h va s b Q.
Puup l P u tA P p_ /
NO te{
j mo en t
O. V
.m.m .
v ...=,0
@f 'U":. :-
O & ,, ey-- ,.
v
, , ,. r , . u ,,. s , . u ,
j g]
co w. t . . e. m..
v 9-Lj fiE >
JJ,ey y!
c'N w i
- l 1
i LAs Hv
- t. - Bi ,.
~
i Gr >'I >
O .
- .\t:(!
tyD;fs% -
c d.
n '
- w A" q,fi iW O.O i Q (L A 4,.V,.._'w 7,'~g u g
g .sa
.n .
y J, ata ?CE
'O ......,s .. . , . .. i..
. - t o ,
O
.- - - . . ~. . . . , ~ , .
6-18
'O .
- .e i
FIGURE 6-3 Comparison of PWR Steam Generator Designs
- Ol
. . . . m,.
4 ,
,y . . .s y
.,A.L-g ,
e .an.*.e.s.
[.AAAs,Apea p ..
s e e .e. r... .s
'r ,4 ,,,,'"
- O St een. t 0.. m s e, sa .e = .
q.,p. r M8Mo******* e g go eg ...
.I f.i,'5 ,- ,
. qfg3 v o-ea so..
rg IM b, Jp.g ai 4.sq g e e r U . :g .o. . . a' W 4.
{
. eat.a g **-g., ee &p' , O, 1
. **p , m e , - e
.r.
.r _-, _ - ] y<
-~~ -
/ r ~ ~ -
%. - '. e[. ..' ,a s.
g\m; r
.. .. 1
,.Q>
,o i 4
.'Ti'i .1 r*vi l -(
1 4 ,
i.
-.-,. .-p .e. ;
2 I{g g,,,,,,
I.
i - *
;..~
.. -4. 3g1 j'j.-
...-t,...... .% -
.t L .
...J-_.;. 4 . p ,: 3 ,m ... . . .,
. . . _ v= o t e.c. .
l
.!^.
,V
-.s s -
.. ' A. , _I $. ---
.-[.. .
',d
..y.
.u..~....:.. .' "
) D .
CD D) I WQ ay - .
f-(f - C (l;s. .&v 1 Kf*). . -
. . . . .. s m] %,
/
....u.
. e e r
- O
.... J
'0 h 6*S"
$.et f*
4 9 N
h 10
~ ' ~ -
.,y*..
} s. . .
..s .
d9y .. .u.
_+,...
pe o8 Qep 4 ..e2 4. . a
, p n . .I
. t.., . m,3 _ ,e.. ** Vr' 8
'g*- 43 t .s.
eo.s s'e,5..ess k
O
---?* .:
- g PS AP4 g
? 8
. = .q._.
. . . . . n i
t.. <....,
"*
- I
- a. .
=9,. e.
- t
+
- s 2,
),% *~*'.q *e.
. s .
. g,,* *. w h m. w engagg*,g q-
. m. e u
e, e n res ... .
g .
i.L.I.
~
s
- "' _ . .t .. ~_,,,D.,, , - .r e ~. . e
- s. ' - * - +
Q-9' _I
.. .. . .,/
w
- . m .s.
., . . 2. ,,
r.
' e'50 8 3
.. g .; a **%==",*
.4'
. . ei
- O '
l
- O. l
6-19 U
FIGURE 68I Comparison Between Top-Entry and Bottom-Entry v In-Core Instrument Tube Arrangements for Combustion Engineering and Babcock & Wilcox PWRs
, i=.coac sta vutur rio.,
SUPPORr ,L AIC
,,. . . . . . . g_ C cCaozzu O . .
. ... con ,,,,.u...,
lGutserust N~
h.*
a ge ;~ ,~,,.uo, ~,,,,o.,
NQZZLE i
- . - 3. It
. ailI.k r .ig. c I i
- -.. ..~.
C 0'e t a OL [.N_(( ; l f;' , .
ALIG w tN,MN UP tm U xu G D P- x ,y . . . . p .
.w ,, E -
'o_ m ,3 j .g 4 {.e.,j ... . lsI
[l 1 4 2
- t3 f'. '
Oz L i =; '-
g
- - ?. . v. P~R a- ou e-f.., . s + -
- . a. i l
$ L 4. tilt a h0 E .
- 0LSEE - - . - q- , * * *
- j ' " g,. * ;
- t, ggg, i /"% i .
6 lV ... ...i ' 4 v LENGrw so* .cTivt CORE
.',m..
l,.
, ,; !" ,[.:. . f f [
- i. ij.
CORE
, , , , , , , ,, , , '4 Q[y S n a o go
[
lM; . k.-*r= gYi.' \l Nf FUEL l LO*Em -
a
.e ., n o, **=1 a $ $EwSLY
' 'f 'i ':4*"li" 'a e,%
lg
! c "s " mea.J.'.i'1 ! I 1 i i fjip 0,*uc"J.c" lo f
- W to. E Pa~s g- -
7.,(
sr ru g,.g g, ~ w, g.,7 4 coat srce / ' . ,. . sN -
,...i 7p g ,n, ,o, ,. .
L~ s. , - ,Jg >e
. . . e..
.,.~.,
. t.
.,,,,n.,
l T . . . ,
"',4.*,",','*, y rs . * * '
X' 4g ~~ /N & / ' s
. .. . _m.i .. . . . . : a :
e a e i ,. .s o c v 1. s ,
T.'.,5 3,,,,,,...m. . t
!,pg 7.3 F
. n y. !
,w: ; , g. ,. _. . _
. , ,.m.
r, t . e l.
I l t.
.} tv I
4" i
j hJ: e.- t. ' ...; -
y.jl][d..f,sr )vi I " '3 f.:n. .x lO '-: N. n?w u -~ ! ' ?!
' Inti" ilp .Illtid ,
1.... In114 . 3:=' ""~ I HUI i ol g !!1111a.h!
? 4. ; .e .
O .[ t s si MEM ~
.n.g'%,u4)ll
- y$ wy, . .. ... ....
yppi n
; -. :a o . ... .. . . . . . . . . . . . - . . . . . . . -
O
!. 6-20 k l FIGURE 6-5 Comparison Between Raised-Loop ) (Davis-Besse) RCS Configuration With Lowered-Loop (Oconee Unit 3) RCS Configuration in Babcock & Wilcox PWRs k,
.* ,md
' I /9 &d r'
+ t 1 ..isu n. s, rJ
)~ stias x
)-[.. ,~ @)t 8
', [d St asatos . s l w e 3 ,g,, . / x . .
;C w., Jj \ tratantse
- i. . =
, p . ._ y g *
- b- ", 'f _7. "
, i i - I l ;
'[-- 1 i 1._ . _._ P. ~
; i , - , - - -
t t I; -
- ? :- i i _. - t, ' i I . ,
li F
- j li' ,
@ l w .h i
- - ,r 1 o;: 4 -
a a w ,
.i. l ! 1;; - ' - i 1 ,i -n. v ..t
, p .d ,r , t, _'.. l,
.i I ,~ ~
r*--.-
"f . 'I' , - [L; E[
l- j' ~ " - j {; _ f. 4 42*. 3 . t
,[ .u.., . .w.% ,
- f x'g T _.s}. -
*47 1 4 ( .Y. , i .
nF
-t 4
i { b
} }
- 1;
.a h.3d l
6
't . .' .. l i ; ,
3 Uk , r -
. .. !, y
- s. . .
i
^
- g. g. I g 3. g, R.-----e ,
j . _ ..
.a ,/ ! 'bi. C ~ ' ~
8
-- 's Q n .,/ - v-k ..m - - . , .9 - i i,,
C --
. . - . _ . l -w J ,
p-v .c- 1 .
.,: , 4 ' '.' ~
f 7
- a , ..
- v. . , .-. , _ .. . ,
4' f ' . . , . t
-e4 .3 i t-e , pts- ., ..c;a *at a tt o e s t b,ri, d- . ese weie 1g149', n't ?. f* Ti .et-ter s Oldet ,. 518' A ' '~ s P (.*~ *' t *I'*69*"'"o e o ..s.>.,. .. . u .>
l f 1
. - _ _ . _ _ . _ _ _ _ _ _ _ _ . _ _ . ~ . . _ . _ . . ~ _ . _ - _ . _ , - - _ - - . _ _ . . . _ _ . _ . _ _ . _ _ _ _ - . -
6-21
.O
- FIGURE 6-6 Interfacing LOCA (Event V) Valve Lineup Variations in PWRs
.C, [HEDDLESON, 1983]
.Q . Syste w S'5f tts eso 40
/
N .i.,T 1, ", . N _i .h _i h ,i
. . r,mb ", ,
e na .N E 26% 5 & d . 4er gente a%%4 4 & 2 e u.2-ama agialoNt 1 aan4454,On t y OCosit t ) 3 ma4LMO SECO Ca*11a6 a.vta l ' watt uitt estasJ t Su"8 v ' 6 2
. N- N{- k 2'--- L y N ,i A.,t at
_ wa. t a .it f eOsam etavte vattty e saggia e a y Devil SE556 4 Q ,7 e. saattv s s) t=at t a'e6 E ast aNO 2 40e f a%%s .. . ee N H1 ~4m t 9 Caos162 i N ) ,& CaLvtaf CLif f t e a 3 wettStu%t 2 es N
.. a,..E s . . . .L .4.
i a -N N h ,I Q 2 0% i & 2 +ws 0
.Nx ': .s,...;......_. - Hgg
- g. .a,,.,,, y 4' m43 no.. so , .ast sacts . ..t. ,0.so . 4
->.. . u aC-P*a *t .s6ar.o i. -H W.%o . . . . .in.i is.a... . soavCat.o => atamo Et ajoav .tsa ++N d484W' pNF ,
sa 4gg -
..a, m.N. t .u Craphie display of interface lineops for f*R I.FCI (lat.
']
- O iO
1 m. 'J' 6-22 I FIGURE 6- 7 PWR Reactor Cavity Design Variations O
, u o .. w., e .
t -
. .. r. . n. , L .. J.'l * .. e .;}; -
4
-m . ,
n . ~. - . v ,1., o ,, l;,..,
~,
(.s
- 7. -
p ;y
- J
- . .*, f: 'c : '..... % . . , c. . (y. ; a.,
p m I ... _ . , . .
-a. .. '
[ ! >
..... m s . 't .
s .,
/ c ~.y. <- 't**'. . , .
s.c. , F ,.., e * . . {. . . . - *
. . . . g. ..t*.... . x '. : ,. ..,a . ._ . .*a -
wea a
, 'I*' - PlaCTORvtsstL
{} TN as inw.e ae. t se.c , mens tas Cong
'I *; ', * .* ,?. ., *, ,.q% ,g cyo, g,y,,,
po.- a a.s s'estauwaseT tum%Et -
,, 7, , ,f . ,-- j - ..
l, _.*.- lion plaq s n eastmat t"l-". '.
. 's l . . p a ..no, ., a 9
g., i
*** een 8 e ,9*
- l ! x- ~
so.ai. e . . :< 'i ,~ a
,s' . e ,
pJ . : , see *.t. '.G {
'1' &. ' . L o'..
i !N
,.*s '. ; l . .#w~s. . h.I.
I . /
* . ' . . . 'E;;. . ' , ,e] .. %. ..? p.; '...,: 9..] _ua f $. .}$ .f! i.$# i ,
{_* .
..* _4 e .*'f,.., . , . . ;3f,;
6 s s.c..a e
. 3: t- *u' .s ..a a --
4 . --
- 3. ,f..C.S.
n V osme c_ . QI
-p.fi .h; .
- 4+,' . O
~ -i ~'.as % : ;. : 6 ;s r., g <f. .m ,g 5-.-. "'*'*a"** ,, y. , .ps.
YMS i" - -f!pdai. i L: 7't
- a. - l, _l , sI 5:
y 9w=.-
'Q ..>- .: , n , ,v ; .: ;s - s, 1 ;; l ',i . ".***.' .. . c.. >
l Q. . . *3 3
"(g - fjg j e .,, :[. 7 g.. . t .s..a.. *:. D, . ;i', '.*',f.7,, ,i .
M
'1' ./**- l __. k,.g , **i".*{
i., V,, *-
.yz..,rg"1g ..,a -s.-;.
yl O
,, ..,, . . . . . a.,.
t.%94 * .
... i ... t w, P
s.r. . f..i * . m
,. r2-s,;:.k:%. .y.-~,. : Uexh)"y:.<+ < .a -s v g._ a .. . ;;r . 3 .. c !. ,'. . a P.*P..-'
T 4 9. s= R. t.- g O F = 9 O
CHAPTER SEVEN o V Limitations of Theoretical Models for Source Term Estimation
SUMMARY
teoretical models are required to calculate source terms for V, each risk-important accident sequence. Wese calculations are extreely ceplex and subject to many uncertainties and limitations. Wis chapter reviews some of the shortcomings in these calculations that arise fra either the neglect or over-simplified treatment of various physical or cheical phencnena . Sate of the issues covered include the need for W the user to make many asseptions in performing the calcula-tions, leading to large uncertainties in the results; the lack of validation of the codes with suitable experimental data; and the inadequate treatment of a large n eber of specific chm ical processes and interactions. .O me theoretical models are, at best, a crude approximation of the actual phenomena occurring during a severe core melt accident. We option to choose input _ assmptions, the general lack of validation, and the substantial uncertainties in the predicted results render the codes unsuitable as a basis for regulatory decisions. a 7.0 Introduction Af ter a given accident sequence has been selected for the purpose of j)
\
=! source term estination, it is necessary to sploy a theoretical model of that sequence so that releases of radioactivity to the environment can be calculated. In all current estimations, such as those sponsored by the U.S. Nuclear Regulatory Conmission (NPC) , the theoretical model consists of a set O of interactive cmputer codes. By appropriate selection of various input paraneters, the user can apply this set of codes to a given accident sequence at a given plant. 9' We existence of this elaborate set of codes is testimony to the skill and effort of many analysts. However, predicting the outccne of a complex and violent event such as a reactor accident is a daunting task. It is, therefore, not surprising that currently available codes have severe O limitations.
- O
l } 7-2 j mis chapter briefly describes some of those limitations. It is not our intent to be exhaustive, but rather to illustrate modeling limitations through ] some exm ples. Other analysts have addressed some of these limitations, frm differing perspectives (DENNING, 1984; GITTUS, E7r AL. , 1982; KRESS, F1r AL. , 1985; LIPINSKI, ET AL. ,1985; RIVARD, ET AL. , 1984; WILSON, ET AL. ,1985] . 3 m varying degrees, each code either neglects, or treats in a simplified manner, various physical or chmical phenmena. Many of these phenmena have implications such that their inclusion or fuller treatment will affect the source term for a given accident sequence, while not substantially altering the progression of the sequence itself. In the present chapter, only this 3 kind of phenmenon is considered. Other neglected or partially treated phenmena can, however, have a more draatic effect, changing an accident sequence quite markedly. Chapter Nine of this report addresses such phenomena. b Section 7.1 provides a brief introduction to the overall model and supporting codes' used in NRC-sponsored work. Section 7.2 illustrates the sensitivity of source tenn estimates to both input ass mptions and the j adoption of different subsidiary models or codes, o Before computer codes can be used with confidence, they must be
" validated" against the results of experiments. Section 7.3 discusses this 3 procedure in a general way, while Section 7.4 assesses the status of existing codes, including the degree to which they have been validated. Finally, Section 7.5 presents scme conclusions.
7.1 The NRC Source Term Model and Its Supporting Computer Codes 3 Through a program drawing upon several national laboratories, but centered at the Battelle Coimbus Laboratories (BCL) , the NRC has developed a source tenn model as illustrated in Figure 7-1. Wis model has been applied 3 by BCL to a variety of accident sequences at several nuclear plants (GIESEKE, ET AL. , 1984a-e] . g Although the model contains a nunber of elaborate cmputer codes, its product -- the estimated release of radioactivity to the envirorment -- is O t
j 7-3 strongly affected by other factors. For exarnple, the first three steps shown in Figure 7-1 -- the selections for study of plant types, of particular 3 plants, and of particular accident sequences -- have a great effect on the outcme. It is therefore noteworthy that, as shown in Gapters Ebur, Five and Six of this report, the set of plants and accident sequences studied by BCL is not sufficient to assess the overall accident risk at U.S. nuclear plants. O Also, the mode and timing of contaiment failure has a major effect on the source tem, as shown later in this chapter. However, contaiment failure is not directly predicted by the source term model shown in Figure 7-1. g Rather, this model predicts contaiment loads (tmperature, pressure) which are used in separate analyses of contalment performance. We results of those analyses -- in terms of the size, location, and timing of a contaiment breach -- are then fed back into the central source term model. 3 tis central model contains eight principal emputer codes, interacting as shown in Figure 7-1. W e reiainder of this section is devoted to a brief discussion of the functions of these models. A detailed discussion is provided in a report prepared by the Oak Ridge National Laboratory (KRESS, 3 ET AL., 1985). Consider first the MARCH code. Wis represents the heart of the source , term model. It predicts mass and energy transfers (scmetimes referred to as J
" thermal-hydraulic conditions") within the reactor core, the reactor coolant systmi (RCS) and the contaiment. Figure 7-2 shows how the plant (a FWR in this case) is represented for the purpose of calculating heat and mass transfers. mss transfers covered by this code incitxie: (i) flows of water, g
steam, and noncondensible gases; and (ii) the motion of molten core material. Ehergy sources include: (i) stored heat; (ii) fission product decay heat; and (iii) chmical reactions such as that between zirconim and steam. 1 3 from the MARCH code, and from them MERGE accepts certain outputs
- calculates the gas (and steam) flows and taperatures, and the structure tmiperatures, within a set of control volmes which collectively constitute the RCS. W ese calculations are necessary because MARCH treats the entire RCS gas-space as a single voline, and more detailed information is necessary when
r 7-4 estimating the retention, within different parts of the RCS, of radioactive material released fr a the core. _) Such releases are estimated by CORSOR, which accepts the core tmperature history generated by MARCH and from this generates release rates of radioactive and inert materials. We inventory of radioactive isotopes is 3 provided by the ORIGEN code, whose function is to determine the initial inventory of radioactivity (fission products, actinides, and activation - products) in the core. 3 While the core reains inside the reactor vessel, releases determined by CORSOR are fed to the TRAP-MELT code. Wis code, which also accepts the flows, taperatures and pressures generated by MARCH / MERGE, estimates the transport and retention of radioactive materials as they move through a set of 3 control volmes representing the RCS. For situations where the core has penetrated the lower head of the vessel and fallen onto ithe concrete floor beneath, the CORCON code is s ployed. O CORCON accepts the cmposition, mass, and initial taperature of the pool of molten core material as determined by MARCH, and calculates the thermal and chmical interactions of this pool with the concrete below. 3 Its principal outputs are: (i) the generation rates of steam and non-condensible gases; (ii) 3 the taperatures and ccnipositions of the melt layers; and (iii) the rates of erosion of the concrete. CORCON's outputs are fed to the VANESA code, whose function is to 3 estimate the release of radioactive and inert materials to the contaiment atmosphere during core-concrete interaction. We initial melt composition used in VANESA will reflect earlier releases frmi the core to the RCS, as determined by CORSOR. O Releases to the contaiment atmosphere, whether from transport throtqh the RCS (calculated by TRAP-MELT) or from core-concrete interaction (calculated by VANESA) are then provided to NAUA. We function of NAUA is to l p calculate the behavior of aerosols in the containtent volme. When provided with a contaiment leak rate -- which is a product of the MARCH calculations, l i 3
7-5 ) given an assmed leak area -- NAUA will also estimate the leak-rate of each nuclide fra the contairment to the enviroment. O
'No additional codes are used to support NAUA for particular reactor types. We SPARC code calculates the retention of radioactive neterial in the suppression pool of a IER, and is used where it is believed that the release g pathway will pass through the pool. For an ice condenser WR contairment, a similar function is performed by the ICEDF code. ICEDF calculates the retention of radioactivity in the ice cmpartments of that contaiment.
7.2 Sensitivity of Results to Asseptions and Models g When estimating source terms with the above suite of codes, the analyst is obliged to make nmerous assmptions, covering both particular nmerical values and types of behavior. For example, the user of the MARCH code must 9 select the tenperature at which fuel begins to nelt. In addition, the MARCH user must select one of three options describing the behavior of the core af ter nelting canences. O Also, containment failure is not -- as pointed out above -- predicted by these codes. Separate analyses are conducted for this purpose, using the contaiment loading conditions estimated by MARCH. These analyses, as shown Yet, their j in Chapter Eight of this report, anbody substantial uncertainties. results have a najor ef fect on source term estimates. ! Wus, a source term estimate for a particular plant and accident sequence will reflect many input assutptions, some of which are uncertain over a wide g range. It is therefore instructive to exanine the sensitivity of the results to the choice of input assmptions. i A sensitivity study of this kind -- the QUEST study -- has been conducted lg ! for one accident sequence at each of three plants, using DCL's methodology
- i. [LIPINSKI, ET AL., 1985). For illustration, consider the QUEST analysis conducted for a station blackout accident at the Surry NR.
Q i O
'O - 7-6 For this case, Table 7-1 illustrates the sensitivity of various predicted source term parmeters (e.g., the mass of fission products released to the O enviroment) to the input assmptions. Two contaiment failure classes are shown: early failure (141-173 minutes); and late failure (900 minutes) . For each class, a "high release" and a " low zelease" cac;e is shown, reflecting a choice of code input parmeters expected to lead to "high," i.e. large, and ,0 " low," i.e. mall, releases, resectively. me assmed size of contaiment 2 2 breach is also assmed to be "large" (10 m ) or " mall" (0.001-0.01 m ) for "high" and " low" releases, respectively.7.5 O It is clear fra this table that contaiment failure tine has a powerful effect on the results. When code input parmeters expected to lead to "high release" are chosen, an early contaiment failure leads to an estimated release of 300 mci of radioactivity. For the same parmeter choices, late O contaiment failure leads to a release estimate of 4.8 mci. Also, the entire range of release estimates is very large. @ e estimated
-10 mass of CsI released ranges over eleven orders of magnitude (from 2.5 x 10
,0 kg to 20 kg), while the predicted total release of radioactivity ranges over four orders of magnittxle (from 2.9 x 10-2 ' mci to 300 mci) . Another illustration of sensitivity is provided by Table 7-2. Wis O table, also drawing upon the QUEST study, shows how the predicted releases to containment of "ref ractory" radionuclides as a result of core-concrete interactions -- following a station blackout accident -- are affected by the degree to which unoxidized zirconium is present in the nelt. %ese refractory O nuclides are of particular concern because sme of thmi are biologically very haza rdous. Core-concrete reactions of this kind occur when the molten core has penetrated the lower head of the vessel and fallen to the concrete floor beneath. Releases to contaiment are estimated by the VANESA code, drawing lO upon results frm CORCON. l Table 7-2 shows that VANESA's results are highly sensitive to the unoxidized zirconim fraction. For exm ple, if the molten core material falls O fra the vessel with 41 percent of the zirconim unoxidized, it is predicted
~4 j that 5 x 10 percent of the Lanthane Group nuclides in the core will be I
i O t
p 6 .O 7-7 released to the contalment. If, however, the nelt falls with 80 percent of the zirconim unoxidized, it is predicted that 30 percent of the Lanthane O croup nuclides will be released to contaiment.7.6 For neny accident sequences, it is probable that the unoxidized fraction of zirconim would be less than 80 percent. However, this may not be true for O all accident sequences. A recent report by a study group of the Aterican Physical Society (APS) has expressed particular concern about rapidly evolving sequences at BWRs (WILSON, ET AL. , 1985] . BWR cores contain more zirconim than those of PWRs, and a rapidly evolving sequence might lead to core melt O before much of the zirconim had oxidized. Uncertainties related to core-concrete interaction can also have significant Implications for contaiment failure, as illustrated by Table 7-3. lO This table shows contaiment conditions predicted by the MARCH code in two variants: (i) in its current form, in which the INTER sub-routine is used for core-concrete interactions, and (ii) in a modified form, in which CORCON (a more recently developed code) is used as a replacment for INTER. O Results are presented in Table 7-3 for one accident at each of two notional plants -- a EWR and a BWR. It will be noted that the MARCH (CORCON) variant generally leads to more severe containment loading conditions.
- O Contaiment emperatures are substantially higher in both cases and pressure is higher in one case. w>re significantly, the CORCON variant predicts a much greater production of fla
- meble gases (CO and H 7) . Explosion of these gases could present a major challenge to contaiment integrity, as discussed in O Chapter Nine.
Wis particular exa.nple shows that the source term estimates made at BCL (GIESEKE, ET AL.,1984] could be misleading, because the MARCH (INTER) variant .O was used in that work. me exmple also illustrates a more general point -- that source term estimates can be highly sensitive to the choice of models mipioyed to nake the. Since many of the current models are somewhat crtrie, it follows that current source term estimates have a wide (and as yet unknown) 10 range of uncertainty. 4 iO
) 7-8 In sutmary, it can be seen that the choice of input assmptions can have a significant effect on the source term predicted for a given accident sequence by current methodology. Likewise, the predictions can be very sensitive to the analyst's choice of theoretical models. Wus , little con-fidence can be placed in the results to date. ) 7.3 validation of Computer Codes Before confidence can be placed in an overall source term estimation, it is necessary that an equivalent level of confidence can be ascribed to each of ) the subsidiary models. It must be shown that each nodel provides an adequate representation of the relevant phenmena, and that the representation of the model by a emiputer code does not introduce numerical errors.7*7 ) The process of acquiring confidence in a code is known as its
" validation". In this process, the predictions of the code are cmpared with experinental results or with the predictions of another code which has itself been validated.
Were are no generally agreed criteria for validation of source term codes. Instead, developers, users, and reviewers of the codes use their judgnent to determine if a particular code is " adequate". 'Ib date, this has ) not been a source of major controversy only because it is generally agreed that most of the source term codes have not been properly validated. A possible approach to experimental validation is shown in Figure 7-3. ) This figure shows two principal stages. First, the code is developed in conjunction with sene "developtental" experiments. Second, the developed code is emipared with a second set of experiments, presmably more representative of the phenatena beirg modeled. If the code's predictions are sufficiently ) close to the experimental outcomes, the code is demed validated. If not, the code re-enters the developtient process. As mentioned above, a code might be validated not directly against ) experinental results but against another code -- a practice of ten known as "benchtiar king". @is would be done, for example, if the code to be validated )
7_, O were deliberately simple so as to economize on caputer time. If the simple code agreed with a more elaborate code over the range of interest, and the Q elaborate code agreed with experiment, then validation could be said to be ! established. Because a reactor accident is a violent and complex event, it is not O possible to directly compare any source term code with a completely representative experiment. 8 Instead, the accident must be broken down into separate phenomena, for each of which a direct emparison with experinent is
-- at least in principle -- feasible. However, such separation may be very difficult because of the neber of parameters which must be considered. For O
example, the behavior of radioactive materials within the RCS can be strongly affected by tmperature, pressure, rate and pattern of flow, radiation effects and highly specific chenical reactions at surfaces. O Even where phenomena can be separated, it is of ten difficult or prohibitively expensive to conduct an experiment which can be directly empared with the ace ident situation. In particular, it is rare for an O experiment to approac't reactor-scale in its physical dimensions, even thotsh it is well recognized that physical scale can be an important paraneter.
% the outside observer, some reassurance would be provided if codes O could, without adjustment, provide reasonable predictions of the results of experiments at different scales and in different parameter ranges. Indeed, smethiry like this is implied by the second stage indicated in Figure 7-3, in which a developed code is tested against " validation experiments". During this stage, it would seen logical that the people makiry predictions with the O
codes should have no involvenent in setting up the experiments. M date, the source term codes have generally not faced tests of this kind. We codes have l generally not moved beyond the developnent stage, during which they are lO adjusted to fit the results of experiments. i Because the codes have not yet been validated, it is not possible to quantify their uncertainty. Indeed, the state of the source term modeling art must advance substantially before this becmes possible. If the importance of O scaling can be fully understood, if all the major deteuninirg factors for each lO
J 7-10 phenmienon becone known, and if theoretical models incorporating all these factors can be developed, it will becae -- in principle -- possible to O quantify the degree of validation of the codes. For a particular accident sequence at a particular plant, this would translate into a statment that the j size and timing of release of a particular radionuclide would possess partic-ular distributions of probability. Such quantification has not as yet even O been attmipted, as illustrated by the following quote from a recent review of the status of validation of the source term codes (KRESS, ET AL.,1985): It is apparent, then that quantification of the level of validation requires a rigorous sensitivity / uncertainty analysis .O that properly accounts for such things as variance and biases in the input, in internal pre-fixed parmeters and correlations, imperfect modeling, finito dif ference approximations, imprecise application of the codes, m.1 results of separate effects and integral experinents. Such a level of validation would require considerable tine and resources and was considered to be outside ,0 the scope of the present review. A quantitative approach to validation would require more than the
" considerable tine and resources", stipulated by these authors. It would also Lg need to rest upon an improved foundation of theoretical understanding.
7.4 Merits of Current Codes Discussion here focuses entirely on the codes described in Sect ion 7.1. g Thesa codes have formed the basis of recent NRC-sponsored source term estimations. A variety of efforts are under way to develop new codes and improve the existing codes. Evaluation of these efforts is best left until 'g their conipletion. As nentioned in Section 7.0, many analysts have addressed the limitations of source term modeling and codes. However, two reports are of particular interest in terms of the nerits of current codes: the report of the APS study g l group [ WILSON, ET AL. , 1985); and a code review conducted by an Oak Ridge National Laboratory (ORNL) team [KRESS, ET AL., 1985). Readers seeking I
- detailed information are advised to consult those reports and the work cited within thm.
lg i Q
7-11 Some of the most prominent characteristics and limitations of the codes are as follows.
'Q 7.4.1 MARCH As is clear from the Appendix to this chapter, MARCH is an elaborate code
,0 which addresses many phenmena. It is equally clear that MARCH is not reliable af ter core nelting has begun. Indeed, the user has the choice of three options for the progress of a core nelt. O m e rate of oxidation of zirconium is an important factor in the progress of a core nelt.7*9 Yet, there is a very limited base of experiments on zirconim oxidation under core-nelt conditions. M3reover, MARCH does not seek to estin. ate the spatial distribution of zirconim after nelting begins. mis O may be particularly important for BWRs, for which MARCH accounts for zirconium in the channel boxes through the crude approach of increasing the effective thickness of the fuel cladding. O Another important factor is the behavior of molten core material af ter it has fallen into water. @is will occur during two phases of the accident: (i) when the core simps into residual water in the base of the reactor vessel; and (ii) when the core material nelts through the base of the vessel 'O and falls into a water-filled cavity beneath -- the presence of water in the cavity being specific to the plant and the accident sequence. hhen the molten core material falls into water, it will be segregated into small particles.7'10 te sizes and shapes of these particles will have a strong .O effect on the rate of the chm ical reactions with water and steam, and on heat transfer fran the particles. In turn, these factors will affect the timing of the accident sequence and the contaiment loading by steam and hydrogen. Yet, !- MARCH assunes these particles to be spheres of a user-specified diameter, with O no segregation of zirconium. In addition, the ut.er must choose one of three i options for modeling heat transfer fran the particles. i As nentioned in Section 7.2, MARCH employs the INTER code as an internal
'~)
subroutine to estimate core-concrete interaction for the purpose of determining contaiment loading. However, Table 7-3 shows very pocr agreetent lO l
O 7-12 between the amounts of flanmable gases predicted using INTER and those predicted using the later code CORCON. Until a core-concrete code is devel- l C oped which can be shown to be accurate by validation with experiments, pre-dictions of contaiment loading will have to be regarded as unreliable. Also, cmibustion of flanmable gases (CO, H )2 is crudely modeled by MARCH. O We contaiment is treated as a single volme (see Figure 7-2) , detonation or quasi-detonation .11 is not addressed, and the extent of burning is lef t to the choice of the user. O te behavior of airborne radioactive materials inside the contaiment will be strongly affected by the condensation behavior of stem. Yet, MARCH models condensation on surfaces using experimental correlations ubich are extended well beyond their empirical data base. Also, steam in the O containment atmosphere in excess of that required for saturation is arbitrarily assmed to form airborne droplets of four micrcmeters dimeter. Interactions between aerosols and steam -- as would be analysed by an integrated code performing the functions both of MARCH and NAUA -- are not .O examined. 9 In smmary, the current Nrch code cannot accurately provide the detailed information necessary for a realistic source term estimation. O 7.4.2 MERGE The purpose of MERGE is to obtain a more accurate picture of thermal-O hydraulic conditions inside the RCS than is produced by MARCH. To this end, MERGE estimates mass and energy transfers within a series of campartments ! cmipouing the RCS. O However, the flow is assmed to be one-dimensional, and that dimension is I approximated by only a mall number of " nodes" (control volmes) . Were is no treatment of circulatory flow within the nodes, as will arise from natural f t convective circulation. tis is a three-dimensional effect which could be ,0 very significant, particularly for high-pressure core neits. Q
3 7-13 Circulatory flow within the nodes will tend to lead to a more uniform taperature distributi;n thrcughout the RCS than would be predicted fr m a O one-dimensional analysis. In turn, a uniform tmperature distribution will act to suppress such flow. Wus, an accurate thermal-hydraulic model of the RCS must be three-dimensional and must realistically represent the heating of RCS structures. O It is not yet clear how the adoption of such a model would affect the predicted retention of radioactive material within the RCS. Transport from the core may be enhanced, but the heating of structures elsewhere in the RCS ? may limit retention in those regions. Such heating of RCS structures could also lead to failure of the RCS, one possible site in m'Rs being the stem generator tubes. This possibility, O potentially significant for high-pressure core melts at W'Rs, is addressed in 01 apter Nine. l MERGE does not incorporate the structural heating effects which would C arise frm the deposition of radioactive materials -- such deposition being estimated by TRAP-MELT. Decay heat from deposited material could alter the cellular flow patterns which would otherwise be established. O Wus, MERGE uses a simple one-dimensional model whose validity will be unknown until MERGE is benchmarked against a detailed, f ully-va lida ted , thermal-h%raulic model of the RCS. In addition, MERGE does not account for j the heatirn effects of radioactive material deposited throughout the RCS. Q i 7.4.3 CORSOR l tis code assmes that the release rate of a specific nuclide frcm molten O fuel decreases exponentially with time, with a time constant which depends only on tmiperature. Wis dependence is based on correlations derived frm a i small nutber of experiments. O It is clear that this model neglects the ef fects of gemietry, surface area, flow conditions, and fuel burnup. We model also -- except in the case of one nuclide -- neglects the ef fect of unoxidized zirconim, although it is O
3 7-14 known that the release rate of certain nuclides is sensitive to the extent of zirconim oxidation. 12 J Moreover, the spirical data on which the tmiperature correlations were based show a wide scatter, suggesting that factors other than taperature can
- be important. Also, the data base is very weak for control rod and structural
- O materials, which can play an important role in aerosol behavior.
Despite experimental observations (at the Power Burst Facility in Idaho) that enhanced release of fission products occurs as molten fuel is quenched on O falling into r.esidual water in the base of the reactor vessel, cORSOR has no neans of estimating such an effect. Likewise, there is no way to account for the enhanced release likely to accompany fuel fragnentation arising from reflooding of a hot core. O In smmary, CORSOR does not account for many of the factors which would affect the release of radionuclides fra fuel in an actual reactor accident. O 7.4.4 OaIGEN The core inventory of radioactivity is predicted by ORIGEN, which is a well-established code. It is, by general agrement, much the best vslidated 'O of the current source term codes. Nevertheless, ORIGEN has a greater uncertainty than is often acknowledged. According to the above-cited ORNL code review (KRESS, ET AL. , O 1985], measured and ORIGEN-predicted values of isotope ratios have generally agreed within about 5 percent. liowever , one emiparison between ORIGEN predictions of isotope content (for nine isotopes) and the measured values for actual fuel rods showed agremient rarging fra a 43 percent underprediction to O a 76 percent overprediction -- suggesting a general uncertainty of about + 30 percent (NRC, 1985).7.13 tis level of uncertainty is instructive. ORIGEN is based on well-LO understood reactor physics and rests on a large mipirical data base. We code l l is also instrinsically much easier to validate than other source term codes. ! Yet, its predictions are still smewhat uncertain. O l-
) 7-15 7.4.5 TRAP-MELT i O This code calculates the transport and deposition of radioactive vapors and aerosols within the RCS. In the recent BCL studies, only four different aerosol and vapor species have been studied: CsI, Cs0H, Te, and "other aerosols." O TRAP-MELT trodels the RCS as a small n eber of " control volmes ," as l adopted for MERGE. Within each control voltme, all paraneters are assmed to l be uniform, although airborne material can be deposited within the volme. O Following MERGE, material flow between c mpartments is one-dimensional -- cellular flow is not considered. Phenomena addressed by TRAP-MELT are: (i) bulk convective transport of O material between control volmes by the carrier stream; (ii) irreversible ( sorption of CsOH and Te onto RCS surfaces; (iii) condensation and evaporation of vapor species at aerosol and RCS surfaces; (iv) aggimeration of aerosols by several mechanisns; and (v) deposition of aerosols onto RCS surfaces by
- O various nechanims.
?
Weaknesses can be readily found in the treatment of each of these areas. In the area of bulk transport, the distribution of material throughout the RCS O will be sensitive to the detailed flow conditions. Yet MERGE does not supply, and TRAP-MELT cannot use, such detailed information. In addition, only steam is considered as a carrier fluid. For an actual accident, a steam-hydrogen mixture -- with quite different properties -- would act as the carrier. O Irreversible sorption is thought to make a small contribution to RCS retention, except in the case of M. In terms of the "conservatian" of the results for nuclides other than 2 , it may therefore not be critical that O TRAP-MELT's sorption model has a substantial -- but unknown -- degree of uncertainty. We model uses a single " deposition velocity" to represent l several emiplex nechanisns, and rests on a limited data base. l l
- O te interactions of vapors and aerosols with RCS surfaces are sensitive
! to detailed flow conditions (incitriitv3 the presence of hy3rogen), surface and
7-16 O carrier fluid teperature, detailed choical reactions at surfaces, and the effects of radiation. Yet, detailed flow conditions are unknown, and the O effect of structural heating by deposited radioactive material is not account-ed for at all. Surface properties are assmed to be unchanged throughout the accident, despite the predicted deposition of substantial amotets of material. Moreover, there is a very limited experimental base in regard to the surface 'O effects relevant to an actual reactor accident. Finally, resuspension of deposited material is neglected. Recent experimental results illustrate the importance of one of these ,0 factors -- the presence of an inense radiation field. Wese results show that it is not appropriate to assm e -- as has been the case with BCL's application of TPAP-MELT -- that iodine will be present in the RCS as CsI. O In experiments conducted at Sandia National taboratories, cesium iodide vapor was passed, with stem and hydrogen, through a reaction tube containing stainless steel coupons (to represent the interior of the RCS) . We tmipera-ture in the tube was about. 1200 K and the radiation received by the stainless steel ranged from about loco to 1480 rad / min. vapors leaving the reaction O tube contained five to ten times as much iodine as cesim, indicating that CsI had strongly dissociated. Moreover, it appears that the liberated iodine will be ionized in this type of situation and will therefore not tend to form metal i 9 iodides. In view of the importance of iodine radioisotopes as a radiological hazard, it is very important in source tem estimations that the behavior of iodine be properly understood and accounted for. Wis has not been done in g the recent BCL sttriies. TRAP-MELT's treat:tient of aerosols assmes that the aerosols are present 9 in a dilute form. However, high mass densities are predicted for parts of the RCS. No experiments on aerosol behavior have been performed at such der.ities. Also, it is assmed that aerosol particles are always spherical, despite experiments (particularly in dry enviroments) showing substantial departure fra sphericity. It is well known that particle shape has an 9 important influence on aerosol aggimieration and settlirg rate and on the condensation and evaporation of vapor at aerosol surfaces. O
3 7-17 y tus , it is widely acknowledged that TRAP-MELT's prediction enbody a substantial -- but unknown -- level of uncertainty. In the words of the ORNL code review (KRESS, fT, AL., 1985]: ] Overall, it is believed that there is large uncertainty in the ability to_ calculate the retention of fission products and aerosols within the RCS. J 7.4.6 CORCON When it is determined that the molten core has fallen into the concrete cavity below the reactor vessel, CORCON is used to estimate the thermal and ) chmical interactions of the molten core material with the concrete. We results of CORCON are used by VANESA, which estimates the releases of radioactive and inert materials to the containment atmosphere during core-concrete interaction. J As mentioned earlier, CORCON leads to results substantially different from those arising from use of its predecessor, INTER. Since considerable effort has been devoted to developing CORCON, there is a tendency to believe J that the code is more accurate than INTER. However, CORCON has been validated to only a limited extent. Its severest test bas been a caparison of its predictions with the results of two D experiments at Sandia National Laboratories (SNL) . In each of these experinents, about 200 kg of molten stainless steel was poured into a concrete crucible. Figure 7-4 shows a comparison of CORCON-predicted and measured melt taiperatures for one of these tests. It will be seen from Figure 7-4 that CORCON's predictions were in reasonable agreenent with the experiment, but that CORCON did not pick up a potentially important trend. We experiment shawed a nelt tmperature which ) initially fell below the CORCON predictions but then rose -- at about 4 degrees C. per minute. Wis tmiperature rise suggests that about 10% of the i heat input to the nelt (fran induction heating in this experiment) was not lost to the environment over the latter period of the experiment. W is could
) have arisen fra the fonnation of a thin surface crust, a possibility not modeled in CORCON.
e )
h 7-18 Further experimental ef fort is required to support the developnent of CORCON. Wis effort should address the possibility of thin surface crusts, as O well as a variety . of other phenomena. Wese phenmena incitxle: (i) the presence of a stable gas film between the nelt and the concrete; (ii) the role of SiO2 when incorporated in the nelt; (iii) the mixing of laprs (CORCON assmes segregation of melt'~ layers); and (iv) the effects of an overlying O layer of water (incitriing the possibility of steam explosions) . Finally, it should be noted that CORCON may not account for all releases
, into the contsiment during the core-concrete interaction phase. Part of the O core may have .been left inside the vessel, and deposited aerosols and condensed vapors within the RCS may be re-evolved due to their self-heating effect. We current source term model does not account for such possibilities.
O 7.4.7 vANesA This code accepts CORCON's estimates of the melt ta perature and the evolution of water vapor and CO 2 fr m thermal decanposition of the concrete. V- This gas and vapor is assuned by VANESA to instantly cane to thermal and chenical equilibriun with the metallic layers of the melt, yieldirg bubbles of gas containing H, H2, OH, 0, 0 ,2 CO, and CO2* O - We.se bubbles -- assuned to be of 2 on dimeter -- rise through the melt, collecting radioactive and inert materials as they rise. Wose materials are released to the contaiment when the bubbles reach the melt surface, the condensible species beconing aerosols at that point. O Although VANESA has shown fair agrement with the results of sane limited For experiments, it requires further develoguent in various respects. exanple, the occurrence of " swarms" of gas bubbles may significantly affect C the formation of aerosols at the melt surface. Little is known about this possibility. Also, the possibility of chlorides, fluorides, and sulphides being present must be considered. Such compounds could be more volatile than the elmental fonns, oxides, and hydroxides assuned in VANESA, leading to O higher releases of radioactive species. In fact, halides and sulphides have been observed in condensed aerosols collected after nelt-concrete interaction experiments. i (
4 7-19
'Ib support developnent Of VANESA and related codes, a more substantial experinental data base must be built. 'Ib date, experinents have not shown the O conbination of scale and cmplexity necessary to represent actual accident conditions. For example, nelts have not been spiked with fission products (in either their stable or radioactive isotopes) , thus renoving the opportunity for direct observation of release rates. Also, although the presence of O unoxidized zirconian is important (see our previous discussion) , it appears that none of the major experinents have incitxled unoxidized zirconitzn in the melt.
O As an indication of the state of knowledge in the entire matter of core-concrete interactions, there renain unexplained differences mong the results of various experiments. In particular, tests at SNL have shown substantial aerosol enission, whereas similar tests in West Germany have not. O In stmnary, to quote the ORNL code review (KRESS, ET AL. ,1985]: We VANESA model represents a major stride in the ability to predict mechanistically the aerosol generation and fission O product release associated with core debris / concrete inter-actions. We VANESA model is, however, of recent vintage. We model does appear to yield results that do not wildly disagree in qualitative ways fran observations made in experiments. We validation of VANESA is not complete by any neans. 'O 7.4.8 NAUA We f ate of airborne material in the contaiment, whether fran in-vessel Wis code, processes or fran core-concrete interaction, is estimated by NAUA. O which draws on MARCH for contaiment conditions and on TRAP-MELT or VANESA as sources of radioactive material, predicts the anount of material which will leak to the enviroment throtqh an asstrued contaiment breach. O NAUA accounts for the followirg aerosol processes: (i) aggiateration (Brownian, gravitational, turbulent) ; (ii) renoval by gravitational settling, containTent sprays, diffusiophoresis, and diffusional plateout; and (iii) steam condensation onto aerosols. 'O only two species are considered: water, and an aerosol representative of all the radioactive and inert materials present. % rough this adoption of a O
i . 4 7-20 single aerosol species, substantial -- but currently unknown -- errors may be introduced. In an actual accident, there will be many aerosol species. f O As for TRAP-MELT, it is assmed that all aerosol particles are spheres, although it is generally recognized that aerosol behavior is sensitive to the actual shape of the particles. In any event, there is a lack of spirical O data.on the shapes of aerosol particles under relevant conditions. We effect of hydrogen burns on aerosols is ignored, although such burns l could alter the shapes and sizes of aerosol particles and could lead to 0 important chmical reactions -- for example, cesian iodide could be oxidized to volatile iodine. Radiation-induced reactions in the contaiment atanosphere, as well as in O the sunps in the floor of the contaiment, could also have a substantial effect. For exmple, nitric acid could be formed by ionization of air in the - hmid contaiment, leading to acidification of water in the contaiment sunps. Hydrogen peroxide formed by irradiation of water in these sumps could, in acid O conditions, completely oxidize dissoved iodides to free iodine.714 NAUA and its companion codes do not have the capacity to predict such effects. Neither NAUA nor its companion codes have the capacity to model radioactive decay chains during the course of the accident. Yet, such decay O chains can be important. In particular, it is expected that telluriun will be released to the contaiment atmosphere over several hours of core-concrete interaction. We decay of Te-132 in aerosol particles to I-132 will thus O provide a continuing source of airborne radiciodine. Wis contribution to airborne iodine may far exceed the mount of iodine previously released from the BCS which reains airborne at that time. Delayed containment failure could thus be accmpanied by a substantial release of iodine. O tus, NAUA requires substantial further developnent if it is to represent actual accident conditions. This developnent should be supported by experiments which address the effects of hydrogen burns and radiation O chmistry, and which cover the full range of relevant aerocol and hunidity conditions. O
O 7-21 7.4.9 SPARC and ICEDF O These two codes supplement NAUA by calculating the decontamination effects *1 of BWR suppression pools (SPARC) or of the ice beds in IWPs with Ice (bndenser Cbntairrnents (ICEDF) . In each case, the decontamination effect will exist only if the radiation release pathway incitries the suppression pool O or ice bed. Other pathways are possible, as mentioned elsedere in this report. 1 SPARC assmes that radioactive material passes through the pool as O aerosol particles suspended inside stem bubbles. We bubbles are assmed to rise relatively slowly, reaining in thermal equilibritzn with the surrounding water. Aerosols are raioved frm the risirg bubbles by: (i) diffusiophoresis; l (ii) gravitational settling; (iii) inertial deposition; and (iv) diffusional 3 deposition. In an actual IMR accident, there may be rapid pressure changes in the drywell. Pressure pulses could arise frm high-pressure melt ejection, steam m' explosions, or (in some ciretrustances) hydrogen explosions. At such times, there may be violent fluctuations in pool level, and the equilibrium assm ptions of the SPARC code (e.g . , bubbles are spherical and in thermal equilibritru with the pool, the pool is perfectly mixed, and the bubble interiors are perfectly mixed) may not hold. On the basis of a substantial research ef fort, it is now believed that pool dynamics will be quite violent for a design-basis loss-of-coolant accident (LOCA). Were has not yet been such a research effort on pool dynanics subsequent to severe core damage, Absent such an effort, it is praature to asstrue the benign conditions modeled j by SPARC. l l In any event, SPARC is still under developnent, and there is a limited experimental base for its validation. Sme relevant tests with bubbles containing CsI aerosols have been conducted at BCL, but the data have not yet been published. O ICEDF is at an even earlier stage of develognent, and has no experimental base at all. Were have appearently been no experiments on the scrubbing of J
7-22 aerosols by ice condensation. Given this state of affairs, it is probably unnecessary to mention the code's failure to account for phenanena such as hydrogen burns. Sus , there is considerable uncertainty about the predictions made by SPARC and ICEDF. 7.5 Conclusions (i) Most of the NRC-sponsored source term codes are still in the developnent phase and have not been properly validated against experiments. (ii) Were are no generally accepted criteria for determining when a code n has been adequately verified. (iii) A substantial -- but in most cases unknown -- level of uncertainty will be present in the predictions of these codes. (iv) Quantification of the level of uncertainty will not generally be possible until improved theoretical understanding is obtained and a stronger experimental base exists.
')
l; (v) It is known that the predictions of the existing codes are sensitive to their input assuuptions and the choice of subsidiary models. (vi) Considerable further research is needed before the predictions of these
-)
codes are used in the regulatory process.
./
Y Q i ._ / 1' L
4 7-23 7.6 End Notes for Chapter Seven 7.1 We version used in BCL's analyses is the MARCH 2 code. Excerpts from the user's manual -- describing the objectives, strengths, and limitations of the code -- can be found in the Appendix to this Chapter. 7.2 Flow-rates and tenperatures of steam and hydrogen leaving the core, together with RCS pressure. s 7.3 MARCH contains an internal sub-routine known as INTER, which addresses core-concrete interactions in the smie manner as CORCON. However, the output of INTER, which is a more primitive code than CORCON, is used only to estimate contaiment loads. 7.4 A station blackout accident features total failure of AC power and of feedwater to the steau generators. @e Surry plant features a large dry subatniospheric contaiment. N Note that the contaiment breach could be much larger than 10 m -- see 7.5 Chapters Eight and Nine of this report. 7.6 he study underlying Table 7-1 found for its " base case" an unoxidized zirconium fraction of 41 percent at the start of melt-concrete interaction. Wis fraction ranged from 16 percent to 69 percent for the range of input parmieters reflected in Table 7-1. We sensitivity analysis represented in Table 7--2 was for the VANESA code only, and was J not limited to this range. Wus, a range of unoxidized fraction from 0 percent to 80 percent appears in Table 7-2. 7.7 A camputer code can jive erroneous results even if the theoretical model on which it is based is correct. Wis occurs because the codeIfuses the approximate solutions to the model's mathenatical equations. J code is badly written, or if computer time suf ficient for an accurate solution is unavailable, those approximations can lead to substantial error. 7.8 The most representative source term experiment to date occurred at the Loss of Fluid Mst (LOFT) reactor in Idaho in July 1985. During this
' experiment, part of the fuel was allowed to melt, and radioactive material was carried throtx3hout the RCS and deposited in the discharge vessel.
7.9 Both WRs and BWRs employ a zirconim-based alloy as a fuel cladding. BWRs also employ zirconim in the " channel boxes" which surround the ~) fuel assenblies. 7.30 Under certain conditions, the falling of molten material into water will generate a violent steam explosion which will change the progression of the accident. Instead of the relatively slow evaporation of water assuned here, the reactor vessel could be breached (for an in-vessel E) explosion) or the core material could be dispersed into the containment (for both in-vessel and ex-vessel explosions) . Gapter Nine provides a discussion of these po sibilities. O
7-24 } 7.11 In certain conditions,. a flame front can accelerate, ultimately becming a shock wave. In its full-fledged form, this constitutes " detonation". Were is a transitional- region of " quasi-detonation", where a cmiplete }. shock wave is not formed. Wese phenmena, described in Chapter Nine, can lead to substantial contaiment loading. 7.12 te exception is Te, for which the effect of the unoxidized zirconim fraction was considered in BCL's source term work. Wis was done in recognition of observations that Te release is substantially reduced when unoxidized Zr is present. 3 7.13 Even this level of uncertainty requires detailed knowledge of the initial cmiposition and power history of each fuel rod. In the absence of such knowledge, uncertainty will be greater. 7.14 If the containment atmosphere is rich in hydrogen, nitric acid 3 production will be much reduced. Also, the presence of alkaline water (fra contaiment sprays) will counteract the effect of the nitric acid. 7.15 " Decontamination" refers to reoval of radioactive material from the release stream. We decontaination factor (DF) is defined as the ratio of the mass entering the pool / ice-bed to the mass leaving. 3 7.16 Were are also ciremstances in which the suppression pool water or ice would be absent. Pool water could leak through a contaiment breach. Ice beds may have melted at a time when releases into the lower conr partment are still occurring. D D D D t I
)_
7-25
'7.7 References for Chapter Seven } DENNING, 1984 R.S. Denning, " Perspectives on %chnical Issues Evolving Fra the Source . Term Reassesment," .Battelle Colutbus Laboratories (Colutbus, .0hio) , paper presented at the NRC's 12th Water Reactor Safety Research Information Meeting, Gaithersburg, Maryland, 22-26 October 1984. J D- GIESEKE, ET AL., 1984a-e J. A. Gieseke, P. Cybulskis, R.S. Denning, M.R. Kuhlman, K.W. Ime, and H.
Chen, Radionuclide Release Under Specific LWR Accident Conditions,
- s. Battelle Colunbus Laboratories (Colunbus , Ohio), BMI-2104, Vols. II-VI, draft, July 1984.
D- GITTUS, ET AL., 1982 J .H. . Gittus, et al., PWR Degraded Core Analysis, Springfields Nuclear (S), Power Developtent Iaboratories, UK Atmic Diergy Authority, ND-R0610 g April 1982. KRESS, ET AL., 1985 g-- , T.S. Kress (Editor) , et al., Review of the Status of Validation of the Computer Codes Used in the Severe Accidnt Source Term Reassessrrent ORNL/TM-8842, Study, April Oak Ridge National Laboratory (Oak Ridge, Tennessee) , 1985. LIPINSKI, ET AL., 1985 g R.J. . Lipinski, D.R. Bradley, J.E. Brockmann, J.M. Griesmeyer, C.D. Iaigh, Tills, and D.C. Murata, D.A. Powers, J.B. Rivard, A.R. Taig, J. K.K. I Williams, Uncertainty in Radionuclide Release under Specific LWR Accident Conditions ,~Sandia National Laboratories (Albuquerque, New Mexico) , SAND 84-0410, Vol. II,-draft, February 1985. NRC, 1985 U.S. Nuclear Regulatory Conmission, Reassessnent of ~ the Technical Bases Regulatory- Commission for Estimating Source Terms, U.S. Nuclear (Washington, D.C.) , NUREG-0956, Draf t 1, 24 May 1985. RIVARD, ET AL., 1984 g Behr, R.G. Easterling, J.M. Griesmeyer , F.E. Haskin, J.B. Rivard, V.L.
. S.W. Hatch, A.M. Kolaczkowski, R.J. Lipinski, M.P. Sherman, A.R. Taig, and Sandia A.J. Wickett, Identification of Severe Accident Uncertainties, National Laboratories (Albuquerque, New Mexico) NUREG/CR-3440, Septenber 1984.
SHAFFER, ET AL' ,1984 . C.J. Shaffer, R.D. Gasser, F.E. Haskin, and V.L. Behr, "MARCON-Developnent and Applications to Containnent Inading Calculations," Sandia National Laboratories ( Albuquerque, New Mexico) , paper presented at the NRC's 12th Water Reactor Safety Research Information t%eting, Gaithersburg, Maryland, 22-26 October 1984. g O
O 7-26 WILSON, ET AL., 1985 R. Wilson, et al., Report to the Anerican Physical Society of the Study Group on Radionuclide Release from Severe Accidents at Nuclear Power O Plants, draf t Ebbruary 1985, to be published in Reviews g Modern Physics. WOOTEN, ET AL., 1984 R.O. Wooton, P. Cybulskis, and S.F. Quayle, IGRCH 2 (l'.altdown Accident Response Characteristics) Code Description and User's Manual, Battelle Colunbus Laboratories (Colunbus, Ohio) , NUREG/CR-3988, Septenber 1984. O O.
.O O
- O O
O
'O O
7-27 8 TABLE 7'- 1
- r. Sensitivity of Source Term Estimation to Code Inputs, O. for a Station Blackout Sequence at the Surry PWR T
e Early Late 4 Contalment Failure Contaiment Failure Highb go ,b gg) Highb- gg)
- O ' Release Release Release = Release. Release (Stiall (Larger
. hole) . hole)
Asstmed Containnent Failure Conditions . Failure 173 141 ~141 900 900 QL time - (minutes) Hole size (m ) 10 0.001 0.01 10 0.001 Ctnulative Release af ter 10 Hours of Contaiment Leakage
- O (i) cMass (k9) _10 _10 CsI 20 0.076 0.72 3.6x10 2.5x10
]- c 0.31 3". 0 4.2x10-2 1,ggyg-9 Cs0H 77 4 c -3 Te 0.6 0.0038 0.36 0.58 4.4x10 'O. RefractorydFission 1.1x10
-3 H. Products 21 0.15 1.5 1.5 ,
Total
- 600 4.5 44 53 1.34 (ii) Radioactivity 300 1.75 16.7 4.8 2.9x10-2
'O (million Ci)
(iii) Fraction of' activity suspended in containent atmosphere at tine of contaiment failure which -is releasal 1.0 0.019 0.18 1.0 0.069 !O l Notes a Wis table adapted from Table 6-7 of [LIPINSKI, ET AL.,1985] . ! lg. ' b Parmeters expected to lead to releases at the low or high end of their i ranges were fed into the l ARCH, CORSOR, TRAPMELT, CORCON, AND NAUA codes. i c We estimated core inventory at shutdown includes: 12.4 kg of I,131 kg of Cs, and 25.4 kg of %. d "Fefra t ry fissi n Pro 6ucts" include: Ba , La , Ce , Sr , til, Pr , and Sti, as !O well as the actinides Np and Pu. l ! e his includes non-radioactive materials. ! i EQ _ _ -_ _ _ _ ,_ _ _ __. _ _ _ __ _ _ _ _ , _ _ _ _ _ _ ]
0: 7_28 TABLE 7-2. O Sensitivity of Core-Concrete nelease Estimates to unoxidized zirconim.. ' Content of Melt, for a Station Blackout Secuence at the Surry PhR Cuuulative Fraction (c) of O Fraction (b) of Zr Radionuclides Released to in telt which Contairunent Atmosphere is Unoxidized After 10 Hours I Sr La Group ( I C,e e Group *) 0.001 sx10-46 2x10-s 0 0 sx10 0.001 0.41 0.12 0.70 0.30 0.20 0.80 0 notes (a) 'Ihis table adapted from Table IV.B.2 of: Wilson et al, 1985. That table in turn drew upon results in Appendix L of: Lipinski et al, 1985. (b) Fraction at time core-concrete interactions begins. (c) Fraction of initial core inventory. (d) Incitzles: Eu, G3, La , t&i, Pm, Pr , Sn, and Y. (e) Includes: Ce , Np , . Pu . O O O
7-29 () TABLE 7-3
'O Comparison of Contairment Conditions Predicted !
by MARCH (INTER) and MARCH (CORCON) Codes PWR Results ID'd) BWR Results IC ' *I -Q MARCn MARCS MARCa MARCS
- (INTER) (CORCON) (INTER) (CORCON)
Contaiment Pressure (psia) 92.2 90.7 124.7 154.1 Contaiment Teperature ( F) 291 332. 505 615 10 Gases Released in Core - Concrete Interactions (lb)
- O CO O 86,10e 3,600 45,870 CO 2 276,500 75,300 87,040 2,710 H 330 2,490 130 1,520 2
H 46,340 16,ese 16,680 34,600
- O 2 i!
Notes O (a) 'Ihese results taken fra Tables 2 and 3 of: Shaffer et al, 1984. (b) A PWR with a Large Dry contalment is assmed to suffer a station
- blackout accident.
(c) A Mark II BWR is assmed to suffer a 'IUN accident. O (d) Cbnditions are shown for a tine 25 hours after accident initiation (23-1/4 hours af ter vessel failure) . (e) Conditions are shown for a tine 3 hours after vessel failure. O .O .O
7-30 I O FIGURE 7-1 SELECTION OF TYPES OF PLANTS
.r SELECTION OF SPECIFIC PLANTS SELECTION OF ACCIDENT SEQUENCES ,e . INVENT 0RT SPECIF! CAT;0N OF PLANT ---- ---
GECMETay A40 ACC! CENT ORIGEN SEQUENCE HENOMENA g
)
CVERALL THERM L HYORAtLIC5 MARCH o 't
) RELEASE FROM FUEL PRIMRY SYSTEM --- --------
THER VL HYORAut.!CS COR50R MERGE
,e ' CCRE. CONCRETE INTERACT!0k PRIMRf SYSTEM TRANSPORT I- -------------- ; ec,co, TRAP-MELT " - CCRE CONCRETE RELEASE q
ej CONTAINMENT TRANSPORT AND VANESA
- POOL SCRU581NG l
MAUA-4 AND $PARC :
,r ^ )
RELEA$t TO ENv!RcNmtNT cy s Roles of Computer Codes in NRC-Sponsored Sourco Term Estimations (Source: Kress et al, 1985) 1 e,
n --- _ O. 7-31 FICURE 7-2 .O i J
- O i
CONTAINMENT SPRAY l0 I h STEAM
<0 I R/S VALVE OR BRExx l
10 t >> l RWST- . SG O
------ ACCUM O CSS AFW G h
- V
<0-- ECCS CORE WDED O CSHX b
O REACTOR ECCHX
~~~---
CAVITY 1- . ~. . _ ,_ 4 Sus? O O Representation of a Nuclear Plant by the MARCII rode (Source: Kress et al, 1985) O
, 7-32 m, .~, FICURE 7-3 . l -; l l0 l l CODE COMPUTER _ CODE MODIFICATION a n ,
- O SENSITIVITY DEVELOPMENTAL ,
P ANALYSIS EXPERIMENTS h (O 7 RANKING OF SENSIT!VITY l<
. l ' SET QUALIFICATION l '
O CRITERIA
! l t
SPEctFY-RANGES FOR = / !O / VARIABLES / 3
] / '
VALIDATION [
- O EXPERIMENTS /
' y
/ ; /
ANALYSIS OF EXPERIMENTS / ADDITIONAL EXPERIMENTS AND COMPARISON & OR CODE MODirlCATION DECISION O. w!Ts CODE y i VALIDATED CODE fO : I f i l 'O : !; An Approach to Code Validation (Source: Kress et al, 1985) l ,O :.
. . _ . . . . _ . . , - - - _ . _ ~ _ . ~ . _ _ . . . . _ _ . . _ . _ . _ , ~ _ , . _ . . _ . _ . _ . _ _ . _ . . _ _ _ . . _ _ . _ _ - _ -
7-33 O FICURE 7-4 y O O 2.sn . . . . U O E X PE RIMC N rA L DATA ? 2.300 - l Temperatiire (thousand oK) g ,,,g ,
< CORCON Predictions . .sco ,
O . 9 g aM r.S I.
' t.Too =
O
,,3,,
2.co 3.co 4.00 o.co 1.co i Time (thousand seconds) O o Comparison of CORCON-Predicted and Measured Melt Temperatures for a Core-Concre interaction Experiment at SNL Notes and Sources Kress et al, 1985. (a) This figure adapted from Figure 4.7 in Part V of: g (b) The inittal melt temperature was 1975 0 K. K. (c) The assumed concrete ablation temperature was 17500
,y
3 7-34 APPENDIX TO CHAPTER SEVEN ,0 Objectives, Strengths, and Limitations of the MARCH 2 Code: A Perspective from the User's Manual The MARCH 2 code is central to the set of codes used in the NRC-sponsored source term estimation effort. 'Ihe following quotations frm the user's manual g for this code -- dealing in sequence with the objectives, strengths, and . I limitations of the code -- are instructive [ WOO';0N, CYBULSKIS & QUAYLE,1984): g THE CODE'S OBJECTIVES:
"The MARCH code has been developed to analyze the themal-hydraulic response of the reactor core, the primary coolant syste, and the contalment "Y8t* i"liht""t***
9 "YSt"8 i" P "S* t "CCid*"t" I"" 1"i"9 * ** lO level of engineered safety features inoperability. While MARCH is primarily intended to address accidents leading to cmplete core meltdown, it can also be used to treat events involving only partial core degradation as well as for assessing the minim m levels of engineered safety feature operability required to cope with various accident events. Depending on the particular application, MARCH analyses can be conducted by theselves or in combination with fission i In the latter case, product transport codes such as TRAP, CORRAL, and MATADOR.
"^"C" " "ld P' "id* th' **'*"ti^1 th*'*^1-hW'*"li i"P"' " "d i t i "" '*9" I '*d lO by the fission product transport codes.
B
" MARCH is designed to cover the entire accident sequence, fra the initiating accident event through the attack of the contalment basmat, for a
'g variety of accident initiators and including coverage of a wide variety of reactor syste designs. More specifically, depending on the reactor design and accident sequences, the code can evaluate: O (1) Heatup of the primary and secondary coolant inventories and pressure rise to the relief / safety valve settings with subsequent boiloff, (2) Initial blowdown of the primary coolant for mall and intermediate O breaks in the primary systs, O
t 7-35 - ns (3) Generation and transport of heat within the core and associated coolant, if any, includirg boiloff of water fra the reactor vessel, 3 (4) Heatup of the fuel following core uncovery, including the effects of metal-water reactions, m (5) Melting and simping of the fuel onto the lower core support a structures and into the vessel bott a head, (6) Interaction of the core debris with residual water in the reactor vessel, v o (7) Interaction of the core debris with the reactor vessel bottom head and the meltthrough or pressure-driven failure of the head, 0- Interaction of the core debris with the water in the reactor cavity, (8) including effects of chmical reactions, (9) Attack of the concrete bassat by the core and structural debris, g (10) We relocation of the decay heat source as fission products are released from the fuel and transported to the contaiment, .O (11) Mass and energy additions to the contaiment associated with all the lt; foregoing phenomena and their effect on the contalment tmperature and pressure response, including the effects of passive and active
" "'****"* **f*'Y f***"'** ' as aPPr Priate, O
(12) We effects on the contaiment pressure and tmperature response of the burning of hydrogen and carbon monoxide, and lO ' (13) teakage to the enviroment after failure of the contaiment." (Pages 2-1 and 2-2]
'O i 'O
7-36 O ITS STRENGTHS: m "te differences between MARCH 1.1 and 2 include corrections of errors and v changes in models, code structure, and progranmirn language. %e new model which are included in MARCH 2 were developed at a nmber of institutions incitrlity Battelle, Sandia National Laboratories, Oak Ridge National Labora-t ries, Br khaven National Laboratcry, and the %nnessee Valley Authority. O Wese new models are in sme cases provided as options to existing nodels. We changes in MARCH were largely undertaken to address some of the recognized deficiencies in the early version related to modeling approximations, time-step n control, and transportability of the code to cther installations. v
" MARCH 2 incorporates the current Anerican National Standard for Decay Heat Power in Light Water Reactors for evaluating fission product decay heating as a function of tine af ter shutdown and time at power, including the contri-O butions due to heavy elment decay. %is replaces the earlier, simplified version incorporated in MARCH 1.1. Alternatively, decay heat as a function of time may be input in tabular form; this approach would be particularly appropriate for transients in the absence of scre where the power history O would be provided by more detailed systes codes. "We representation of the properties of water and steam in MARCH 2 has been improved over that in MARCH 1.1. Wis has included expansion of the O,
property tables and correlations incorporated in the code such as additional properties required by the new phenmenological nodels.
"We MARCH 2 treatnient of the primary systs includes improvments in both O the initial (early) primary systs response as well as the addition of several phenmenological models to treat the processes followiry core collapse into the bottmi head. Included are changes in the steam generator model to raiove some of the restrictions and limitations of the earlier version, improved break flow
'l' models, changes in the flashing model in response to primary syste pressure changes, provisions for simultaneous break and relief / safety valve flow, changes in the treatnient of gas phase heat transfer to structures, and consid-eration of the holdup of released fission products within the primary cystm. m O
O 7-37
" MARCH 2 retains the basic model of the core as developed for the earlier version, but incorporates a nmber of additional models for a more detailed O treadient of heat transfer processes. We heat transfer between the fuel rods and the stean-hydrogen gas mixture is now calculated using either the full A
Dittus-Boelter correlation for turbulent flow or a laminar flow correlation. subroutine has also been added to approximate axial conduction heat transfer in O the fuel rods using the Eburier law of heat conduction and the BOIL-calculated node taperatures. We ef fect of axial and radial thermal radiation heat transfer within the core as well as between the core and surrounding structures and water surfaces can now be calculated. Included is the heatup of the core O support barrel by thermal radiation. Additional changes have been made in the heat transfer analysis of partially covered core nodes and in the metal-water reaction model. A BWR core model providing an explicit treat::ient of channel I boxes and control blades is incorporatal. O "A nmber of phenmenological models have been added for the treatment of core debris and water interactions in the reactor vessel. tese range fran a flat plate critical heat flux model, to a fragnented debris-to-water heat O transfer correlation, to several options that consider formation of debris beds within the core and bottom head s ile water is still in the vessel. We bottan t head heatup model utilizes a calculated heat transfer coefficient between the molten debris and the vessel head. O "A major area of concern and controversy in the analysis of core neltdown accidents has been the behavior of core and structural debris upon contact with water in the reactor cavity. We isolated-particle models incorporated in O MARCH 1.1 have been supplatented with the addition of the treatment of several debris configurations; these include a flat plate critical heat flux model, a particulate heat transfer model with improved heat transfer coefficients, and [Pages 2-25 and 2-26] several debris bed heat transfer correlations." O ITS LIMITATIONS:
"Ebr phenomena which are not well understood, there are a neber of O user-specified options in the code that may be selected to explore the effect of various modeling assumptions. In all cases, mass and energy are conserved O
O 7-38 ) so that calculated sequences are self-consistent. Were is generally no deliberate bias in the coding to produce ' conservative' results; however, the [O code user may by choice of input cptions produce ' conservative' results. We i choice of these user-specifled options can in many instances greatly influence the results of the calculations." (Page 2-2] .O mese calculations indicate MARCH can do an adequate job of calculating core tmperatures during the initial heatup phase prior to nelting and gross core distortion. However, this agreenent should not be extrapolated beyond the experimental data base into the melting and slunping phase. We neltdown O models resin unsupported by direct experimental validation for this regine of a behavior." (Page 2-28] 4 "In stmnary, the MARCH 2 cmputer code models a large nunber of emplex 0 interacting phenmena. In sme cases, the basic processes are well known, and the models are supported by a significant data base. In general, however, there is very little experimental basis with prototypic materials and severe enviroments to support the modeling of the emiplex interacting phenomena O associated with core meltdown itself. MARCH 2 was not developed with the intent of modeling all aspects of core neltdown accidents in detail. Rather, i the intent was to provide a tool for risk analysis which described in reasonable detail the approximate timing and conditions of fuel degradation and O contaiment failure. It is well suited for application to paraneter and g sensitivity studies. We user should be aware that because some of the - impo r tant phenomena are not well understood, the uncertainties in the predictions of a given MARCH 2 calculation may be large. It should also be O recognized that there are many assmptions not related to the MARCH code that are required in the analyses that may also introduce substantial uncertainties; anong the latter are definition of the accident sequences incitriing failure and success criteria for various engineered safety features, nature and timing of O operator intervention, contaiment failure criteria, etc. We major areas of uncertainty in the MARCH 2 models are believed to include: 0
- 1. Pelocation and simping of nelted fuel as they af fect steam and hy3rogen generation
l . (' 7-39 f).
- 2. Debris behavior in the lower plenun of the reactor vessel
- 3. vessel failure mode
- 4. Debris-coolant interactions in the reactor cavity
- 5. Ignition of cabustible gases (hydrogen and carbon monoxide) .
Many f the above phenomena will always have large associated O uncertainties that cannot be eliminated by more sophisticated modeling. In general, however, we feel that the MARCH code predicts thermal-hydraulic conditions that are representative of those . in severe accident sequences." (Page 2-31) h O O O 4 O O O O
T . CHAPTER EIGHT ) Contaiment Performance Under Internally-Initiated Accident Conditions )
SUMMARY
We contaiment buildings of nuclear power plants are large steel and/or concrete structures designed to contain the radioactivity that might escape fra the fuel during a severe ) power plant accident. If the contaiment maintains its integrity during the accident, little radioactivty will escape and the source term will be small; conversely, if the contalment fails, the source term for the accident could be large. Wis chapter revietws the uncertainties involved in contaiment performance during accidents. Were are a nmber of serious accidents which contalments are not designed to withstand: external blast waves, internal missiles fra coolant pmp fly
- eels, the rupture of a major emponent such as the pressure vessel, or severe hydrogen explosions. Additionally, there are sequences where 3 radioactivty could escape without containment failure:
so-called " bypass" events or sequences where penetrations (hatches, valves and the like) through the containment builditg were left open. Given the wide variation in plant designs, it is clear that 3 generic conclusions cannot be drawn fra even detailed attriy of so-called " representative" plants. Current source term estimates rely on arbitrary assmptions about contaiment performance. Substantial improvments are required in both the analysis of contalment performance and in the supporting program of experimental research. 8.0 Introduction We contaiment buildings of nuclear power plants are large steel and/or concrete structures surrounding the pressure vessel and sme of its associated ] equipnent. In turn, the pressure vessel contains the highly radioactive core. tese contaiment structures are intended to "contain" radioactivity which might be released fra the core during a major accident. 'Ib the extent that f contaiments can survive accidents without releasiry redioactive materials, ] l l
)
8-2 the corresponding source terms will be small. Conversely, containment failure can lead to large radioactive releases, that is, to large source terms. We ) purpose of this chapter is to review the present understanding of, and the uncertainties surrounding, contalment performance during severe accidents. Reactor accidents can be initiated by either " internal" or " external" ) events. Although the distinction between these two types of event is imprecise, it is generally understood within the community of accident analysts. Exmples of internal initiators are: loss of electrical power, pipe break, or operator error. We category of external initiators includes: ) earthquake, sabotage, aircraft crash, or an exploding gas cloud. In this chapter, only internally-initiated accidents will be addressed. Wis is not because the external initiators are insignificant, but because current source term investigations, such as those sponsored by the US Nuclear Regulatory ] Cmmission (NBC), have focused on internally-initiated accidents. We begin with a brief outline (Section 8.1) of the history, design criteria, and categories of reactor contaiments. Ebilowirg this is a brief discussion (Section 8.2) of the importance of contalments in mitigating accidental D radiation releases to the enviroment. Section 8.3 discusses sme of the design aspects of U.S. contalments, along with a record of their integrity. A brief discussion of contalment loads during accidents is provided in Section 8.4, and the nature of the responses of contaiments to those loads is ] addressed in Section 8.5. @cn, Section 8.6 reviews the status of knowledge on contalment response, the need to validate emputer models, and the relevance of existirg research progras. Section 8.7 smmarizes sme of the ] research questions that still need to be addressed. Section 8.8 presents our conclusions. 8.1 The Role of Reactor Contalments D 8.1.1 A Brief History l During the cmparatively short history of the international nuclear power j industry, there have been widely differirn opinions on the need for ard rolo of reactor contalments. In the United States, contalments have been l D
jO 8-3 1 provided from the beginning. By contrast, the Soviet Union has only recently ! begun to require then. West Germany, on the other hand, uses more elaborate lC contaiment systens than does the United States. ! We Anerican experience has been samarized (AHEARNE, 1982] by John Ahearne, then a Oxteissioner of the MC. MJch of the natorial in this section O is drawn fra that source.9*1 One of the earliest statments on the need for contalments was made in 1956 Congressional testimony by Willard t,1bby, Acting Chairman of the US O Atmile D'er9Y C='nission (AEC) : It is expected that power reactors such as that now under construction at Shippingport, Pennsylvania, will rely more upon '. the philosophy of contaiment than isolation (that is, remote . sitim , ni.] as a means of protectim the public against the .O consequences of an improbable accident... Although the ADC had fra the beginnim required contalments, there was, ; and has reained, confusion as to their exact purpose. It was smetimes i O stated that the contalment should protect the public fra any conceivable accident. Yet the criteria to which contalments were designed did not usually encanpass the full range of accidents. In illustration of the " total contalment" [hilosophy, Dr. Clif ford Beck of the ABC's regulatory staf f wrote 1" 1959 O We external contalment vessel, as a barrier of last resort against releases of radioactivity to public areas, offers a unique protection, cmpletely independent of all other safety devices and engineerim safeguards, and its dependability is O unaffected by errors in safety analyses and strigment of the reactor assa bly. At certain sites with high population densities, reactor designers actually sought to provide such a high level of protection, for example at the O Ravenswood plant sich the Cbnsolidated niison Oxnpany proposed in 1962 to construct in Queens, New York. Each of the two reactors was to be surrounded by two consecutive structures consisting of stool shells backed up by 5 1/2 feet and 2 1/2 feet of concrete, respectively. We spwe between these O structures ws to be filled with porous concreto. Wis contalment synts ws O t
g 8-4 intended to ensure "no leakage" of radioactivity following any conceivable m accident, including core melt. Eventually, however, the entire Ravenswood
'v proposal was withdrawn by Consolidated Edison. In none of the reactors constructed later were the contairment design criteria as stringent as they wre for fuvonsw:cd.
By the mid-1960s it had becme widely recognized that the contairnents of , most reactors would leak or fail in the event of core nelt. Contalment failure becane even more likely as early reactor designs were scaled-up and as zirconim e me into use as a fuel cladding material.8.2 By the time the NRC's Reactor Safety Study (WASH-1400) was empleted in i 1975, there was little dispute that the contairnents of existing power plants could fail as a result of certain severe accidents. For the past few years iO debate has centered on estimating the probabilities both of core daage itself and of accmpanying loss of contaiment integrity. Most recently, as part of the source tem effort, there has been an expanded research progre on contairment failure modes during severe accidents.
.O 8.1.2 Containnent Design Criteria i
Contairnent design criteria were developed in parallel with siting and emergency planning regulations.8.3 These began in 1950, when the AEC reemmended a population " exclusion radius" *ich varied with powr icvel. For a 30 megawatt thermal (MWt) research reactor of the early 1950s, this radius would have been 1.7 miles, Whereas for a modern 1000 negawatt electric (MWo) powr reactor, it would be 17 miles.8.4 As larger reactors were built,
! more reliance came to be placed on contairnents, and the AEC's exclusion radius reemnendation was ignored. Until 1962, however, siting decisions wre ,
mad n a plant by plant basis. O In 1962, the AEC adopted the siting regulations in 10 C.F.R. Part 100. In ef fect, these regulations requirtd that a contairnent ord its associated f systms such as sprays and cican-up filters, together with appropriate siting, O should assure that no mmber of the public would receive more than a 25-tm whole-body done or a 300-tm thyroid dose af ter any " design basis accident" i
.O
8-5 r ~ , (DBA). We DBA as a regulatory concept originated in 1959, under the name l l " maxima credible accident." Although the DBA concept has been central to !O I reactor safety regulations since 1959, the dividing line bethn " credible" and " incredible" accidents has never been precisely defined.' Indeed, the
~
definition of a DBA has been some4at fluid, depending on cirematances. We spectre of DBAs considered for existing plants includes events such as the breaking of large pipes in the reactor coolant sybi.e (RCS), but does not include core melt or extensive core d mage. For the, purposes hf'10 C.,F.R. Part 100, it is assmed that a DBA might lead to a release frca the core to O the contaiment of up to 100% of the radioactive noble gases, 50% of the I radioactive iodines, and it of solids. These essentially arbitrary percentages were taken fra a 1962 AEC docment, TID 14844, which in turn was s based on 1960 recomendations of the Advisory Comittee on Reactor Safeguacds (ACRS) and the AEC staf f (DiNUNNO, ET AL.,19621 Generally speaking, in designing nucloat plants there has been no requirment to consider contsiment loads arising fecm severe core dmage. A ' lO partial exception is that some reactors have in recent years been required to mploy hydrogen control measures designed to prevent or mitigate the cebus-tien of the hyltogen produced if 75 porcent or more of the zirconim in the fuel cladding reacts with water. However, most operatirn reactors are O designed with the assmption that no more than 5 percent of the zirconim will be oxidized. l i. In contrast with 10 C.F.R. Part 100, other federal regulations,10 C.F.R. O Part 50, are intended to provido "roasonable assurance" that reactors are so designal and constructed that sovero core damage will not be likely. We rotation betwon Parts 50 and 100 10, however, confused. The siting i requirements of Part 100 do assumo that Part 50 will not be totally l 'O successful, thoujh thoro is no explicit 'assmption that it will totally fall. Instead, there is a vague and inconsistent overlap, so that some failures to noet Part 50 are allowed for under Part 100, and scrte are not. Wis confusion is reflected in the critoria adopted for contalment design. O i O
3 8-6 Contalments in the United States are designed to withstand tornado loads and a Safe Shutdown Earthquake (SSE) .8.5 are aim desigM for loss-of-g
! coolant accidents (LOCAs) involving broken pipes in the RCS, and for accmpe-nying phenmena such as the reaction loads fra broken pipes. A concurrent LOCA and SSE are considered in the design, but a IDCA in a pressurized water reactor (IWR), concurrent with a secondary circuit failure inside contalment, g ;
is not. t Neither external blast waves -- for example frei exploding gas clouds -- I nor the inpact of missiles fra a ruptured turbine are considered in contain-g Nor is the impact of large aircraft, with exceptions such as ment design. plants located near airports, as is Wree Mile Island. O We 9erati n f internal miniles fra Ge WPture of 6e flyseels Nor is the rupture of a i coupled to reactor coolant peps is not considered. i major ca ponent of the RCS (vessel, pmp, stem generator, pressurizer) incitaled in the spectra of teas. Also, there is no general requirement to l l consider the failure of a support for any piping syste or emponent. 4 J Finally, there is no requitstent to design the contalment for loads . (static and dynmic pressurization, taperature, and radiation) which could .
' d d"' i "' ""' * *
- d "*9' *** Id*""' ' ^ * "Sh ' * * ' '*d "* ' ""*
- O hWrogen control systes are now required, there is no requireent that the i contaiment be able to withstand severe hy!rogen explosions.
In stmtiary, the present generation of contalments has not been designed to cope with many of the severe accidents which are being considered within 1 i the source tem controversy. Any ability which these contalments possess in this respect is largely the result of conservatim in their design. ,. y' 8.1.3 contatritent categories . t Pressurized Water Reactors (WRs) and Bolling Water Peactors (BWRs) dif fer significantly in design, with important implications for their contain-0 nents. For exmple, BWRs contain more cooling water in their RCS than do l WRa, and have a heavier coro, as shown in table 8-1. Partly as a result, all i i i 10 I a- .v ,,--, .- -.,--wm.,__....nn,_. y,m-.,,.nn n .n m . _ _ . _ , _ , , , , _ . ..---~,,w,m,-m, .e-.-- ,e,,-
'3 8-7 h ) ' .i modern DWR contaiments sploy pressure suppression features, whereas most m'R L contaiments do not. Such features are intended to reduce the contaiment [ pressure following a severe IDCA. 0 2 Pressure suppression is achieved in BWRs by directing the stean generated during the LOCA to pass through a large pool of water. $ree distinct types of contaiment have been developed for modern BWRs in the United States, and are designated Marks I through III. Figures 8-1 through 8-3 illustrate these t j
! contaiments. We current inventory of BWR contaiments in the United States, categorized by the nethod of pressure suppression, is given in Table 8-2. - hhere pressure suppression is mployed at m'Rs, condensation on ice is used. In the Ice Condenser contaiment, a large mass (about 1,000 tons) of ice is held in baskets in such a manner that stean released during a LOCA will g
A pass through these baskets. Figure 8-4 illustrates the systs. All other W'R contaiments consist of large structures without any specific pressure suppres-sion equipnent, although the contaiment sprays and air coolers can serve a limited pressure suppression function. Figure 8-5 illustrates the arrange-g Scne of these contaiments, described here as large dry subat:nospheric ment. contaiments, are partially evacuated during reactor operation. 6 We reain-j der, described here as large dry atmospheric contaiments, are operated at
;s atmospheric pressure. Table 8-3 gives the current inventory of W'R contain-O
- ments in the United States, categorized by the nethod of pressure suppression.
't Wese various BWR and W'R contaiments are characterizcd by widely , varying free volunes and design pressures, as shown for see typical plants by the, data in Table 8-4.
8.2 Significance of Contaiment Behavior During Accidents q ' 8.2.1 In;plications of Timing, Location, and Size of Breach i
\ In the event of a core nelt accident, the amount of radioactive material that escapes to the atmosphere will depend critically upon the contaiment 9 failure mode. The timing , location and size of the breach will each be significant, to differing degrees for each plant and each accident sequence.
I A
0- ,_, Assmptions about such parameters are important inputs to the NRC-sponsored AL., 1984] source term estimations {GIESEKE, ET AL., 1983; GIESEKE, ET O conducted by Battelle Cblmbus Laboratories (BCL) . I We importance of the length of time until the contalment fails is illustrated by Figures 8-6 and 8-7, which show the estimated containment in the contaiment, for a specific
" pressure and mount of airborne material accident at the Surry WR. We accident is a 12 rge-break loss of coolant accident occurrire at the see time as a loss of all electric power, both on and off site. With no electric power to operate the cooling systes, the core O overheats and melts. Although a rapid hydrogen burn is assmed to occur early 5
We pres-in the accident, it is assmed that the contairment does not fail. sure in the contaiment continues to increase until the contalment eventually 9 fails, about 42 hours into the accident. O Figure 8-7 shows that a contaiment failure during the first few hours of the accident, from either the hydrogen burn or a failure in contaiment isolation, could lead to a much larger release than would arise fra a delayed - O contaiment failure. More material would be airborne in contaiment at earlier times, and thus be available for release.8.7 i: We location of the contaiment breach is particularly important for BWRs ib and NR ice condenser plants. If the breach is located so that escaping radioactive materials will pass through the suppression pool or ice bed before then the magnitude of the release may be being released to the atmosphere, ! significantly reduced. Alternatively, the breach may be so located so that Moreover, if a n V such " scrubbing" of radioactive material does not occur. breach occurs in the vicinity of a BWR suppression pool, then the pool We possibility of integrity may be lost, leadiry to loss of pool water. scrubbing in the pool is then eliminated entirely. In see instances, a release of radioactivity will be mitigated by structures or systens outside the contaiment. Wis possibility, addressed e
- further below, provides a further illustration of the importance of the O location of a contaiment failure.
i 0.; I
A V 8-9 he size of breach is clearily important, but the implications of breach size must be considered carefully. Notably, a mall breach early in the acci-O dent might relieve pressure sufficiently to prevent a large breach later. For example, leakage around a personnel hatch or drywell head could be sufficient to prevent catastrophic failure of the contaiment wall at a later time. Such a " leak-tef re-break" phenmenon might, or might not, reduce the source tem, O 8*8 depending on the accident sequence. 8.2.2 Contaiment Bypass Events Q In many accidents, radioactive materials will be released first frm the RCS to the contaiment, and then fra the contaiment to the envirement. However, in another class of accidents - the contalment bypass events -- n materials will be released directly fra the RCS to the enviroment, v his possibility is widely recognized for PWRs. In those reactors, failure of a check-valve in the low pressure coolant injection systs (an event V) r multiple steam generator tube ruptures can lead to a loss of O coolant and subsequent release of radioactive material without a direct breach of the contaiment. Fr BWRs, the main steam line offers a direct path frm the RCS to O Wus , inccruplete closure of the main steam regions outside the contaiment. valves wul create a bypass event. 8.2.3 Mitigation of a Release by Plant Features Outside Contaiment O Were are a nmber of possible neans whereby a release frmi containment may be mitigated. For exanple, many bypass events involve passage of the As an illustration, an i radioactive discharge throtsh adjacent buildings. v event V release might be mitigated by scrubbing in a pool of water in the auxiliary building. Releases through a direct contaiment breach may also be mitigated by the structures and supporting equiptent in the release path, mny plants have a reinforced concrete shield building surrounding the contaiment, as shown in !
)
1 O
O 8-19 Figures 8-3 and 8 this building could be the site of fission product deposition. In the case of Mark I and Mark II BWRs, as shown in Figures 8-1 O and 8-2, adjacent buildings empletely surround the principal contaiment. 8.3 Design and operation of US Containments O 8.3.1 Types of contaiments As mentioned in Section 8.1, several types of contaiments are used in U.S. light-water reactors. Tables 8-2 and 8-3 give the current inventory, O categorized by the means of pressure suppression. Within each category, there is a variety of construction types. Table 8-5 shows the current inventory for BWRs. We Mark I steel contaiment is the most cmmon type, accounting for 22 of the 46 BWRs now operating or under construction. Table 8-6 shows the inven-tory for IMRs. Here, the large dry contaiment in the form of a prestressed O concrete cylinder is the most emmon type, accounting for 41 of the 85 IMRs in the current inventory. Typical design pressures of the various contaiments are shown in Table 8-7. O Within some of the construction types shown in Tables 8-5 and 8-6, there is a further variety of designs. One Mark II BWR contaiment (Shoreham) sploys a truncated-cone body instead of the truncated-cone and vertical-O cylinder body shown in Figure 8-2. Large dry IMR contaiments with pre-stressed concrete vertical cylinders may have shallow or heispherical dmes and zero, three, four, or six buttresses in their vertical walls. It is clear, therefore, that there is a wide variety of different
- O contaiment designs in use. Using the categories of Tables 8-5 and 8-6, there l are 9 types of BWR contairment and 8 types of PWR contaiment. We sub-categories mentioned in the preceding paragraph add one further type of BWR contaiment and four further types of PWR contaiment. Wese basic variations f
are further ccmpunded by widely varying designs at the more detailed level, l as illustrated by the subsequent section on contaiment penetrations. g Earlier (Section 8.2) there is a discussion of the mitigation of a radioactive release by plant features outside the contairment. Were appears ' l l 1 ~O
O 8-11 to be no systsatic capilation on a plant-by-plant basis of the layouts and structural designs of relevant features such as shield buildings, auxiliary n v etc., nor of the characteristics of systes -- such as fire buildings, protection sprays and ventilating systens -- which might mitigate a release. Such a empilation has been undertaken for contaiment penetrations, as discussed below. O 8.3.2 Contairment Penetrations and PurgeWent Systems A typical light-water reactor containnent has 60-100 electrical pene-n U trations and 100-300 mechanical penetrations [SUBRNiANIAN, 1984]. We mechan-ical penetrations include equipnent and personnel hatches, fuel transfer tubes, drywell heads (for Mark I and Nrk II BWRs) , and a variety of piping penetrations. Argonne National taboratory (ANL) has empiled a catalog of i lO typical nechanical penetrations, based on detailed design information obtained fran a sanple of 22 plants. Each penetration surveyed has been given eight figures of merit, which are intended to indicate the likelihood that the pene-
}
tration will leak under severe accident conditions [ BUMP, ET AL.,1984) . O We ANL work is part of a broader NRC-sponsored effort under which various contracting laboratories are studying aspects of contaiment integ-rity. We four major elenents of this program are: the Contaiment Safety .O mrgins Progran, the Contaiment Penetrations Progran (which enempasses ANL's cataloging ef fort) , the Electrical Penetration Assmblies Program, and the Isolation Valves Progran [SUBRAMANIAN, 1984}. It seens that a emprehensive survey such as that conducted by ANL is not being undertaken in the case of O electrical penetrations or isolation valves. Figures 8-8 through 8-11 illustrate some typical contaiment penetra-tions. Figure 8-8 shows an equipnent hatch in a reinforced concrete EWR con-O tainnent. It will be seen that this hatch is of the " pressure-seating" type, in that increased contaiment pressure will force the hatch against its seal. Another feature to note about this hatch, however, is its relative weakness against " punching through" the containnent wall. Figure 8-9 shows an equip-
.O ment hatch fran a Mark I BWR. In this case, the hatch is of the " pressure-unseating" type, in that increased contalrnent pressure will tend to separate 'I O e b
,t
O s-12 the sealing faces. Another " pressure-unseating" penetration is shown in Figure 8-10, which shows a drywell head fra a Mark II EWR. It can be seen O that a large leakage area can be created by separation of the sealing faces of this penetration. Figure 8-11 shows a pipe penetration in a prestressed con-crete IWR contaiment. As this pipe (a feedwater pipe) is hot, it is therm-ally insulated fra the contaiment wall. .O Contaiments are equipped with purge and vent systens to control the camposition of the contaiment atmosphere. As these systens employ large-diameter (tens of inches) pipes, they form potentially significant leakage Iblatively .O paths. Figure 8-12 shows a typical valve used in these systens. minor leakage could occur if this valve were closed but did not seat correctly. Major leakage could occur if the valve failed to close when required. Such failure is referred to as an " isolation failure" of the
- O contalment.
Figure 8-13 shows how the various peneu ation.1 and their associated systens fonn a set of potential leakage paths frca the contaiment (or 'O directly fran the BCS) to surrounding buildings or to the enviroment. It is clear that a wide variety of such potential paths exists.8.9 ANL's cataloging effort makes it clear that penetration designs vary
- O greatly fran plant to plant. Indeed, nuclear plants in the United States vary greatly in many aspects of their design, even where the reactor itself is supplied by the sme vendor.
- O 8.3.3 The Record of contaiment Integrity Clearly, the ability of contaiments to withstand their design pressure, without more than the specified leakage, is relevant to their ability to withstand severe accident conditions. However, the record shows that contaiments often leak at above the specified rate.
Periodically, contalments are subjected to " integrated leak rate tests." .O Frmi the reports of these tests, and the reports of other inspections, one can estimate the frequency and magnitude of leaks. 'Ihe first nejor study of 'O
O. 8-13 4-this type was performed for American Nuclear Insurers (WEINSTEIN,1980] . Wis
,s study found that, during the period 1967-1979, BWR contaiments were
.V adequately leak-proof only 77 percent of the time, with a corresponding value of 96 percent for PWRs. Isakage rates ranged up to 900 times specified levels (in this case, 6 inch valves on a EWR contairr.#r.t were open for 2 months) . O A group at Pacific Northwest Laboratory is currently extending Weinstein's work, adding more recent data (PELTO & COUNTS, 1984}. Wese researchers intend to generate frequency distributions for leakage area for reference PWR and BWR plants. 8.4 Contaiment Loading During Accidents Durim a severe accident the containnent structure of a nuclear power 'O plant represents the last major engineered safety feature preventing the escape of radioactivity into the enviroment. Wis section describes the various possible threats to contaiment integrity. o 8.4.1 Static Pressurization Accidents can occur in which contaiment pressure rises relatively slowly (static pressurization) or very quickly (dynanic pressurization) . In the present context, " static pressurization" means that the characteristic time of the pressure pulse is greater than tis natural vibration periods of the contaiment structures. Sources of pressurization which fit this specifi-cation include: production of noncondensible gases fran clad oxidation and O concrete attack; steam production (excluding some steam explosions); hydrogen or carbon monoxide burnirg (excitx3ing detonations); and direct or combustion-induced heating of the contaiment abnosphere following high-pressure melt ejection (HPME) . @e excluded sources (scme stean explosions, and hydrogen detonations) lead to dynamic pressurization. test of these static pressurization sources are well recognized, and are addressed (to varying degrees of completeness) in source term estimations such O as those by BCL. However, HPME is a recently-recognized phenatenon which .is 'O 6 !
-m - - n---n - , - - - , , _ - - , - , , - - - - - , , , . - - - - - , - - - , - -
O 8-14 poorly understood and is not addressed in those estimations. Oapter Nine of this report provides a brief discussion of HR1E. O Hydrogen ccanbustion could present a severe threat to certain types of contaiment. Accordingly, Mark I and II BWFs normally operate with their contaiments filled with inert gas, while Mark III BWRs and ice condenser cont iment WRs are equipped with hydrogen igniters (which are meant to burn O off hydrogen before it reaches high concentrations) . As discussed in Gaptar Nine of this report, these measures reduce but do not eliminate the threat. g Analysis of contaiment deformations under static pressurization is not straightforward, as explained in a later section of this chapter, but it does avoid some of the difficulties of a dynmic structural analysis. 8.4.2 Dynamic Pressurization O Dynmic pressurization occurs when the duration of the pressure pulse is less than or ccxnparable to the periods of the contaiment's natural vibra-ti ns. Tw s urces of dynamic pressurization arise: certain types of steam O explosion, and hydrogen detonations (or, in sme cases, ccxnbined hydrogen and carbon monoxide detonations) . Wese phenomena have not been addressed in source tenn estimations such as those by BCL. However, they are potentially n significant, as explained in Gapter Nine of this report. v Experinents have shown that powerful steau explosions can occur in reactor cavities. W ese explosions could have dynamic effects in their
- g innediate vicinity, but the consequent overall pressurization of the contain-ment would proceed more slowly, probably in the static range.
Hydrogen (or hydrogetV' carbon monoxide) explosions can take the form of g detonations or quasi-detonations, in which shock waves spread throtgh the con-taiment atmosphere. tese waves can yield pressure pulses of several tens of atmospheres, with a pulse duration of several milliseconds. We effe ts fs h pulses on contaiment structures will depend on the O peak pressure and on the ratio of the pulse duration to the period of the O
O 8-15 i contaiment's most critical natural vibration. When that ratio is greater than one, the contaiment will behave as if statically pressurized. Ebr O ratios less than one, structural deformations (and stresses) will be less severe than for static pressurization with the see peak pressure. Because of j this additional emplication, analysis of the response of a dynanically pressurized contalment is even more difficult than analysis of a static O response. 8.4.3 High T e peratures O High tenperatures within a contaiment can pose a direct threat to items such as electrical equipnent, hatch seals, purge valve seats, and to equipnent which supports contaiment integrity.8*IO High tenperatures will also lead to themal expansion which may increase the strains already produced at the O contaiment boundary by pressure effects. In some instances, however, thermal expansion may be beneficial. As an exanple, for the drywell head shown in Figure 8-10, a higher tenperature inside than out will actually promote sealing of the flange. .O One potential source of high local tenperatures is a standing flame, occurring at a point where a strean of hydrogen-rich gas issues into a region where canbustion is possible. 'Ihis possibility is currently receiving atten-O tion in the sttx!y of hydrogen flanes above the suppression pool surface in BWR Mark III plants. In principle, increasing the tenperature of the containment boundary O could present a direct threat to contaiment integrity due to the decline of material strength at higher temperatures. Mark I BWR contaiments would be particularly vulnerable in this respect. However, in practice the contaiment would most likely have failed for other reasons before this point was reached. O 8.4.4 Ionizing Radiation Ionizing radiation can pose a direct threat to cmponents and systems in a manner analogous to high tenperature. Also, radioactive materials may O settle out from the contaiment atmosphere and accunulate to form local areas O
8-16 O of high concentrations which could cause high local tenperatures and intense radiation exposures to seals and electrical equipnent. %is possibility has largely escaped notice in the source term arena. 8.4.5 Missiles As nentioned earlier (Section 8.1) , the generation of massive internal O missiles has not been considered in the spectrun of DBAs. Morcover, this form . of contaiment loading has largely escaped attention in current source term efforts. Missiles could be generated by explosive failure of a major ccupo-nent of the RCS, or by rupture of the flywheel on a reactor coolant panp. A severe in-vessel stean explosion could generate a missile. Most prom-inently, such an explosion might lead to the head being separated from the reactor vessel and propelled upward with sufficient velocity for it to pene-O Considerable controversy surrounds this event, as trate the contairrnent. described in Gapter Nine of this report. However, given the present level of understanding, a contaiment breach in this manner cannot be ruled out, g 8.4.6 Molten Core Material In a severe accident, the fuel core would overheat and eventually begin to melt. In the past, molten core material has been seen as providing a direct containment threat in that it can melt through the bottcm of the pressure vessel and then through the basenat, leading to the release of radio-activity to the enviroment via atmospheric and liquid pathways. % is is popu-larly known as the "Gina syndrcme." However, molten core material may lead to a contaiment breach in other ways. Mark II BWR plants (see Figure 8-2) provide a good exanple. telten core material may melt through the drywell l floor, thereby weakening the support structures for the pressure vessel. I Collapse of the vessel and its associated piping may then lead to a direct O breach of the contaiment boundary.
- O O
O 8-17 b 8.5 Containment Response to Accident Loads O From the preceding sections of this chapter, it is clear that the mode of contaim.ent failure will be specific to each accident sequence at each plant.8.ll tereover, uncertainty about the physical processes of a particular accident sequence will be reflected in a corresponding uncertainty about the O mode of contaiment failure for that sequence. In turn, the mode of contain-ment failure will have important implications for the source term. Despite all these variations and uncertainties, there are same important common measures of containment response. For example, we must fully 'O understand the effects of any given level of static pressurization on a typical centaiment before claiming confidence in source term estimations. O mis section focuses primarily on the effects of static pressurization on the overall contaiment structure, on the penetrations, on the interaction between penetrations a xl the overall structure, and on isolation valves. We adoption of such foci does not mean that contalment loads such as missiles, O molten core material, or dynamic pressurization are insignificant. It is more a reflection of the focus of current source term investigations on contaiment failure by static pressurization. ,0 8.5.1 The Overall Structure At the internal pressures generated during severe accidents, contaiment structures will be subject to large strains which will lead to their O deformation and, possibly, to their failure. Figure 8-14 illustrates the calculated deformation of the Indian Point Unit 3 contaiment just before failure from overpressurization. W e calculations indicate that the dame apex wculd be displaced vertically by 6 inches and the mid-height region of the .o cylindrical wall would be displaced radially by 1 1/2 inches. Figure 8-14 also shows that sharp transitions in contaiment gemetry (in this case, at the wall-basenent intersection) will generate local areas of .O high stress. It is noteworthy that the standard contaiment configuration for West German reactors is spherical, thus reducing the potential for stress con-O
O 8-18 centrations. U.S. contalments , however, are all characterized by one or another type of sharp gemetrical transition (see Figures 8-1 through 8-5) . O- Contairrnent penetrations add further sites for stress concentrations. It is possible that the contaiment envelope will leak sufficiently that the overall structure will not be pressurized to its failure point. If the
.O contaiment failure point is reached, howver, failure is likely to be explosive [MURRAY, 1979). At present, there is virtually no basis for pre-dicting the size of breach arisire frcxn an exploaive failure.8.12
.O 8.5.2 Hatches and BhR Drywell Heads Equipnent and personnel hatches came in a wide variety of designs. Figures 8-8 and 8-9 show just two of many configurations. Figure 8-10 (which
,0 has been discussed earlier) shows a typical BWR drywell head. Due to the large size of these penetrations, their potential for opening leak paths to the atznosphere is considerable. .O Some hatches and drywell heads are of the " pressure-unseating" type (see Figures 8-9 and 8-10). With this type of penetration, high contaiment pressure will lead to leakage at the sealing flanges. For this kind of structure, analysis of flange separation is relatively straightforward.
- O Behavior of the seals is, however, difficult to predict, particularly if rad-iation effects are considered.
Where hatches -- as in Figure 8-8 - present a convex face to the
- O contaiment interior (i.e. they bulge inward into the contaiment) they are prone to buckling by " snapping-through," which can open up a substantial leak area at the flange or even lead to gross hatch failure.
- O Deformation of the contaiment's vertical wall may be non-uniform in the vertical and ciretzuferential directions, as illustrated by Figure 8-14. 'Ihis can lead to ovality in the flanges of hatches passing through that wall. In turn, that ovality may open up a leak area at the flange, due to relative rotation of the flange and hatch cover.
79 iO
4 8-19 0 8.5.3 Piping Penetrations O Figure 8-11 provides an exmple of a of pipe penetration. Wese are generally quite robust and likely to withstand acciden: loads. However, the fuel transfer tube may be relatively fragile, particularly at its expansion joints. O of more significance are the effects which arise from differential strains in pipes and contaiment walls. For exmple, accident loads could cau::c a pipe to produce a large bending mment at the penetration. An even m re armatic exmple is provided by Figure 8-15. mis figure shows the O effect of clamping heavy pipes (main-steam and feedeter pipes) inside and outside the contaiment. Large local defomations (and stresses) are intro-duced in the containment wall in the vicinity of the pipe penetrations. Con-t iment failure s sh wn in Figure 8-16 has been postulated for a contaiment O pressure of about 130 psig. 8.5.4 Electrical Penetrations O Wese penetrations will be at risk from pressure, tmperature, and radi-ation. An NRC-sponsored experimental program is under way, which will subject full-size penetrations to pressures and tmperatures typical of accident condi-ti ns, fter pri r simulati n f thermal nd radi ti n aging (SUBRAMANIAN, .O 1984). Until the program is completed, relatively little can be said about the behavior of electrical penetrations. Also, it should be noted that this test program will be cor. fined to penetrations manufactured by only 3 of the 18 n vendors who have supplied electrical penetrations for U.S. reactors -- the a reaining 15 vendors are no longer active in this business. Caution must therefore be exercised in applying the forthemirg test results to electrical penetrations in general. O 8.5.5 Isolation valves A variety of valves are represented in the contaiment envelope. Of greatest importance are the purge and vent valves, such as those shown in Figure 8-12. Preliminary estimates have been made of the leakage area for O
O 8-2e purge / vent valves under accident conditions [NBC, 1984]. According to these
~* estimates, at a contaiment pressure of 134 psig the purge and vent valves can O contribute up ,to 7e percent of the total leak area in a large dry PWR
, contaiment (Zion) . A particularly interesting, and currently unexplored, feature of valve O _ behavior is the effect of radioactive materials leakirg past the valve seals. Degradation of the seal might be significant under these conditions. 8.5.6 The Effect of Lcakage on Contaiment Integrity
- O As mentioned earlier (in Section 8.2) , contaiment leakage early in an I accident might prevent a major contaiment breach at a later time. Wis j might, or might not, lead to a reduced source term, depending on other factors.
-O Calculations have been performed to estimate the effect of leakage area on overall contaiment integrity. For a station blackout sequence8 .13 at the Zion EWR, Figures 8-17 and 8-18 show the estimated contaiment pressure (and IO t*perature) and leak rate for three leakage areas: " low", " medium", and "high". We assmed leak areas within each of these categories are shown in Table 8-8. W ese leak areas, which are essentially hypothetical at this stage of understanding, are based on various assmptions -- such as total failure of
'o the purge / vent valve seals in the "high leakage" case.
Based on these calculations, it appears that the "medim leakage" case may be the most benign. We "high leakage" case leads to a substantial con-lg taiment leakane rate early in the accident, while the " low leakage" case leads to eventual contaiment failure, perhaps in an explosive manner. l
, - 8.6 Status of Knowledge on Contaiment Response IO We previous section makes it clear that contaiment response to accident loads will be quite emplicated, even if loads such as dynamic pressurization, i molten core material, and missiles are neglected. % predict contaiment jO resp nse, we have recourse to theoretical models (which are incorporated into ; elaborate computer codes) and to the results of experiments. Wis section lO
8-21 p addresses the state of knowledge in these theoretical and experimental dcmains, and the degree to which predictions based on that knowledge are I applicable to actual reactor contaiments.8.14 9. 't I
. 8.6.1 Theoretical Models s
1 1 Because of the multitude of contaiment designs, the size and expense of G contaiments, and the variety and severity of accident sequences, theoretical methods must be used to predict contaiment performance under various accident corr 3itions. Unfortunately, despite considerable efforts, state-o f-the-art analysis is incapable of accurately predicting the behavior of contaiment .O . This holds true even if one neglects structures during severe accidents. dynamic pressurization and the behavior of small penetrations and isolation
- i valves.
O Tables 8-5 and 8-6 provide information on the gecmetrical form and construction materials used for the contalments of U.S. nuclear Inwer plants. Steel and concrete (both reinforced and prestressed) are used in a variety of O I Carmenting on the probles of modeling steel contaiments, researchers at Sandia National Iaboratories (SNL) concluded [BLEJWAS, LT AL.,1982): O If penetrations are included, a full finite-element model of the entire containment structure is required. With nonlinear behavior, such full analyses may be impractical because the time and cost beccrte prohibitive. .O As implied by this quotation, the models must be used for loading conditions in which the relation between stress and strain in a material is nonlinear.8.15 Such application of theoretical models is more deanding than I is their application for design purposes, where material properties will reain in or near the linear (elastic) range. It is particularly difficult to .O model the ultimate failure processes, as the SNL researchers have pointed out (BLEJWAS, ET AL. , 1982) : Although the post-yielding plastic behavior of steel is well ,O-understood, no ultimate f ailure mechanism for biaxial and triaxial stress has been accepted. Crack-propagation theory .O
s-22 O requires that the location of an initial crack be known; this location cannot be predicted with the existing programs. Werefore, predicting functional failure due to crack initiation, cra k gr wth, and resulting leakage will n t be attsupted at this O' time. Althotgh a nunerical instability may develop in the mathmatical model at scue level of loading, functional failure may occur prior to this level. We difficulties in predicting the behavior of concrete contaiments are O also severa. We sme SNL researchers concluded (BLEJWAS, ET AT.,1982): , te modeling of concrete contaiments in fine detail with high confidence is not possible at present and may not be practical in the near future. Modeling, using simplified approaches such as O layering of concrete and steel, is possible, but qualification of 4 existing or new codes is required before simplified procedures can be used with confidence. We probles of modeling concrete contaiments have been described by C other researchers as well. A recent review by Brookhaven National Laboratory (BNL) of analytic methods for modeling reinforced concrete containment behavior concluded, inter alia, [SHARMA, EP AL., 1983]:
. . . various modeling and analysis approaches, which have been O proposed and carried out for reinforced concrete structures, y
differ substantially in their predictive capabilities. Even for simple structural mmbers such as a biaxially loaded panel, the analysis results show wide variations. Wis simple example does not even involve many of the major nonlinear features occurring in large-scale cmplex structures. Fra the study it sees n 9 obvious that a systematic ef fort for developing benchmark y problens that will enable the verification of various analysis and ccrputer procedures currently used for safety evaluations of reinforced concrete structures would be very beneficial. O As shown in Tables 8-s and 8-6, many concrete contaiments (particu-larly large dry atznospheric PWPs) are of prestressed concrete construction rather than reinforced concrete construction. Analysis of prestressed contaiments encounters an additional difficulty, in that the prestressing O tendons can slide lengthwise -- subject to friction -- relative to the Wis frictional force may be difficult to predict -- for concrete.8.16 instance, it might vary with curvature. l O r s O
O 8-23 An illustration of the difficulties of analyzing concrete containment structures is provided by three different analyses of the Indian Point Unit 3 O contaiment.8.17 We difficulties were manifested even though penetrations were not modeled in these analyses. We first stMy, performed at the Massachusetts Institute of Technology (MIT), concitried that if all reinforcing bars and splices have the see properties, the contairnent would not fail at
- O 170 psig, despite radial wall displacaents exceeding 20 inches [FARDIS, EP AL., 1982]. Although the analysis was terminated at 170 psig, it was estimated that failure would occur at 200 psig, at about mid-height of the contaiment wall.
'O A study conducted at Ins Alamos National Iaboratory (Lt.NL) concluded that the contaiment would fail at 118 psig, but at the wall-base intersection [ BUTLER & FUGELSO, 1982). 'Ihis is a point of high stress concentration, as O mentioned above. A more recent BNL study [SHARMA, ET AL. , 1984] puts the failure pressure at 125 psig, with the same failure location as identified by the LANL group. However, the BNL group calculated a bassat uplift of 1.8 inches at 118 psig, compared with an LANL estimate of 4.4 inches. O Although the Ios Almos and Brookhaven models are in scme ways more sophisticated than the MIT model, both fail to incorporate the effect of material variability, which is included in the MIT study.8.18 When the MIT .O analysts considered the variability of strength of reinforcing bars and splices which might be encountered in practice, their estimate of failure pressure fell from 200 psig to 140 psig. Failure was again predicted at cylinder mid-height, due to failure of hoop reinforcing bars. It is important !O to note that the variability assumed in this analysis was within the
" acceptable" range - the possibility of substandard reinforcing bars or splices was not addressed.
O Various investigators have attempted to quantify the many uncertainties in the structural performance of contaiments. In a review of the state of understanding of contaiment performance, researchers at Ames Iaboratory, Io m , concluded [GREIMANN, FANOUS & BLUHM, 1984]: 'O 'O
~ ~ - - - - - - , -
-4 0 8-24 . . . the uncertainty assessments to-date must be considered preliminary and the results interpreted only as notional probabilities. We results can become meaningful only af ter d, these methods [that is, methods of analyzing the probability of containment failure under accident conditions, Bl.] have been calibrated with real-world failure experience.
8.6.2 The Experinenal Base for Theoretical Modeling O It is evident fra the preceding discussion that there is a need to develop and validate more sophisticated analytic models of contairnent perfor-mance. At present, there is a very mall base of experimental data to support
. such an effort. Model tests have been conducted at various scales, the largest until recently beirg a 1/10-scale prestressed concrete contairment pressurization test conducted in Poland. Wese earlier tests had a neber of deficiencies, but have provided some information. For exmple, the Polish . test ~O illustrated the importance of asynenetries in the contairnent structure (pene-trations, buttresses, etc) . Because the earlier tests were conducted at only one scale, however, they cannot be used to validate theoretical models for use with full-size contaiments [BLEJWAS, ET AL.,1982] . -Q A series of pressurization tests on steel contairunent models has recently been conducted at SNL, with the intent of overecmirs sme of the limitations of previous test programs. Four 1/32-scale tests and one 1/8-O scale test were conducted with the following outccmes: - Two 1/32-scale models with no stiffening rings or penetrations were tested. We first model failed explosively at 135 psig, g while the second experienced a small tear at cylinder mid-i height at 110 psig, following which that test was suspended.
Predicted failure pressure was 132 psig [HORSCHEL, 1984].
- One 1/32-scale model with stiffenity rings but no penetrations was tested. Af ter pressurizing to 120 psig and then depres-O surizing, two anall tears near the top of the cylinder were repaired, along with some areas where construction defects were found. When the test was resuned, explosive failure occurred between 135 and 140 psig. We predicted failure pressure was 138 psig [HORSCHEL,1984] . - ne 1/32-scale model with some penetrations was tested.
.O Snap-through buckling of the equipnent hatch door occurred between 40 and 50 psig (predicted pressure: 76 psig). A plate 'O
i 8-25
- was then welded over this hatch and the test was restrnal.
3 Explosive failure occurred at 120 psig. We predicted failure h pressure was 122 psig [HORSCHEL, 1984]. i - A 1/8-scale model with stiffenirg rings arx! sczne penetrations ) was tested in Novenber 1984. Althotgh a full report of the ) test is not available at the time of writing, it is known that
- the containnent failed explosively at 195 psig. Although g' sleeves around equipnent-hatch openings becate oval as pre-dicted by analyses, leakage did not occur at these openings (see our discussion of the " leak-before-break" question in Section 8.2) . It appears that failure started in a stiffener near an equipnent hatch [NRC, 1985]. Precedirg theoretical analyses had suggested that significant containment leakage
, would occur at the equipnent hatches at about 185 psig, and that the ultimate contairrnent capacity was 226 psig [CIAUSS, 1984].
i Wese tests showed reasonable agreenent with theoretical analyses. Considerable work will be required, however, to show that this agreenent could V be expected to hold for the more ccxnplicated and larger structures of actual ' steel containnents. Also, the quality of actual containnents may net match ' that of the test models. We tests denonstrated the impartance of small flaws in structural steel and in welds. O Testing of a 1/6-scale reinforced concrete contairrnent model is planned at SNL during 1986. It is unfortunate that only one test is planned, so that validation of theoretical models at more than one scale cannot occur. Also, O it seens that no test of a prestressed concrete contairrnent model is planned. i All of these ccxnpleted or planned tests involve static pressurization. Were is no ecznparable progran to subject containnent models to dynanic O pressurization. 8.6.3 Contairrtent Quality in the Field O clearly, the strength and performance of contairunent structures will depend heavily on the care taken in their construction. For example, the strength of a reinforced concrete contairunent depends on the integrity of long reinforcirg bars with multiple splices - these bars are only as strong as the O weakest splice. voids in concrete, which are particularly likely where con-centrations of reinforcing steel (and stress) are high, can substantially O
~O 8-26 weaken the contaiment. Similarly, steel structures are sensitive to the care taken in welding.
.O In addition, the gemetry of the actual contalment may not be exactly as specified. For exmple, out-of-roundness of the contaiment cylinder can occur, causing . local stress intensification and instability. Such asyrmetry U could arise during construction, or subsequently due to factors such as creep distortion caused by long-term insolation on one side of the structure (GITTUS, ET AL., 1982].
.O Experience with reinforced and prestressed concrete structures shows many problens with corrosion of steel reinforcing bars and tendons [GITTUS, ET ' AL., 1982]. Although this problen is recognized and guarded against for reactor contairments, it is impossible to guarantee totally that corrosion has !O not occurred. A recent incident with a prestressed concrete containment is instruc-tive. In January 1985, it was discovered that two tendon anchor heads had .10 fractured at Farley Unit 2 (a large dry atmospheric PWR) . Nmerous tendon wires were broken near the fractured anchor heads.8.19 , Ibis failure was detected about 8 years after the tendons were stressed, and it is speculated that the breakages occurred during a minor seisnic event in October 1984
- O [HUDGINS, 1985].
Containment defects such as these may not be readily detectable nor beccne evident durire leak tests (which are conducted at anbient tenperature
- O and 115% of design pressure) . Yet, they can becme very significant when the containnent is stressed well beyond its design limits.
8.7 Major Outstanding Questions O As this chapter has shown, there are many factors related to con-taiment performance which will play a critical role in determining how mtx:h radiation will be released durirg a severe accident. 'Ihese factors incitxle: !O the timing, location, and size of a breach; the occurrence of bypass events; the possibility of " leak-before-break" of the contairment; and the potential !O
, g 8-27 1
( for mitigating releases by plant features outside the contaiment. 'Ihe sen-
-)
g sitivity of source term estimates to these factors has not yet been exanined, AJ and it seens that a full sensitivity study of this kind is not planned by the N E 8.20
- t
) Before such a study could be cmprehensively conducted, answers would i,$ need to be found to a number of oustanding generic questions, incitxling the C fi following: 1 (i) Wat effect does material variability (within specified .j limits) have on analyses and tests of containment performance? (ii) How confidently can the results of theoretical analyses
- and model tests be applied to actual contaiments?
- l
; (iii) Wat relative confidence can we have in theoretical analyses of steel and concrete containments, Q
respectively? (iv) Wat will be the effects of dynamic pressurization? (v) Could BWR contaiments fail in such manner that the O suppression pool is bypassed or pool water is lost? i (vi) Could PWR ice condenser contaiments fail in such manner that the ice compartments are bypassed? (vii) Could isolation failure or leakage of inflatable hatch seals be precipitated or aggravated by accident O ., conditions? I (viii) Can we predict the structural effects of molten core
- material and missiles?
g (ix) Wat are the implications of hydrogen explosions in buildings adjacent to the contaiment? I (x) Can we predict the outcome of explosive contaiment failure? "D In some cases, these questions cannot be properly addressed because the nature of the containnent loading is not well understood. 'Ihis is true for HPME, hydrogen detonation, and missiles generated by steam explosions, phenmena which are discussed in 01 apter Nine. As understandirg of these O
- O P
c - - _ ~ , .w.- , ..
O 8-28 phenmena improves, their implications for contaiment response will becme clearer. O 8.8 Conclusions (i) We base of knowledge on contaiment performance has
'O widened in recent years, largely as a result of NRC-sponsored research.
(ii) Nevertheless, substantial improvenents are required in the analysis of contaiment perfonnance. At the m ment, there are major qualitative and quantitative uncertain-O ties in the analyses that have been empleted. (iii) A larger experimental base is required in order to support such improvanent. NRC's research program on relevant issues stretches for at least two years into the future. Even so, this progra as planned is not suffi- -O cient to answer all the important questions. (iv) BCL's current source term estimation relies on rather arbitrary asstrnptions about contaiment performance. O (v) Given the wide variation in plant designs, it is clear that generic conclusions cannot be drawn from even detailed sttx3y of so-called " representative" plants. In-depth study of each plant is required. (vi) There will always remain an uncertainty about the lO difference between " designed" and "as-built" contaiments. (vii) Were is little basis for relaxing regulatory require-ments on the grounds of improved knowledge about contaiment performance.
- O l
O
- O l
l
- O
O 8-29 8.9 End Notes for Chapter Eicht O 8.1 A more cmprehensive source is [OKRENT,1981] . 8.2 At high tm peratures, zirconium reacts vigorously and exothermically with water, leading to cladding abrittiment and the production of hydrogen. Either directly or throtgh embustion, that hydrogen can lead to large increases in contalment pressure and possibly to the loss of Q contaiment integrity. 8.3 For reviews of contaiment criteria and of siting and mergency planning regulations, see [BERNTHAL, 1984; SHOLLY & THOMPSON, 1983; STEVENSON, 1982).
/ /100, where P is the o 8.4 We formula for exclusion radius (in miles) was P reactor's thermal power level in kilowatts.
8.5 Each nuclear plant is supposed to be capable of being shut down, and having its long-term decay heat renoved, af ter enduring an SSE. We magnitude of the SSE, expressed in terms of ground acceleration, is O selected for each plant based on probabilistic argtnents. 8.6 For example, the Surry Plant operates with a contaiment pressure of 10 psia (GIESEKE, ET AL. ,1984] . 8.7 We estimated responses shown in Figures 8-6 and 8-7 are from an early O phase of the source term investigations conducted at BCL [GIESEKE, ET AL., 1983]. Subsequent analysis of the same accident assumed containnent failure by h%rogen burn at 4.5 hours, with a different developnent over time of the amount of airborne material [GIESEKE, ET AL., 1984]. However, in view of the various uncertainties as to actual behavior, Figures 8-6 and 8-7 provide a good qualitative illustration of g the importance of contaiment failure time. 8.8 We magnitude of the radioactive release will depend on the size of containnent breach and the amount of radioactive material suspended in the contaiment atmosphere. We amount of suspended material tends to decline with time -- thus a later containnent failure will tend to yield n a smaller source term. Wus, a high leakage rate early in the accident
- could lead to a larger release than a delayed containnent failure. For an illustration of this effect, see Figure 8-18.
8.9 Since Figure 8-13 is only a simplified illustration, it anits several potential leakage paths -for example, via stean generator tube ruptures r through the spent fuel transfer tube. Also, BWRs have different 'O types of leakage paths. 8.10 We phrase " equipment which supports containnent integrity" refers to mechanical and electrical equipment whose failure may lead to leakage at some point in the containnent boundary. For example, some equipnent and n personnel hatches are equipped with inflatable seals, which will be
" vulnerable in the event of damage to the seal-inflating systens. Also, containnent isolation might rely on the correct functioning of electrical or mechanical valve actuators inside the containnent. Were
.O
-_ .=. _
Y 8-36 f has been no study of the vulnerability of such supporting systens ar.d the implications of their failure, although NRC officials have intimated that such a study will be ccanissioned. U 8.11 As a further emplication, there are many potential accident sequences for each plant, rather than the handful which is often suggested.
, Sequences can exhibit different timing, or branch in various ways, according to factors such as variation in the initiating events (s) ,
operator actions, or intrinsic variability of physical processes. O Chapters Four, Five, and Six address this subject. 8.12 Explosive failure late in an accident could lead to a release of radioactive material through violent boiling (flashing) of contaminated water pools inside the contaiment, or through re-entraiment of material deposited on surfaces -- see Chapter Nine. y 8.13 A station blackout sequence at a PWR involves a total loss of AC power and of feedster to the stean generators. 8.14 Useful reviews of these matters are provided by (GIMUS, W AL.,1982; GRIENANN, FANOUS & BLUHM, 1984). .O 8.15 Contaiment behavior at high internal pressure can beccxne nonlinear in the gecznetrical sense, as well as in the sense of material behavior. Geometrical nonlinearities arise when deformations in one dimension have a significant effect on deformations (and hence on stresses) in other dimensions. In that event, two-dimensional or three-dimensional O analyses will be required. 8.16 In Canada's CANDU reactor containments, grouted tendons are used. However, most EWR contalments snploy ungrouted tendons (GIMUS, ET AL., 1982]. O 8.17 This containment is a large dry atznospheric PWR type, with a defonned-bar reinforced-concrete vertical cylinder, henispherical dczne, flat base, and steel liner. 8.18 It is noteworthy that the MIT study also analyzed the effects of hydrogen detonations. If the asstrned detonation initiated at the center
.O of the containnent base, the contaiment was shown to fail near the dome apex, whether or not the effects of variable materials were incitried.
Detonation at mid-height on the contaiment axis led to failure only if the variability of materials was included in the analysis. 8.19 The fractured anchors were at the bottczn of 2 of the contaiment's 130
- O vertical tendons, each of which is made up of 170 1/4-inch dianeter rods. Were are 357 tendons in the entire contaiment structure. We
~ contaiment is supposed to withstand its design pressure if 9 tendons (incitriing no more than 3 vertical tendons) are defective (HUDGINS, 1985). O 8.20 A limited source term sensitivity study using the BCL methodology -- but for only one accident sequence at each of three plants -- has been performed (LIPINSKI, W AL. , 1985]. We role of contaiment failure is addressed quite crtriely in this work.
O 8-31 8.10 References for Chapter Eight O AHEARNE, 1982 J.F. Ahearne, U.S. Nuclear Regulatory Commission, prepared speech published in, Proceedings of the Workshop on Contaiment Integrity, U.S. Nuclear Regulatory Conmission (mshington, D.C.) , NUREG/CP-0033, Vol. 1, 0:tober 1982, pp. 13-20. O BERNTHAL, 1984 F.M. Bernthal, U.S. Nuclear Regulatory Conmission, prepared speech to the Canadian Nuclear Society's International Conference on Contairment Design, Toronto, Ontario,17-20 June 1984. BLEJdAS, LT AL.,1982 O T.E. Blejwas, A.W. Dennis, R.L. Woodfin, and W.A. Von Riesemann, Background Study and Preliminary Plans for a Program on the Safety Margins of Contaiments, Sandia National Laboratories (Albuquerque, New Mexico) , NUREG/CR-2549, May 1982. BUMP, F?r AL. ,1984 O T.R. Bump, R.W. Seidensticker, M. A. Shackelford, V.K. Gambhir,. and G.L.
?ttennan, Characterization of Nuclear Reactor Contaiment Penetrations:
Preliminary Report, Argonne National Laboratory (Argonne, Illinois) , NUREG/CR-3855, June 1984. BUTLER & FUGELSO,1982 O T.A. Butler and L.E. Fugelso, Response of the Zion and Indian Point Contalment Buildings to Severe Accident Pressures, Ios Alamos National Laboratory (Los Alamos, New Mexico) , NUREG/CR-2569, May 1982. CARLSON, ET AL., 1981 D.D. Carlson, W.R. Cramond, J.W. Hicknan, S.V. Asselin, and P. Cybulskis, O Reactor Safety Study Methodology Applications Program: Sequoyah #1 PhR Power Plant, Sandia National Laboratories ( Albuquerque, New Mexico) , NUREG/CR-1659, Vol.1 (SAND 80-1897/1 of 4) , April 1981. CIAUSS, 1984 D.B. Clauss, " Analysis of a 1/8 Scale Steel Containment tt> del Subject to .O Internal Static Pressurization", in Proceedings of the Second Workshop on Contaiment Integrity, U.S. Nuclear Regulatory Conmission (Washington, D.C.) , NUREG/CP-0056, August 1984, pp. 635-657. COFFIN & SCHLESINGER,1984 B. Coffin and R. Schlesinger, Nuclear Power Plants in the United States: Current Status and Statistical History: Status as of 1 Auqust 1984, Union O of Concerned Scientists (Cambridge, tessachusetts) , August 1984. FARDIS, LT AL., 1982 M.N. Fardis, A. Nacar, and M.A. Delichatsios, Reinforced Concrete Containment Safety Under Hydrogen Explosion Loading, Massachusetts O Institute of Technology (Cambridge, Massachusetts) , NUREG/CR-2898, Septanber 1982. 'O
'O 8-32 GIESEKE, W AL., 1983 J.A. Gieseke, P. Cybulskis, R.S. Denning, M.R. Kuhlman, and K.W. Ime, Radionuclide Release Under Specific LWR Accident Conditions, Ihttelle
,d Colunbus Laboratories, (Colunbus, Ohio), BMI-2104, Vol. I, draft, July 1983.
GIESEKE, W AL., 1984 J.A. Gieseke, P. Cybulskis, R.S. Denning, M.R. Kuhlman, K.W. Lee, and H. Chen, Radionuclide Release Under Specific LWR Accident Conditions, _O, Battelle Colunbus Laboratories, (Colunbus, Ohio) , BMI-2104, Vols. II-VI, draft, July 1984. GI'ITUS, LT AL. ,1982 J.H. Gittus, et al., PWR Degraded Core Analysis, Springfields Nuclear Power Developtent Labor.atories, UK Atanic Diergy Authority, ND-R-610 (S), U April 1982. GOU & LOVE, 1982 P.F. Gou and J.E. Inve, " Pressure Carrying Capability of the Containment Structural Systen of the Mark III Standard Plant," in Proceedings of the Workshop on contalment Integrity, U.S. Nuclear Regulatory Cmmission C (Washington, D.C.) , NUREG/CP-0033, Vol. 2, October 1982, pp. 263-317. GRID 4 ANN, ET AL. ,1984 L. Greimann, F. Fanous, and D. Bluhn, Final Report, Contalment Analysis Techniques: A State-of-the-Art Sutmary, Anes Laboratory ( Anes , Iowa) , NUREG/CR-3653, mrch 1984. v HASKIN, LT AL. ,1984 E.F. Haskin, V.L. Behr, and L.N. Snith, " Combustion-Induced Icads in Large-Dry IVR Contaiments," Sandia National Laboratories, in Proceedings of the Second Workshop on Contaiment Integrity, U.S. Nuclear Regulatory Catmission (Washington, D.C.) , NUREG/CP-0056, August 1984, pp.129-143. J HOFMAYER, ET AL., 1984 C.H. Hofmayer, G. Bagchi, and V.S. Noonan, " Containment taakage During Severe Accident Conditions," in Proceedinos of the Second Workshop on Contalment Integrity, U.S. Nuclear Regulatory Cmmission (m shington, j D.C.) , NUREG/CP-0056, August 1984, pp. 305-317. O HORS TEL, 1984 D.S. Horschel, " Experimental and Analytic Results of Steel Containment Tests", Sandia National Laboratories (Albuquerque, New Mexico) , in Proceedings of the Second Workshop on Contaiment Inteority, U.S. Nuclear Regulatory Cmmission (Washington, D.C.) , NUREG/CP-0056, August 1984, pp. U 621-634. HUDGINS, 1985 C. Hudgins, "Farley-2 Tendon Anchor Head Failures Seen as Possible Generic Problan," Nucleonics Week,14 February 1985, pp. 5-6. O O
s, 8-33 , LAKSHIMPATHIAH, 1983 K. Lakshmipathiah, "Contaiment Failure tedes Analysis", Stone & Webster O Engineering Corporation, Appendix 4-F of Millstone Unit 3 Probabilistic Safety Study, Northeast Utilities Service Omnpany, August 1983. , LIPINSKI, LT AL.,1985 , R.J. Lipinski, D.R. Bradley, J.E. Brockmann, J.M. Griesmeyer, C.D. [eigh, K.K. Murata, D.A. Powers, J.B. Rivard, A.R. Taig, J. Tills, and D.C. g Williams, Uncertainty in Radionuclide Release under Specific LWR Accident Conditions, Sandia National Laboratories (Albuquerque, New Mexico) , SAND 84-0410, draft, February 1985. MURRAY, 1979 D.W. Murray, "A Review of Explosive Characteristics of Prestressed g' Secondary Contaiments Near the Ultimate toad ," Nuclear Engineering and Design, Vol. 52, pp. 157-164, 1979. NERO, 1979 A.V. Nero, Nuclear Reactors, University of California Press (Berkeley, California) , 1979. O NRC, 1981 U.S. Nuclear Regulatory Cmmission, Technical Bases for Estimating Fission Product Behavior During LKR Accidents, U.S. Nuclear Regulatory Conmission (Wshington, D.C.) , NUREG-0772, June 1981. n NRC, 1984 Contairment Performance Working Group, Contaiment Leak Rate Estimates, U.S. Nuclear Regulatory Cmmission (Washington, D.C.) , NUREG-1037, draft, 4 April 1984. NRC, 1985 U.S. Nuclear Regulatory Cmmission, "Wst Indicates . Significant Design 'O Margin in Large Steel Containment Buidings," Office of Public Affairs, U.S. Nuclear Regulatory Cmmission (mshington, D.C.) , press release, 14 January 1985. OKRENT, 1981 , n D. Okrent, Nuclear Reactor Safety: On the History of the Regulatory Process, University of Wisconsin Press (Madison, Wisconsin) ,1981. PELM & COUNTS,1984 P.J. Pelto, and C. A. Counts, " Reliability Analysis of Containment Isolation Systens," Pacific Northwest Laboratory (Richland, mshington) , in Proceedings of the Second Workshop on Containment Integrity, U.S. O Nuclear Regulatory Conmission (Washington, D.C.) , NUREG/CP-0056, August 1984, pp. 31-46. PHIIADELPHIA ELECTRIC,1982 Philadelphia Electric Company, Probabilstic Risk Assessnent: Linerick cenerating Station, Philadelphia,' Electric Company (Philadelphia, 'O Pennsylvania) , Rev. 5, Septaber 1982. O
O 8-34 SHARMA, ET AL. ,1983 S.M. Sharma, M. Reich, and T.Y. Chang, Review of Current Analysis g' Methodology for Reinforced Concrete Structural Evaluations, Brookhaven National Laboratory (Upton, New York) , NUREG/CR-3284, April 1983. SHARMAN, ET AL.,1984 S. Sharma, Y.K. Wang, and M. Reich, " Nonlinear Failure Analysis of a Reinforced Concrete Contalment Under Internal Pressure," Brookhaven g* National Laboratory (Upton, New York) , in Proceedings of the Second Workshop on Contalment Integrity, U.S. Nuclear Regulatory Catmission (Washington, D.C.) , NUREG/CP-0056, August 1984, pp. 519-532. SHOLLY & THWPSON,1983 S. Sholly and G. 'Ihmpson, "Buergency Planning in the UK and the US: A o Canparison", Annex 0 of Safety and Waste Managment Implicataions of the Sizewell PWR: A Report for the town and Country Planning Association and Local Authorities, Town and Country Planning Association (London, England) , Novenber 1983. STEVENSON, 1982 J. . Stevenson, " Current Status of Containment Develognent", in 'O Proceedings of the Workshop on Containment Integrity, U.S. Nuclear Regulatory Cmmission (Washington, D.C.) , NUREG/CP-0033, Vol. 2, October 1982, pp. 9-34. SUBRAMANIAN, 1984 C.V. Subramanian, " Integrity of Containnent Penetrations Under Severe 3-)
- Accident Conditions," U. S. Nuclear Regulatory Catmission, in Proceedings of the Second Workshop on Contaiment Integrity, U.S. Nuclear Pegulatory Cmmission (Washington, D.C.) , NUREG/CP-0056, August 1984, pp. 401-412.
WEINSTEIN, 1980 M.B. Weinstein, " Primary Containment Isakage Integrity: Availability and M- Review of Failure Experience," Nuclear Safety, Vol. 21, No. 5, i Septanber-October 1980, pp. 618-632. C O 4
- O
- O
[ t Q'
^I 8-35 ',, -1 i sy 3
- TABLE 8-1
- s. -
O i water Inventories and core Masses - Typical PWR and BWR Plants i (kg/MWt)
. . n ., }4f ,
Zion (pwR) crana culf (SWR)
- O- ,
Beactor Cbolant Systen 2 ter Inventory 76.1 85.5
,4 ,
UO2 in Core 30.3 > 43.4 T ZircaloyibCbre 6.2- 20.7
'i 7(- - ..\ ; t. + . .t ?- , ~, ;
Notes r I (a) Zion is a 3238 MWt Westinghouse PhP. ,. (b) Grand Gulf is a 3833 MWt Mark III General Electric BWR. (c) Data are from: Gieseke et al, 1984. O
-)
O
\
s s
- O i
t t
- O
} i d [k, r.O l,,
' i. . ._ . . . - - , _ . _ . - . . _ - . . . - . . - _ . - - _ _ . , - , - . . . - _ . . . _ _ , _ , _ . - . - , -
D. 8-36 TABLE 8-- 2 U.S. Inventory of BWR Contairments (Status as of 1 August 1984) Pre-Mark Mark I Mark II Mark III q v Licensed to operate 3 22 5 1 Lbdcr 0- construction 0 2 4 9 Total 3 24 9 10 O (Sources: Blejwas et al, 1982; Coffin and Schlesinger,1984) O O
'O O-O
^O:
I 8-37 TABLE 8-3 n U. S. Inventory of PWR Contairunents M (Status as of 1 August 1984) Ice Condenser. Large Dry Large i Subatmospheric D_ry r Atcmospheric O. Licensed to operate 6 5' 43
.Under construction 4 2 25 teal le 7 68 iO (Sources: Blejwas et al, 1982; Coffin and Schlesinger,1984) ;O 10
!O i ~O i 10 i .10 lO
l
- O !
s-38 l TABLE 8-4 Contaiment Volmes and Design y Pressures: Sone Typical U.S. Reactors Ratio of Free (Gas) Design Pressure Volme tQ Reactor (psig) Power (m'/MWt) Peach Bottom (BWR Mark I) O drywell: 1.4 56 wetwell: 1.0 56 Limerick (BWR Mark II) drywell: 2.1 55 wetwell: 1.3 55 .O 4 Grand Gulf (BWR Mark III) drywell: 2.0 30 contaiment: 10.5 15 Sequoyah (Ice condenser PWR) ^O 7.1 11 upper empartment: lowr ccmpartment 3.1 11 Surry (Large Dry Subatmospheric PWR) contaiment: 20.9 45 O zion (Large Dry Atmospheric IMR) contaiment: 23.6 47 4 Sources g
. Data for Limerick: Philadelphia Electric Company,1982 Design pressures for Surry, Zion: Haskin et al, 1984 ^11 'h*' d*'** i* *** ** *1' 1984'
- O lO
'O
.O
- g 8-39 TABLE 8-5 U.S. BWR Contaiments: Inventory V by Type of Construction (Status as of 1 August 1984) 2 Licensed Under Total To Operate Construction O~ (i) Steel contaiment.with 1 .0 1 henispherical dome, vertical-cylinder body, and ellipsoidal base.
(ii) Steel sphere. 2 0 2- "O Mark I (i) Steel contaiment of 20 2 2 light-bulb / torus configuration. ]O (ii) Deformed-bar reinforced-concrete 2 0 2 contalment with steel liner both in drywell and in toroidal suppression chanber.
- O Mark II (i) Deformed-bar reinforced-concrete body, 2 4 6 steel closure cap, flat base, steel liner.
(ii) Prestressed concrete body, steel 2 0 2
- .0 closure cap, flat base, steel liner.
(iii) Steel contaiment, ellipsoidal base. 1 0 1 Mark III .O (i) Deformed-bar reinforced-concrete 1 3 4 vertical cylinder, henispherical dame, flat base, steel liner. (ii) Free-standing steel cylinder with 0 6 6
- O. shallow dctne, deformed-bar reinforced-concrete base with steel liner.
(Sources: Blejwas et al, 1982; (bffin and Schlesinger,1984) !O e a f 10
O :- 8-40 L
~ TABLE 8-6 O' d.S. PWR' Containments: Inventory by Type of Construction L (Status as of 1 August 1984)
Licensed Under O- To Operate' Construction Total'
' Ice-Condenser (i) Deformed-bar: reinforced-concrete 2 0 2 vertical cylinder, hemispherical 0 -dame, flat base, steel liner.
(ii) Free-standing steel-cylinder with' 4 4 8 meispherical dome, defonned-bar
. reinforced-concrete base with steel liner.
O. . . Large Dry Subatmospheric Deformed-bar reinforced-concrete 5 2- 7 vertical cylinder, hemispherical dame, flat base, steel liner. Large Dry Atmospheric (i) Deformed-bar reinforced-concrete 7 8 15 vertical cylinder, hemispherical dome, flat base, steel liner. (ii) Vertical' concrete cylinder with pre- 2 0 2 stressed vertical reinforcement and deformed-bar hoop reinforcement. Deformed-bar reinforced-concrete hemi-spherical dome, flat base, steel liner. (iii) Prestressed concrete vertical cylinder 26 15 41 and dome, deformed-bar reinforced-concrete flat base, steel liner. l (iv) Steel sphere 2 0 2
- O
- (v) Steel containment with hemispherical 6 2 8 deme, vertical-cylinder body, and ellipsodial base.
O (Sources: Blejwas, et al., 1982; Coffin and Schlesinger, 1984.) 1 i f !O
TABLE 8-7 Typical Design Pressures (psig) of U.S. Contairments Concrete Steel Prestressed Deforned-Bar Other Light Bulb Sphere Vertical Steel Done Cylinder, & Body, Vertical Vertical -& Torus j Cylinder & Cylinder & liemispher- Reinforced i Done, Flat tiernispherical, ical Dane, Concrete Base Done, . Plat ETITpsoidal Base Base Base 1 i BWRs i A Pre-Mark - - - - 27-30 - - 7 Mark I - - 56 56-62 - - - p E Mark II - - 45-56 - - - - 15 - - - - 15 Mark III - i PWRs 12 - - 11-15 Ice Cbndenser - - Large Dry - j Subatmospheric - 45 - - - -
- Large Dry Atmospheric 47-60 42-55 42-60 -
25-46 34-44 - (Source: Blejwas et al,1982) l l 1 1 i
O 8-42 TABLE 8-8 Hypothetical variation of Leak Area with Pressure, .Q Large Dry Atmospheric PWR Contaiment 2 Contaiment Pressure Leak Area (in ) (psig) Low Mediun High Estimate Estimate Estimate O Nonnal Operating 0.1 0.5 1.0 23 0.1 0.6 1.5 47 0.1 0.6 1.8 105 0.1 2.1 11.0 134 0.1 5.3 23.7 O Notes and Sources
;O (a) 'these estimates are fran: NRC, 1984.
(b) Estimates are based on stu3ies of the Zion plant. (c) Leak areas contributing to these estimates arise at:
- equipnent and personnel locks O
- purge and vent valves
- electrical penetrations iO
- O i
O lO l O
> 8-43 ~
FICURE 8-1 .O
',..- t O)
O u y f M J
- r r F
I O q l l J REACTOR
\ ) . k7 '
BUILDING f _ RY WELL O f) d C 1 l Lk ip .,
/ T
- O -
,, WET WELL L. l 1 - . _ l # SUPPRESSION POOL J b b L
.O O Containment for a Mark I BWR (Source: NRC, 1981) 'O O
t 8-44 FICUKK 8-2 i
. G RPV MOVABLE l SHIELDIN6 PLUss , 2*l ELLIPTICAL HEAD 9 '
I -REMOVABLE l EL. 343'- G AG" . .'
- . . f % BOLTED COVER t !
f3Y.lo* 1.D. '- /
^ I EL.32 5'- 8" j 5. -
f- - nh O g .- ); 36'-4* l .D. 3I ' TRUCTURAL
'.. SEPARA. TION ~
- .'/
EL.312' ,B" ,, g : ','.
} .1l. - - ^ -5 TEEL LINER PLATE ANCHORED TO O I CONCRETE G : 2 ' .#
24 '- 7' 3 . .. . . ., p:.. PLATF. RA-
- j ,
O
'
- EGUIPMENT HATCH E L . 2 G 5'- O* t :.'f j
f-v' h /W/ PERSONNEL
*/ .: '$
g= LOCK IF HATCH DOOR EQUIPMENT k PLATF. \ .. . ,q. . HATCH -- .5.. . . ( 7 zst3El.o. - c . 5 +- 2 lg EL.237'-Il'
/ 1 -e i i __._._
JET DEFLECTOR A t
'- PLATES / ls __
IZ-3'- G'pf s , i6 ' DOWNCOMER 'O COLUMNS EN- "
. . 4' d. zo'-r 3"_ --' 7 a
VENT PlPES
*4 '
t f.
#00L WATER . + + ~-' ' ~
T ,, LEVEL l 5 ' ' 7NVENT PIPE
- 87 ' HIi' D ', ~ " <. cf ' ~
5 RESTRAINT SYSTEM O FL ISI'-II' / T '
~ ~
Y' ' -
~~ - ..- .:. . -l ' ,
h Le-o* - SECONDARY REACTOR BLD CONTAINMENT WALL iO WALL (SECONDARY CONTAINMENT) !O Containment for a Mark II BWR (Source: Philadelphia Electric Co., 1983) i lO
8-45 j i FIGURE 8-3 l l 50 SHIELD EUILDING Q A DR WELL EllNG g 9 p /sLAs O f 3' /! g y, ,
--r -- / g *f i
_ STEEL f g1.;~EN1 g[.\. O . g p a- .w&' ~a+w ay
?+ @'@ --
9.1-b ~'*m 0 '; ' ' 1 DRY L vi ( ) I h
- O LL xgx r. h' .I =
(,NN M E-
- d g b
f ,.. I pjg "k'pi. i d e m <. .
, om N Q.. ie - . . . . , , , _ $r p ) {a]- '9M 9 2
n .O f r ' .
, M -
ce y.. [ l J' U-d U- ' 2
]gSUPPRESSION sumE.smou POOL 9g _
l
- =- yg POOL WALL l l l ,
~ ~ " '**
!O \
-. na >4 g.: .10; -' ' ' .6' - "x
- D'
_ _ , , . ; m .% .
- : m L_ s 'np- ' o-
- o. e e- t-y l
I ', ,(.S' .- CONTAINMENT
')[g',h.':*3,. ,r p, , g 3'j p ,g *g ANCHORAGE O
l i O Containment for a Mark III BWR (Source: Gou and Love, 1982) O
8-46 a FICURE 8-4 A A V CONCRETE SHIELD , BUILDING , O a POL AR CRANE
- gr ,
g
.O TOP oEcR PA~Et .p 6, ,0 .. )
o V __ s' STEEL r ~
~
g C c~ ~ SHELL *( CRANE WALL
- Q 7 ~'.
7 o
,C~
7 T
~
p, d
~[ ~
p .
~1 'D d f' [ ~ ~ /
r ICE CON DE NSE R STE AM 2 o 2,
. GENERATOR e 3- en '
C C ~~~y CONTROL ROD
.)
g
~~~~
(
*C i * '
60 f ORIVE MIS $4LE M'
) SHlf LD $' c ., , ) *
- 0/ ---3 REACTOR COOLANT r- I
-~~~ *i g
PUMP L-Q }d a 0
- }1
,J. --- . 3 ---
(
. /P. L O . " " .1 -- ,
e c .O , t R$
.- 0 & P "r
[ {
- I n ! _
9 __._ 0
*( .
Q h ei g i,_ e 3
.( '.
l 7 f\ '*,
- c 3,cJ
.- a . . .'}l. ; 8 a t i ,/ 5 ",c . ,3 7 2 C i r.
1 i
)
J. :'
~ ~
O .O a ---~ %'
;'.' 4O' v
g,' ;c .< 3
>c e 0- 3 3
hI o " _b - D o ,'J c. .0e, -
'*' .- Y 6 :ta.m ; ogg,,6 ',0
_%y.o~;.n
.h . e.o. . f;gQ T o . 'oyg
^ _m 6n :. ew. .0c d\ L
- O
' ~5a .
l
.Qt' VENTILATION FAN AND ACCUMULATOR ) { j- )
7 .'o EQUIPMENT l.D,C . ,, .. Oc **P< elevation 3.*.
@ '_ npD . A n* h n b O
Containment for an Ice Condenser PWR O (Source: Nero, 1979)
- O
i 8-47 h-F .?I C U R.. . ..E, * .8-5 t ........**,r..*.. s m v
;/ FILTE RS SPRAYS . .~ 9 O :: .t . . . .c, .
9 2
'.. F- , ' h/\M ..
o ,.
~~ 7 _
Q.
. . .i * - t J . ..., .' e r m * 'e' . ... , i e ,., c >
- q O
. ./.-C. *s r , , .
l , .
** l - * ...s I;'
W [ ( j . r- .
,.s. + ;
r n
~-n . :.
O ,; ms. . 1 l ... r... g .. o :_,- --
- e .:: , .) :. ., , y i - ,-l . l. . . . .
.o e. -f=9 ... .
- g,. , .w. .
y G:p ,,
,:. ,- .:r
- s .. .
.' + * . R' E ACTO 4.".ll*r.] . < ,/r l _ ', */ *..t.' .
R *'*.". 4 'Jr.*. ..:.*
- l
.J VESSEL s * *ll- 'l l '. ..; * .<* ./ ;;f J..
J. t .'
....._'l O , .l,. ...:. *.,.......... .*4..* . *;.
o % . . . . . i...*** .... . . .... . ..
..s..,=. * . . . . . . _ .s- . ; : , ...,, s. ,.I ... g ... . . .*. . . . .. . ~*. . - . ':. lls . t; . ; . : .'. . . i's
- . .'.*;;.' '::*l.': *;***,. . . ?.: ...
. . , . :. , ,l r.'..r...t :' .
- 1 4.-
, ,:. ::. '.',}'.p ' . ' -l]
Q ', .: .. * ' Law on containunt for a na
- O (Source: NRC, 1981)
.O
~'
6
~
O O O O ~ '6 O b O O' -O 120.0 - 0.8 0.7 i 100.0- n j ,a e hydrogen g
; burn -
0.6 5 a .I w 80.0- 3 m
- containment -
0.5 5 h failure i
, 0 a:
3 m
" 0.4 g 60.0-i $ E H
IE .
= -
0.3 , a ]
;5 40.0- y c z w a 8 IN 0.2 , L
(
- t I 20.0-i ,
0.1 j * ( i 0.0 5d0.0 1000.0 1500.0 2dOO.0 2500.0 3d00.0 3500.0 4000.0 j TIME - (MINUTE) I i Estimated Containment Pressure for an AB Accident Sequence at the Surry PWR l ' l (Source: Gieseke et al. 1983) I i.
e- 8-49
> FICURE 8-7 O-10 AB-8 Hot
- O -
q [ - - - Solid 8 Water [l Solid I g llI Q l n If i 0 iO S ll \ I i g a -
\l \ ; _ yI 3 \ ! \
.o e i \
$ - \ \
Q \ \ assumed .O I0 5 ( 7 e railure time 4 - N hydrogen
-
- burn k 0 !
- l V I f # I I I I I I I I I I I I f I 10
.O l 2 3 4 io 10 10 10 Time, min .O Estimated Mass of Airborne Material in the Containment Atmosphere O f or the Accident Sequence Described in Figure 5.6 (Source: Gieseke et al, 1983) O
. 8-50 FICURE 8-8 (Part 1 of 2) s i / -- N ' ..k;, . -
_n.. s.... . M.. I ' *\ . ~ . ' ' [ . .- I n+. .,- .; ,13 . . . s,a.2
. . ,I J
4<- } Q. Il 4,,4 0 - J - { l
? I j' - -.<:11!L-G' 4[- ! !. I ll -, ! 4 !' p J O l !!
3 i I ,
! l lI i. .a' - .i _ _ 0 ' l ~ , t i t I, -
ii . m i "i \(o . O , ,S o. ,O i i , I - a 4
. / :l bp g. R .o - i 0 l \p t ll covt e. ;. tl 'o. i; lI ~a c I g , 'l . ,, .I y ., }
i ....# ~ . -'
=
y 6 .' ' ' ,1 ' 4%a. w& -
,e)4 ::- w, ov-o,- - -jL'(;; n .
af _'J. ,.**. ,
. p w.ecN GE E. .~ -- 4'.6 -. : .
e l
.y 9 .A m,
N ./ _ f W ,_,. m Equipment Hatch in a PWR Containment U (Source: Dump et al, 1984)
/ %.
I 8-51 h~
- FICURE 8-8 1
i (Part 2 of 2) O i 1 1 7 8 3*uuEw s I
- yj -
.. - m eiu r. We
[ it OtA. bWIMG BCL Tb 2.I9 ** /, 8 ha O l (tro EEG) N-)- 4.eiuGS f l *,'."a & 'oF tSo bTUD6 8 4 , e h .6_.G /$ Wi= O }
-hfp)y _ 'l i . 4 j '/<, . B \\\ \ \\\\7. -
b
; - 3" 6" -ik. ! -i.r -I 63 - i 63'-=* /
l(o Io. OtA. O is N .. gd.O 6m. uA.o.
- O ETA \ \
O PzF_6bucE rusto v E F tic.A L k*k i Guu C E.o P .b E At . I- < m s s l . .. O [/ . l .
"r.,t\s . .t y ,, y; s l ., 4 s I . ll. %4jO. --- ,--. ,s w'4 k,o. ~ g%-
O ' 6EAL DETAIL
- O Equipment flatch in a PWR Containment
,g (Source: Bump et al, 1984)
O 1
8-52 p FIGURE 8 -9 I i I
', < to'-toy; & < 12' cd \ -
A
.I )
t s-r i.e. it'd A
.O 8* 6[i.o.
l m- I
- ml-e \7'-o t. O.
p: ;p; 3 PEDSUR- ,w. A f
' 1 l
[ k w! ef J St C
\ h g
comu nutwv svAut i L- t. uTEt:.t O C'. E.4 r E R.t o R o oo sz. __ o coc'_ r~. v
/~
Equipment IIntch (with Personnel Airlock) in a Mark I B'#R Containment O (source: Bump et al, 1984) b
'v!
uu
~ '
FICURE 8 - 10 1 I
' SECTION @
CONTAINMENT p REMOVABLE g HEAD SHELL SECTION @ V
/ it BOLTS O 53 (TT' / k -. SECTION @ / /
.O / ll 2 ) EL.328L 4' SECTION @ , 3 g ,
\ \
,O I SECTION @ e e- \ , O ' N i O \/ SECTION @
\ .
O CONTAINMENT WALL 'O Drywell Head, Mark II BWR Containment O sSource: Philadelphia Electric Co., 1983) 'O
1 0 8-54 FICURE 8 - 11 y es 1 1,- L O = i .. 13 " 6- 6= 55TG
= ^
CONCRETC W At.L g , ANCHOP. *T E
~ *5*b' STf2A PS 9 PLCS, + ~ *-
2.k- >-3 *
-PRESSURE n CONN. .
_ ij .. e '; . Ib SLEEVE
#' 1 k, / / / , , i , / ., , - /T:8MZD > i . . n L bY 5 f e
I -T is4- e3 ' ' ' g g' * ~Tf 3 2.~o.0. G'A90 U ins 0 Lelou l o O D- -iNNEe-~ ANCaet FRG SLTEVE TuSE~ e t [ c o0Liwca c citt
~~
j l %* 9 Q~ <//r si //i: //ir e. <>.p'issi/A , )f d . _., _ i m i
. N %
ONVV i p r
~. .r..... ,
7 b M A s eJ ANCHOR
~- . le
- 1 'l_WELO RIMGJ i l oy F O RGiN G.
4 ) . .. p '. 3 OUTER.
- o & - we etNE?_
U RAo. 2- , FL'JE O HEAD r-- 37i~ ,- 4-Oh .spnCce. S'.00 K S cxp. JOLNT 7 CONVOL, r 3 PLCS. \ P r.ES S U RE 'C.OVt.R PL A iE J 3 llo t Pipe Denetration in a D'SR Containment 3 (Source: Bump et al, 1984)
'% /
l
-g' , 8-55 FICURE 8 - 12 p, SHAFT SEALS 9 q pBEARINGS ( @$ ad a 0 0 0o o SEAT 0 0 b 0- o g( , NkNNV g " < (/ o\0 o
) o , / ciSc f
O /)
- s ,/ otSc 3
Ng N
/3 '
j $5bE?"* o BOY s 4 RETAINING RING
] '8# / SEA G fh G e N N - < i= -== \wlh x sci}n u o IO
) Typical Butterfly Valv_e Used in Containment Purge and Vent System
- O (Source
- Hofmayer, Dagchi, add Noonan, 1984)
,0
8-56 n F1GURE 8 - 13 i- .] Construction Vent 8" Vent g MOV Containment Soray CSIS Header T CS 101 A anjection System VS 02 MOV 8" VS100A ( CSIS Header g CS 13 1018 - M 00A g 36" L 2 m Purge p.1 A r1 7m' Supply , VS1006 CS 1009
-- Pe t on CS24 G ^
'V 101D1 '890C#'IY P 18 V Low Prenure Pen t on Id VS 1000
' ^ ._ 32 as- m92 _ . Pu,,e Hf,P 1890A 7' F ' ' - E nhaust T m 19648 0 9 1863^ Hot l'9s VS101 To HPIS Airlock
{ToHPIS 18638 .- b p 1860A 12" L] d Sump
% P1A 4 .
12" P 18 18608 O O Simplified Illustration of Containment Leakage Paths at the Surry PWR (source: carlson et al, 1981) O O
'8-57 l x, FICURE 8 - 14 O
O k' O i- .. o m .' ,. . N,. Displacements are 9 i magnified by a
.I. -
1 factor of 50
- n. ,
9 N ~ D N_- O g . O i-0- 6- s#-)' l! o.o cio.o soIo.o iso.o ido.o gado.o R O O Estimated Deformation of Indian Point Unit 3 Containment at a Pressure of 118 psig (Source: Butler and rugeuo 1982] O O
F;. 1 ?-
. 8- 58 l $ ' FICURE 8 - 15 . Q CONTAINNENT .Part
( 1 of 2)p - - ' / P y l . DEFLECTED / LO SHAPE h _ p
/ . /
o j ' L / I I
- o CONTAINMENT WALL I
o I L j L J
' ( MAINSTEAM y,
o , (( \ Nf' dd'-3' ' ( FEEDWATER lo \
\ CRANE WALL \ \
l \ l0 x
\ - MAT 1
- O l
l' Estimated Deformation of Millstone Unit 3 Containment Under Internal Pressure, l Showing Effect of Pipe Anchorages Lakshmipathiah, 1983) [ Source ,O l lO l
m J FICURE 8 .15 (Pcre 2 cf 2) I t CRANE CONTAINMENT
, WALL WALL a g \[\ CONTAINMENT DEFLECTED SHAPE \ MAIN STEAM VALVE D \ BUILDING \
RUPTURE I RESTRAINT [\ o m I' SEISMIC PIPE ANCHCR VALVE
) RESTRAINT 4' - :, \
O )
- Xl , C X :
s> !' x O d{/ ' I i i
/ - /
l R e.yl I
%. /
O PLAN EL 66'-3" ' O Estinated Deformation of Millstone Unit 3 Containment Under Internal Prensure, Showing Effect of Pipe Achorages 'O ISource: Lakshmipathiah, 1983] l l .O 1
\
8-60 l F1 CURE 8 - 16
,) l ' \ = '
LNSOE FACE OF CONTAINMENT WALL *e s (* EMS'-- I r
- \ -
w.___*/_ . . . -
\ - },~:: '
d ,
'J , p,* ' . . /-
( r# s \ , M. . m i I l , .5* o l 1 s I us -) 1 O Postulated Failure Mode of Pipe Ponetration Following Deformation Shown in Figurn 5.15 O (Source Lakshmipathiah, 1983) D
FICURE 8 - 17 g e i . ! 160 ,
... -~ ~ -
140 - LOW LEAKAGE x ., llATER DEPLETED 120 - ls MEDIUM ~ 2 - HIGH LEAKAGE 'LEAXAGE 100 l g ~ e,
- s -
- g y 80 -
l g . m ., W 60 - I 40 - f0 J l 20 Q
'f h 0 ' '
IO - - MEDIUM LEAKAGE I
# f' ~
300 - HIGHLEAKdGE +- LOW LEAKAGE l c0 c - 250 - , .. E ' l i 5 - : lg g 200 . N 150 - t' !O ' i e . > e i i ! 100 1400 1600 600 800 100U 1200 ' 0 200 400 TIME-(MINUTES)
,g ' '
s Egtimated Containment Conditionc for a TMtB' Sequence (with Flooded Cavity) at
- O the zien twa, ror contain%ent 1.cak Arcan Shown in Table 5.8
~~
(Source: NRC, 1984]
- O
l 800
- g. .
700 - i 5 HIGH - E l 600 - LEAKAGE
.06 h l .w 500 - .- 9 l & -
E u 400 - w E w
-' . .
- 04 m M 300 - CONTAINMENT $ -
FAILURE N g "$$"E n E 200 - (LOW LEAKAGE) $ e y [ , - .*02
~
in E n 8 100 f~~
% e. )
I g , o 0 *
~
5 0 200 400 600 800 1000 1200 1400 1600 TIME (MINUTES) , Estimated Containment Leakage for Conditions Shown in Figure 5.17 [ Source: NRC, 1984]
~ _ ,
l
% I U
l
/ ;
CHAPTER NINE
,/ /
Neglected Phencxtena Which can Alter the Nature of Accident Sequences
SUMMARY
t .O . Current source term estimates do not take into account a nmber of important phencmena that might occur during severe accidents. This chapter describes several, which if incorporated, could significantly change current estimates. O , we four most important are steam explosions, high press melt ejection, hydrogen explosions, and steam generator failures. Were is considerable uncertainty over the importance of these phencnena. Until the relevant processes can either be dismissed as insignificant or properly addressed, source term estimates cannot be regarded as convincing or ccruplete. 9.0 Introduction Severe reactor accidents will give rise to a wide variety of phenmena. In many ases, there is great un ertainty about the nature of these phenomena. .O Although the underlying principles of physics and chenistry are well under-stood, the application of these principles to the complex processes of a reactor accident may be poorly understood. .O tis chapter addresses several phenmena which have been " neglected" in the sense that they are not accounted for in current source term estimations such as those [GIESEKE, ET AL. , 1984a-e) 'made at Battelle Cblmbus Labora-g tories (BCL) . mch phenomenon has received sate attention in other contexts. We purpose here is to describe several major phenmena which, if incor-porated into the BCL (or other) source term estimations, could lead to major
,, qualitative changes in the results. Wis discussion is not intended to be v
encyclopedic, and the subject area is in any case in a state of flux. A nmber of reviews addressing phenmenological uncertainties about severe accident behavior are available (GI'ITUS, ET AL., 1982; LIPINSKI, ET AL., 1984b; RIVARD, ET AL.,1981) . O O
9-2 3' Source term estimations begin with the definition of accident sequences. ; men, calculations are performed, using elaborate caputer codes, to predict fg the outcomes of these sequences. Inevitably there will be limitations to the'se codes, arising from difficulties in constructing neerical models, from inem plete understanding of the processes to be modeled, from errors or oversight on the part of the analysts, or fra the prohibitive expense of writing or rtmning codes of the necessary emplexity.
- Such limitations, if they can be overcee, will most certainly lead to quantitative changes in the results for given sequences. We amotmt of a radionuclide estimated to escape by a given pathway, and the timing of its Some of the limitations of current theoretical models
^ escape, will change. (as expressed in cmputer codes), and the implications of such limitations, are addressed in Chapter Seven of this report. ' O 2e present chapter focuses on phenomena which can alter the nature of an accident sequence. Different escape pathways can be opened up, and processes can be initiated for which new or modified codes . are required. Wese phenmena thus imply accident sequences significantly different from those
.O :
previously considered. In part, BCL's failure to account for these significant phenmena arises from their being required to address a limited set of specific accident
- O sequences.9*i Also, these phenomena are not well understood, and thus are difficult to model. However, until the neglected phencuena can either be dimissed as insignificant or properly addressed, source term estimations cannot be regarded as convincing or emplete. .
O Four of the selected phenomena addressed in this chapter have major implications for contaiment integrity. Wese are: steam explosion (Section j 9.1); high pressure neit ejection (Section 9.2); hydrogen explosion (Section
.O 9.3); and vulnerability of IWR steam generators (Section 9.4) . We first two of these phenmena also have implications for the amount of radioactive material suspended in the containment atmosphere at the time of contaiment failure. ' O-
1 0 V ! O 9-3 1 1 l 1
; Section 3.5 deals with two phenomena dich could contribute to secondary ! releases of radioactivity to the enviroment. In this context, it is not O often appreciated that a large inventory of radioactivity is held in spent fuel pools which are adjacent to or within the reactor buildings. A release 1
- frcm such a pool, consequent on a severe reactor accident, is one of the phenmena addressed in this section.
O In each of Sections 9.1 through 9.4, conclusions are drawn specific to the phencnena discussed in those sections. Sare overall conclusions are presented in Section 9.6. O 9.1 Steam Explosions It is possible during a core melt accident for molten material to rapidly O mix with water. Wis could occur within the reactor vessel, or in the reactor cavity beneath the vessel. If such mixirg occurs, heat will be transferred very rapidly fran the molten material to the water, causing a violent expansion of steam - a stean explosion. O Many incidents of this kind have occurred in industrial contexts. We appendix to this chapter describes sate illustrative incidents in the metals, paper, and nuclear industries. O A stean explosion during a reactor accident could have three types of effect. First, were it to occur inside the reactor vessel, it could lead to a breach of the reactor coolant systen (RCS) . Second, an explosion in the vessel O (or, less likely, in the reactor cavity) could precipitate containment failure, perhaps through the generation of a missile which penetrated the contalment wall. Wird, an explosion in the vessel or the reactor cavity could increase the anount of radioactive material suspended in the contairunent O atmosphere. 9.1.1 Stages of A Steam Explosion O From a series of experiments conducted at anall scales (1-20 kg of molten material), four main stages have been observed (GI'ITUS, ET AL.,1982): .O
g_4 O i
"(i) Mixing Stage j O A very hot liquid can intermingle with water to form a quasi-stable coarse mixture, with material zones having dimensions of order 1 to 10 nm. W is mixing process may be assisted by steam formation which can both prevent rapid cooling of the hot liquid, and prmote mixing by stirring the syste. me essential features of this stage are that O heat transfer rates are relatively low and that no signi-ficant increase in system pressure occurs. Typically mixing can be initiated by pouring the hot liquid into a vessel of i water, but many other modes of contact are possible. In i experiments the duration of this stage is often on the !
order of a second. l O (ii) Trigger Process Some localized disturbance, or trigger, must initiate a local increase in pressure. %is can either be spontaneous or external. For exmple, a minor impact could result in 4 O collapse of the ste m layer surrounding a few drops of the hot liquid. Wis could lead to a fine fragmentation of that hot liquid, and hence to a local increase in pressure because of the enhanced vapor generation on the increased surface area available for heat transfer to the water.
.O (iii) Propagation Stage Once the process of fine fragmentation and hence much more rapid heat transfer to the water has started at one point -in the syste , then experiments have shown that rapid propagation through the reainder of the syste may occur.
O n e propagation process has been identified as a shock wave moving through the mixture. me essential feature of the propagation stage is that it results in the coherent behaviour of a large systs. A large sc, ale steam explosion can only occur if rapid heat transfer frcm a large mass of hot liquid occurs coherently. Typical propagation rates O are of order several hundred meters per second. (iv) Expansion Stage Mechanical energy is generated by the expansion of the heated coolant vapor. mis process can be closely coupled O with a propagation process, in which energy needed to sustain a shock wave is derived frcm coolant expansion. Surrounding structures can be loaded both by the enhanced pressure and by the impact of materials flowing away from the explosion."
- O l
O 9-5 9.1.2 The Great Debate: In-Vessel Steam Explosions Which Generate Massive Missiles "O Although explosions outside of the pressure vessel have received some attention, many analysts have focused on one potentially very serious event: an explosion within the vessel which could blow the head off the reactor vessel with a velocity sufficient to penetrate the containnent building. For O such an event, the source term would be large. It is often assuned that if this event can be ruled out, other steau explosion events would not present a direct threat to containment integrity. O Blowing the top off the pressure vessel could occur if a sufficiently large steau explosion occured within the lower part of the vessel, which would then function like a cannon barrel. A slug of water and core material would be pr pelled upwards, driven by the force of the expanding stean. Wien this O slug struck the underside of the vessel head, it could have sufficient manentuu to blow the head fran the vessel and propel it upward at a high velocity. Even if the botton of the vessel were blown off by the explosion,
'h*** *iih* "'ill h* *" 9h *"*"9Y ' " ""* 'h* "PP*" h**d * **P*'*'**
O Were is considerable controversy over the likelihood of such an event. Experts take varying positions as to its probability, with sane arguing that it is physically impossible. Were is consensus, however, that no position !O can be unambiguously excitried on the basis of present technical evidence. 'Ihe levels of uncertainty about various stages of the event are such that experts must rely on technical jtrigment at several points. O 'Ib blow the top off the pressure vessel a large mass of molten core material must first fall rapidly into a pool of residual water in the base of Figure 9-1 shows one way in which this could happen. Molten the vessel. l material fran the initial melt region would re-solidify as it reached lower 'O (cooler) parts of the core. In this manner, a molten pool would be suspended on a crust of re-solidified core material. A similar crust could entirely surround the melt pool with a thermally insulating blanket, as shown in Figure 9-1. Eventually, for instance when the melt region reached the lower core O plate, the crust would be penetrated and molten material would flow rapidly into the residual water. O
2 9-6 e 4
; Another plausible mechanism for rapid formation of a falling mass of j molten material would be initiated by restoration of cooling water flow to an k already overheated core. For illustration, if the icwer one-tenth of a dry core near its melting tmperature were stridenly covered by water at atmos-p pheric pressure, it has been calculated that an instantaneous heat transfer rate of 1.5 W could result. We resultant steam, if it fully entered into a f
f heat-generating chemical reaction with the zirconim in the fuel cladding, I would lead to a peak power output of 10 GW (3 times the nminal thermal j operating power for a typical plant) . Since the energy required to destroy j the entire core gemetry might be as little as 6 GJ (1 GI = 1 m for 1 O second) , a large mass of molten material might be formed very rapidly (RIVARD, ET AL., 1984]. i { Given the rapid falling of molten material into residual water, an O explosion would occur as discussed above. We various stages of such an explosion are shown in simplified form by Figure 9-2. In the first stage of mixirs, the molten material would be dispersed into relatively large segments, as indicated by the " coarse mixing" stage of Figure 9-2. At some point, an O explosion would be triggered and a pressure wave would spread outward from j this point (Stage 2) . Molten material would be dispersed by this wave, forming fine fragments which would rapidly transfer heat to the surrounding water. We resultant generation of steam would generate pressure leading to D further propagation of the wave, which would spread through the entire mixture
- j (Stage 3).
In the final stage, a violent expansion would occur. ihe lower head of
- O the vessel could be blown off at this point. mether or not this occurred, a slug of water and core material would be driven upward at high speed. If its mcrrients were sufficiently high, this slug could blow the head frcm the vessel with sufficient velocity for it to penetrate the contaiment.
O It is very difficult to theoretically model such a cmplex process, and extreely expensive to perform experiments at a scale approaching full size. Moreover, many of the contributing phenmena are poorly understood and may in O any event have a high degree of intrinsic randeness. 2 -O
i J 9-7 Despite these uncertainties, various analysts have sought to attribute probabilities to the overall process. Sate have assigned a probability of J zero - that is, sufficient mmente cannot be imparted to the vessel head under any ciremstances. Others have assigned median probabilities r'arging from 10-2 (one in one hundred) to 10-4 (one in ten thousand) per core melt. 2 O S=e interesting calculations were recently performed at the sandia National taboratories (SNL). A simple theoretical model of the process described above was constructed, dependent on five major parmeters. We possible range of each parameter was estimated, and then divided into three O regions. ten, a Monte Carlo cmputation was perfomed, in which a cmputer code representing the model ms run 10,003 times, drawing upon different parts of the ranges of each major parmeter.9.3 O Table 9-1 shows the results. If the full range of each major parameter is included, then 20% of core melts (2017 cases out of 10,000) are predicted to lead to a steam explosion which is sufficiently vigorous to breach the botta of the reactor vessel. About 5% of core melts (460 cases out of O 10,000) are predicted to lead to an explosion severe enough to separate the vessel head and impart to it a velocity over 50 Ws. Likewise, about 3% of core melts (267 cases out of 10,000) are predicted to result in a head velocity over 90 Ws. Wese two velocities were selected as representing the Q upper and lower bounds of the range of minimm velocities necessary for the head to penetrate the contaiment. l l 1 If each major parameter is confined to the lower third of its range, ! Table 9-1 shows that neither vessel nor containtent breach is predicted for O any of the 10,000 core melts. By contrast, if each parameter falls in the l l upper third of its range, contaiment breach is almost certain [BER%N, ! SWENSON & WICKETT,1984] . l O he actual nebers shown in Table 9-1 should not be viewed too seriously. However, the exercise is very instructive in that it shows the importance of reaching a better determination of the possible ranges of the major O parmeters. Other approaches to this proble, which often rely on personal jtx3gment without being explicit as to its implications, are less helpful.9*4 O
O 9-8 W is SNL work did not address BWRs. Nor did it address core melts during which the RCS renains at high pressure. Some analysts claim that steam O explosions are less likely at high pressure, but this claim lacks a foundation of experinents at appropriate scales. In stmmary, there is large uncertainty about the probability that a core
.O nelt will lead to generation of a missile which breaches the containnent. Such a breaching, however, can by no means be ruled out a priori. Large-scale experiments and further analytic efforts could reduce the uncertainty, but there will always remain an uncertainty deriving from the unpredictability of -O the various related phenmena.
9.1.3 Steam Explosions in Other contexts O Smaller steau explosions -- those with insufficient energy to generate a contairnent-danaging missile -- can also threaten the contairment integrity of W'Rs . We vulnerability arises because m'R steam generator tubes could be ruptured by a relatively snall steau explosion. Section 9.4 explores this O issue in more detail. For many EWRs, the cavity below the reactor vessel will contain water at a tine when molten core material is also present. If same of this molten O material were to rapidly mix with the water, one or more steam explosions could occur in the cavity, generating a pressure pulse and possibly increasing the amount of radioactive material suspended in the contairunent atmosphere.
'Ihis phenatenon is discussed in more detail in Section 9.2, in the context of O high-pressure -- and hence, high-velocity -- melt ejection. However, ex-vessel stean explosions could also occur if the melt were not ejected at high velocity.
O BWRs of Mark I and II types could also suffer ex-vessel steam explosions resultirg fra a gradual flow of molten core material into their suppression pools. W is possibility has received relatively little attention. i
- O
!O
9-9 9.1.4 conclusions
~
Although there is great uncertainty as to the likelihood, the possibility exists that large stem explosions can occur within the reactor vessel with a vigor sufficient to blow off the top of the vessel and breach the centalment. Much more experimental and analytic effort is needed to understard the various i O phenmena associated with such explosions. Smaller steam explosions inside reactor vessels could also threaten NRS by overpressurizing steau generator tubes, thereby openirg a pathway to the l 0 enyiroment, Explosions outside of reactor vessels could also occur and threaten contaiment integrity. Such stean explosions may also lead to an increase in C the amount of radioactive materials suspended in the contalment atmosphere -- an important effect even if contaiment failure occurs for other reasons. 9.2 High Pressure Melt Ejection 'O For many accident sequences, it is expected that the reactor core will melt while the BCS reains at high pressure.9* In that event, molten core material could flow into the bottom of the reactor vessel, and melt through
- O the vessel wall, while the BCS remains pressurized. Molten material could then be ejected fran the vessel at high velocity, driven by pressure inside the BCS.9.6 C In recognition of this possibility, various theoretical and experimental investigations have been conducted or are in progress. Weir purpose is to better understand the process of high pressure melt ejection (HRiE) and its implications for the source teun issue. We concerns raised by HPME are O twofold. First, HR4E provides mechanisms for the suspension of radioactive material in the contairinent atmosphere. Second, it can lead to a substantial increase in contaiment pressure, potentially leading to contaiment failure.
Wat pressure increase could cme frcm direct heating of the contaiment O atmosphere, fran cmbustion of the molten material, or fran an ex-vessel steam explosion. O
i
'O 9-19 9.2.1 Conditions Leading to HPME
.O If HPME is to occur, the BCS bomdary must maintain its structural integrity until the molten core has formed a pool inside the botta of the reactor vessel. Further, the core must melt through the vessel wall in such a O way that molten material flows into the reactor cavity at high velocity. Were is some dispute as to the likelihood that these two conditions will be 4 met. !O tere are several effects which could decrease the likelihood of the core 1 melt conditions needed for HmE, as illustrated by the following two effects. First,-an in-vessel steau explosion could blow open the lower end of the reactor vessel, thus precitriing HME. Wat possibility is addressed in 3 Section 9.1 of this report. Second, the taperature of the BCS boundary might 'O closely follow the core tenperature for accidents in which the RCS is pressurized. If so, the decline of material strength in the RCS at higher j temperatures could cause a loss of structural integrity, leading to O depressurization before the molten core simps into the bottom of the vessel. For example, L Were are calculations supporting this second possibility. one simple theoretical model of a IWR [NOURBAKHSH, ET AL., 1984] shows that, ! if convective circulation through the steam generators is assuned, the upper Q internal strtx:tures of the BCS will closely follow the core temperature while
?
the stean generator tubes will lag by only 70-150 C. If convective circula-ff' tion is limited to the vessel, the temperature of the upper internal struc-O tures will lag the core tenperature by only 68-12e C. Since the core i temperature can substantially exceed the point at which RCS structural integrity is lost, depressurization would occur before core slunping if this model is representative of accident conditions.9*7 ,7 Y his model, however, is based on a nunber of assunptions, see of which may not hold. For exanple, the model assunes that the core is porous, thereby In pennittirg convective heat transfer away fran the central core regions. !O reality, core material may melt in one region and re-solidify in cooler regions. In this manner, a large melt pool may form, thennally insulated fran l its surroundings by a crust of re-hardened melt (see Figure 9-1) . In this .[. o d i i ! 1
< -, )
;O 9_11 ciremstance, convective heat transfer may be sufficiently inhibited that heat cannot be transferred to structures forming the RCS boundary in sufficient O amount to raise the tenperatures of these structures to their failure point.
As discussed in Section 9.1 of this report, fonnation of a thermally insulated melt pool could be a precursor to a severe in-vessel steam _O If collapse of this melt pool does not lead to a stean explosion explosion. which breaches the RCS boundary, then the molten material will be available for HPME. l
'O ~
If a melt pool does form in the base of the vessel, with a bubble of stean and gases pressurizing it fra above, HPME will only occur if the vessel wall is penetrated in such a way that high-velocity ejection occurs. In particular, if the entire botta head drops off (due to cire mferential
'O melt-throtgh of the vessel wall) before penetration occurs, then the bulk of the melt will be unavailable for HINE. 'Ihis problen has been investigated theoretically, with some results available for the Indian Point NRs [ POWER AUTHORITY OF NW YORK & CONSOLIDATED EDISON, 1982). Calculations were O performed for cases representing eighteen different combinations of paraneters, yieldirg the following outemes: - In twelve cases, only the instrmentation tubes would fail, and the core melt would be entirely dispersed through the O holes left by ejected instrmentation tubes. Failure of instrtunentation tubes would lead to high-velocity melt ejection in a manner similar to that shown in Figure 9-3. - In four cases, instrmentation tube failure would be followed by botta head failure. For these cases, the first part of
'O the core melt would disperse through the holes left by ejected instrmentation tubes, again in the form of high-velocity melt ejection. 'Ibe second part of the core melt would escape through a tear in one side of the vessel, and the third and last part of the melt would stay in the bottczn head as it dr pped ff the vessel.
..O - In two cases, the bottczn head of the reactor vessel would fail prior to (or simultaneously with) instrtunentation tube failures. For these cases, a part of the core melt would escape through a tear on one side of the vessel, and the rest f the melt w uld renain in the bottczn head as it dropped off .O the vessel.
- O
.O 9-12 In sm , therefore, HmE is a credible event. As in the case of steam explosions, it is difficult to estimate the probability of an accident O involving high-pressure melt ejection.
9.2.2 Effects of Melt Ejection in the Reactor Cavity
.O If the cavity below the reactor pressure vessel is dry, the molten material will sweep through the cavity and enter the contaiment atmosphere.
Should the cavity contain water, an energetic steam explosion could oce:ur.9.8
'O As part of an experimental program at SNL, experiments have been condu:ted in which HmE into a water-filled WR cavity .has been simulated.
We results of two of these experiments are instructive, in tenus of the potential for a violent steam explosion: O
- Ten kg of molten thennite was ejected (at a speed of about 20 m/sec) into a water-filled cavity (constructed of aluninun) simulatirg the Zion plant at 1/20 scale. At about the time of impact of the melt jet on the floor of the cavity, a steam explosion occurred, with a peak pressure of 30,000 psig on the O walls of the cavity structure and a pulse rise-time of about , 75 micro-seconds. We test apparatus was destroyed (TARBELL, 1983]. - In a later test, 80 kg of molten thermite was ejected into a water-filled concrete cavity simulating the Zion plant at 1/10 O scale. Again, a steam explosion occurred, with a peak g
pressure of 500 psig on the cavity walls. Water slugs 1 containing quenched debris were ejected frcxn the keyway and travelled up to 100 ft, followed by dispersed melt. We
- cavity was destroyed (REHM, 1984).
D Such a steau explosion durirg an actual accident event would generate a pressure pulse on the contairunent. We pressure spike might present a threat to contairunent integrity, particularly for NRs with Ice condenser contairinents, which have relatively small volunes and low design pressures. g In addition, the stean explosion could increase dispersion of molten material into the contairinent atmosphere. With or without a steau explosion, some of the molten material will enter g Once there, it will heat that atmosphere 1 the contairunent atmosphere. 6 't: )
i 9 9-13 I
} (directly or through embustion) and some of it will renain suspended as an aerosol. At present it is difficult to model the processes by which molten ; material will pass through the cavity.
9.2.3 Effects of HPME on Contaiment Pressure i O Wat part of the melt which enters the contaiment at2nosphere as fine f particles will lead to heatirg of the atmosphere, both directly and by canbustion. Direct heating will be important even if combustion cannot occur, I for instance in those contaiments which operate with inert atmospheres. If O sufficient oxygen is present, however, the fine melt particles will burn. We principal constituents of the nelt will be zirconiun, iron, and U02, each of which will react exothermically with oxygen.9*9 O A variety of effects will contribute to the total contalment pressure. First will be the pressure increase due to the escape of steam and hydrogen fran the RCS. Second, any stean explosion in the reactor cavity will yield an additional pressure pulse. Wird, melt particles entering the contairunent O atmosphere will directly heat that atmosphere. Eburth, melt particles may j burn exothermically in that atznosphere. Finally, the melt particles may ignite hydrogen already present in the contaiment atmosphere. O Table 9-2 shows some of the results of calculations undertaken at SNL to
' estimate contalment pressures at the Zion plant for various possible outcanes of a HmE event. In at least one and probably two of the three cases shown, the peak contalment pressure (202 psia) exceeds the estimated contaiment O failure pressure (two estimates - 149 psia and 126 psia -- are cited in the - footnotes to Table 9-2) . Given the assunptions underlying SNL's calculations, contairinent failure is evidently possible for a variety of HmE events.
O tese calculations are conservative (worst case) in that they assme that 100% of the core participates in the HME event. Wis will be an unrealistic assunption in most cases, but not necessarily in all. It will be recalled from Section 9.2 that rapid melting of the entire core could occur in the O event of delayed delivery of coolirs water to an already overheated core. 0 0
'O 9-14 siowever, the likelihood and implications of this for high-pressure sequences j deserve more exmination. O In other respects the SNL calculations are not conservative. @us, they assme no ccabustion of UO 2 in the melt, and do not include the effects of hydrogen combustion. Finely divided melt particles could initiate hydrogen O ccabustion over large regions of the contaiment atmosphere in a very short time. With such a " distributed" ignition source, the normal limits to ; hydrogen ccrobustion may not apply. Wus, very lean or steam-rich mixtures of hydrogen may ignite even though they would not do so for localized ignition O sources. We Zion plant possesses a relatively strong contaiment with a large interior volme. @us, other plants with weaker or analler contalments - O both EWRs and BWRs -- may be at risk even if it can be demonstrated that the Zion contaiment can withstand all credible HENE events. 9.2.4 HPME and Airborne Materials in the Contaiment .O It is an open question as to how much of the ejected melt will beccrne suspended in the contaiment atmosphere, with or without a steam explosion in the reactor cavity. Were are nmerous obstacles in the path of the ejected O strem, and some fraction of the melt will be deposited on these obstacles. Likewise, the distribution of sizes of the suspended particles is an open question. 'O Four experiments have been performed recently at SNL, in which 80 kg of thennite melt was ejected into a concrete reactor cavity simulating the Zion plant at 1/10 scale. In two tests the cavity was dry, while in the other two tests (one of which is mentioned above) the cavity was filled with water. For O all four tests, nearly all (more than 95%) of the melt was ejected frcm the cavity. A wide distribution of particle sizes was observed, with a mean of about 0.5 nm.9*18 Large particles renained molten throughout their flight, which carried than up to 100 m frcm the cavity [TARBELL, PIICH & BROCKMANN, O 19841-O
O 9-15 3 Future experiments of this kind will be conducted in a 90 m chamber rather than in the open, which will pennit the measurenent of important O parameters. Also, obstructions will be placed in the melt flight-path, to simulate obstructions in an actual contaiment. Heating of the contaiment atmosphere, directly and via melt combustion, O will be affected by several factors, incitriing the velocities and dimensions of the nelt particles. In turn, the characteristics of these particles will be altered by their own cmbustion or their initiation of a hydrogen burn.
'Ihus, experiments must simulate accident conditions fairly closely if their O results are to be useful.
If the HmE event leads to prcznpt contaiment failure, the source term will consist of material suspended in the contalment atznosphere prior to the 0 event, aerosols generated during HmE, and subsequent releases arising from evolution of aerosols into a breached contaiment.9*11 Each of these contributors to the source term will exhibit a different O chenical ccrnrositon. Aerosols present in the containnent prior to HME will be dminated by volatile species (such as Cs, I, 'Ib) , while those generated by HmE will more closely represent the cmposition of the molten core. Aerosols generated after containment breach, particularly those generated by O nelt-concrete interaction, will incitrie refractory fission products (POWERS, ET AL., 1983]. 9.2.5 Conclusions O HmE is a potentially significant event for all types of PWR or BWR. At worst, it could lead to a prompt failure of containnent with an accmpanying large release of radioactivity. Much more theoretical and experimental work O is required before its full implications for the source term issue are apparent. O l O
3 9-16 9.3 Hydrogen Explosions lO During a severe accident, large anounts of h%rogen will be generated by ) chenical reactions between steam and zirconim in the fuel cladding (and elsewhere in the core structure) . Escape of this hydrogen frcIn the RCS to the > contaiment can lead to substantial concentrations of hydrogen in the contain-O ment atznosphere. Figure 9-4 shcws the predicted hydrogen concentrations, by volune, for various kinds of contaiments and assuned metal-water reactions. It can be seen that for extensive retal-water reactions the hydrogen concentra-tions can be very high, especially for BWRs. Explosions of such hydrogen-air O mixtures could present a major threat to containnent integrity. We implica-tions of such hydrogen explosions have not been fully considered in source tenu estimations made up to this date. 4 g Were are a number of processes, aside frcan steam-zirconian reactions, that can generate hydrogen during accidents. An iron-stean reaction can also occur, although its initiation tenperature is scxnewhat higher than that of the zirconium-stean reaction. Radiolysis of water, and corrosion of zinc-based O paints, g lvanizing layers, and aluninun structures in the contaiment will
; contribute further hydrogen. When the molten core falls into the reactor cavity, melt-concrete interactions will release hydrogen and carbon monoxide.
Wese various processes can contribute an anount of hydrogen ccznparable to that frcru the zirconiun-steam reaction, although several tens of hours may be O
- required before such a ccrnparable contribution is made [ CAMP, ET AL.,1983) .
Because of the explosion threat, Mark I and II BWRs normally operate with g their contaiments filled with inert gas (although short periods of operation with de-inerted contaiments are allowed -- Chapter Four) . Although this will prevent explosion in the containnent itself, hydrogen may reach explosive concentrations if it is released into the reactor building sich surrounds the ntaintent. Mark III BwRs and Ice cnndenser contaiment mRs are equipped O with electrical igniters which are meant to burn off hydrogen before it reaches high concentrations in the containnent atmosphere. Il0 wever, these igniters would not operate during accidents where electrical powr was lost. g Large Dry (bntainnent NRs have no provisions to prevent hWrogen explosion.
n j>- 9-17 9.3.1 Types of Explosion O There are two qualitatively different types of explosions. A deflagration occurs- when heat is transferred relatively slowly (primarily by condtx: tion) from burned to unburned gas. W e flame front may travel as slowly as 1 Ws . In a detonation, a shock wave moves with a speed which is O supersonic relative to the unburned gas. Heating of gas to its ignition tenperature occurs primarily by shock canpression. Flane speeds may reach 2000 W s (SHERMAN, W AL., 1980]. 10 Transition fran deflagration to detonation may occur in a manner which is not well understood. It has been observed that turbulence in the flate front, either generated by the presence of obstacles or occuring naturally, can accelerate that front. If the turbulent flate speed exceeds about 1/10 of the O sound speed, shock waves will form ahead of the flame front. Wis process can continue until a full-fledged detonation wave is created [ SHER %N, W AL., 1980]. O Transition phenanena are being studied experimentally using the FIR 4E apparatus at SNL.9.12 Durirg these experiments, transition to detonation has been observed even in the absence of obstacles, leading to high overpressures which in one instance danaged the heavily reinforced structure [SHERMAN,1984] . O 9.3.2 conditions Favorable to Explosion Neither very high nor very low hydrogen concentrations (in air) are O conducive to explosion, tereover, high concentrations of steam may prevent explosion, a process known as " steam inerting." However, the limiting concentrations of hydrogen (and steam) are not always well defined. O Deflagrations will propagate in steam-saturated air (at ambient tenperature and pressure) when the volute fraction of hy3rogen exceeds 4-9%. Wey will not propagate if the hydrogen volune fraction exceeds about 75%. 1 Stean inerting is fully effective when steau is present in a volute fraction C of about 60% [ CAMP, W AL. ,1983] . l O
.O 9_18 2e limits of hydrogen concentration within which detonation can occur are narrower, although not well defined. It has until recently been assmed O that lower and upper limits to detonation of hydrogen-air mixtures arise when hydrogen volune (mole) fractions are 18% and 59%, respectively. However, recent experiments at SNL have extended these limits to 13.5% and 70%. It is significant that SNL's experiments are at a larger scale than previous tests O [SHERMAN, 1984].9.13 9.3.3 Effects of Explosions O A deflagration will spread relatively slowly through the contalment at:nosphere. @us, dynanic effects will be unimportant. However, substantial pressures can be generated at high hydrogen concentrations. For exanple , Figure 9-5 shows that the burning of a 30 percent hy3rogen mixture in the O contairrnent can lead to a contairrnent pressure of more than 170 psi. Detonations can produce severe shock waves. Figure 9-6 shows that a detonation of a 30-35% hydrogen mixture can produce a pressure behind the O reflected shock of more than 500 psi. Due to the rapid motion of these waves, dynamic ef fects must be considered. The level of stress induced 'in contaiment structures will depend on the relative timescales of the pressure pulse and the structures' natural vibrations. Short-duration pulses will O produce less stress than pulses whose duration is canparable to the period of } one of the containnent's natural vibrations.9*14 It will be noted (see Figure 9-6) that the pressure pulse fran a O detonation is expressed in terms of the ratio of final pressure to initial pressure. 'Ib a first approximation, this ratio is independent of the initial pressure. Wus , pre-heating of the atmosphere before the detonation will increase the peak pressure. O The pressure ratios shown in Figure 9-6 are for detonation waves propagating in one dimension. However, even higher pressure ratios can occur where two-dimensional and three-dimensional effects are important, as would O usually hold for detonation in a reactor containnent. Wese higher pressure ratios will arise from wave focusing effects. 'No-dimensional calculations of i I
O 9-19 ) I the detonation process have been conducted for the contairunent of Indian Point Unit 3 (a Large Dry Contairstent IWR) . Wese calculations indicate a peak O pressure at the contairunent wall which could exceed 70 times the initial contairnent pressure. A nunber of successively lower pressure pulses follow this initial pulse, because of shock reflections frm other surfaces, but af ter a few reflections (3 or 4 at most) , pressures decay to a steady value O IDELICHATSIOS, ET AL., 1982].9*10 As noted above, deflagrations can change into detonations through acceleration of turbulent flame fronts. We transition may be inemplete, O yielding a pressure pulse less severe than that for a emplete detonation. Wis phenatenon, known as a quasi-detonation, is poorly understood. However, it is reasonable to assune that the phenmenon widens the range of conditions under which danaging pressure pulses can occur. O 9.3.4 Likelihood of Detonation Detoriation requires the accunulation of hydrogen in relatively high
- O concentrations, greater than 14 percent by volute. Wis requires that embustion be suppressed until the hydrogen has time to accunulate.
As mentioned above, Mark I and II BWRs normally operate with inerted O containment atmospheres, thus ruling out an in-containment detonation (provided that inertirg is correctly carried out) . For Mark III BWR and Ice Condenser PWR plants, high hydrogen concentrations could occur if the electrical igniters were inoperative, potentially leading to detonation or O quasi-detonation. For example, calculations have been performed to estimate the hydrogen, air, and stean concentrations in different empartments of an Ice Cbndenser O PWR contairunent in the event of a severe accident in which igniters are inoperative.9.16 Detonable mixtures were predicted for the done region and for the ice cmpartments (HASKIN, BEHR & CAMP,1984] . O Large cry contairment swr plants are not equipped with igniters. mus , hWrogen could accunulate to high concentrations before burning, at least in O
9-20 local regions of the contaiment.9*17 Althotgh the contaiment atmosphere would be steam-inerted under many such conditions, there are accident ) O, sequences in which the steam could be condensed. For exanple, if contaiment cooling were restored late in an accident sequence, af ter considerable hydrogen (and carbon monoxide) had acetmulated, the steam could be rapidly condensed. Figure 9-7 illustrates this possibility, showing how the '# restoration of contaiment cooling could lead to steam condensation and hence de-inerting. 'Ihe pressure rises shown in this figure arise frm hydrogen deflagrations; detonations may also be possible. O 9.3.5 conclusions All U.S. plants are liable to hydrogen explosions in their containnents, with the exception (in normal circtmstances) of Mark I and II BWRs. Even n 's these latter two categories of plant are liable to explosions in their reactor buildings. Under a variety of conditions, hydrogen explosions can take the form of O detonations or quasi-detonations. such explosions would lead to large pressure pulses which can breach the containnent or danage structures or equipnent within the containnent. O The basic phenomena, particularly transition from deflagration to detonation, require more experimental and theoretical work. Source term estimations conducted to date have not reflected this state of uncertainty. In particular, the likelihood and consequences of detonation have not been 0 addressed in the tcL work. 9.4 Vulnerability of PWR Steam Generators O Both BWRs and IWRs have a variety of interfaces -- primarily in the form of pipes and valves -- between the reactor coolant system (RCS) and systens outside the containnent. In certain accidents these interfaces may becme imgortant pathways that radioactivity can take fra a danaged core into the 0 environennt. twRs are of particular interest in this respect, in that their stean generators represent a large interfacing area between the BCS and the O
O 9-21 secondary coolant system. Since the secondary systen is designed for lower m pressure than the RCS, the opening of a path to the environtent, especially J through a stuck-open safety valve in a secondary loop, is a likely consequence of a substantial failure of this interface.9.18 A typical steam generator consists of thousands of thin-walled tubes.9*19 We uter fran the core passes through these tubes at a pressure in excess of 2000 psi and generates steam in the secondary loop, typically at a pressure less than 1000 psi. Degradation of these tubes has been a chronic problen for operating PWRs, but their vulnerability in severe accident situations appears not to have been fully appreciated. Because of their anall thermal mass, steam generator tubes can readily overheat, with a consequent loss of strergth. Were is sate reason to believe q " (see Section 9.2) that heat will be readily transferred from the core to the RCS boundary durirg high-pressure core degradation sequences, leading to structural failure and depressurization. Steam generator tubes would be articularly vulnerable to such overheating and failure. Another possible source of heating of the generator tubes would be fission products transported fran the core region by convective circulation or by passage to an PCS breach downstream of the steam generator. We con-9 " siderable height of the stean generator tubes (18 m for the Surry Plant) would provide a driving force for convective circulation [RIVARD, ET AL., 1984].9.20 te heating effect of fission products on stean generator tubes has been estimated for the Surry plant. A simple calculation has shown that deposition of the core inventory of Cs, %, and I in the stean generator tubes would lead to a tenperature rise in those tubes of 6 degrees C. per minute [RIVARD, ET AL.,1984].9.21 O Wis calculation neglects heat losses fran the tubes, and makes the unrealistic assunption that 100% of the core inventory of these three elenents would be deposited in the tubes. However, deposition would in practice be O uneven, and " hot spots" would occur . Wus , deposition of a much snaller O
1 O 9-22 l l fraction of the core inventory could still be dangerous, by leading to the overheating of a snaller part of the total tube area.
'O Failure of tenperature-weakened steam generator tubes could occur under stress of the high - but relatively constant -- RCS pressure experienced during progress of a high-pressure core melt. However, failure would becme O more likely if a pulse of RCS pressure occurred as the molten core simped into residual teter in the base of the vessel. Even a mild in-vessel steam explosion could produce a substantial breach in the steam generator tubes.
O As nentioned above, degradation of steam generator tuber has been a problen for many EWRs. Wis is a safety issue in the absence of core melt, in that one or more tubes could fail suddenly while the plant is operating, thereby creating a breach in the RCS. Pereover, str$ den loading of the tubes O -- for exanple, by the pulse of scs pressure arising from a " scram" failure -- would increase the chance of tube failure. In the event of a high-pressure core melt, the loading conditions could be even more severe. Yet, the degree of degradation may not be well known for a particular plant.9.22 O j 9.4.1 Conclusions
@e failure of steam generator tubes in IMRs could open up a direct path O for radioactive materials fran the core region to the environnent. mis j vulnerability is enhanced by the likely weakening of these tubes by heating effects occurring during a high-pressure melt sequence. We problen is cmpounded by the possibility of " steam spikes" or in-vessel steam explosions iO occurring af ter the tubes have become weakened in this manner. In addition, the tubes may be weakened to an unsuspected degree prior to the accident.
We entire problen of stean generator vulnerability deserves much closer iO attention.9.23 9.5 Secondary Releases O 'Ihis section briefly describes two phenomena which could lead to .'. secondary releases. In this context, " secondary" means that the release fI
- O l
S I 9-23 f 1 - process would be separate from the main processes exanined in current source tenn estimations. [] I We first phenomenon is the pcssible re-evolution of radioactivity from various sites within the contaiment, subsequent to an explosive containnent t I failure. In this case, the radioactivity would have been previously released
,) frm the reactor core to the containment. We second phenmenon is release fran a spent fuel pool, an entirely different source of radioactivity.
In each case, considerable further investigation is required. Our O discussions are therefore brief. 9.5.1 Re-Evolution after Explosive Contairrtent Failure O As explained in Qiapter Eight of this report, contaiment failure could occur explosively. In that event, pools of accident.-heated water in various parts of the contaiment (for exanple, in the reactor building sump) would boil due to the rapid reduction in pressure. Some of the radioactivity O present in those pools could be released to the atmosphere as a result of violent boilirg (flashing) . Also, the containnent atmosphere would flow out of the breach at high velocity. Wat flow could entrain radioactive material deposited on surfaces within the contaiment. O
+ Preliminary calculations9 .24 suggest that, for hole sizes above about 15 2 up to 25 percent of deposited material could be resuspended fran dry m,
surfaces as a result of high atmosphere velocities. We mass of water flashed O fran pools within the contairment could reach 100 tons (for a hole size of 1000 m ) . We fraction of dissolved and suspended fission products in the 2 pools which would be resuspended in the event of such flashirg could be as 2 high as 60 percent (for a hole size of 1000 m ) .9~25
^ ***i""" V 1
- f h le O 2 size would be 1200 m -- the horizontal cross-sectional area of the contaiment [LIPINSKI, ET. AL., 1984). In practice, the hole size could be much snaller. Resuspension and flashing phenanena would be correspondingly 1*** "I l'"
O O
-_ . . _ _ ., ___ _ _ . , - . m._. - _ _ __ _ _ _ _ _ _ _ . . . _ _
O 9-24 At the time of contalment failure, much of the radioactive material O released to the containnent atmosphere could have settled on surfaces or become suspended or dissolved in water pools. @us , these resuspension phenonena deserve close attention. O 9.5.2 Release from a Spent Fuel Pool Large quantities of spent fuel (often the equivalent of several reactor cores) are stored in spent fuel pools which are adjacent to reactor O contaiments (or, in some instances, inside the contaiment) . Wese pools contain large inventories of long-lived radioactivity which, if mobilized, could lead to substantial enviromental contamination. O mis possibility is often discounted because of the apparent lack of a driving force for such mobilization. However, such a driving force can be created because the fuel assenblies are catmonly stored inside tubes with walls made of neutron-absorbing materials, so as to suppress criticality. In .O the event of water loss fran the pool, those tubes will also suppress heat transfer fran the spent fuel, leading to cladding tanperatures which may be high enough to initiate zirconium-air or zirconium-stesn reactions. We heat released by such reactions provides a large driving force. Volatile O radionuclides (such as Cesium-137) can be released fran the fuel. We surrounding buildings are often flimsy, and may in any case be destroyed by an explosion of hydrogen generated in the zirconium-stean reaction. @us , a significant release to the enviroment might occur. O A severe reactor accident could lead to loss of water fran the spent fuel pool in two ways. First, violent phenanena such as hydrogen explosions could lead to a breach of the pool. mis would be most significant for those plants O (such as mrk I and II bht <s) where the pool is above grade level. Second, the pool cooling systens may be disabled as a part of the reactor accident sequence. Repair of these systens might then be precitxled for several weeks or longer, due to high radiation fields around the plant. Mster would then be O 1 st by evaporation, leading to uncovering of the spent fuel in times of the order of a week or two (the time depending heavily on the age af ter discharge of the most recently discharged spent fuel) . O
i 'Q 9-25 Calculations have been conducted at SNL to estimate the cladding tmpera-O ture in a typical pool in the event of water loss. 'Ibese calculations focused l on the case of total pool drainage - partial drainage is actually a worse f case -- and used a simple model for radiation heat transfer. Nevertheless, they showed that cladding tanperatures could exceed 1000 C, at which tenpera-O ture both the zirconium-air and zirconium-steam reactions proceed vigorously (BENJAMIN, E?P AL., 1979). The most recently discharged fuel will be most at risk of comnencing a .O self-sustaining zirconiun oxidation. However, the heat so generated may raise the cladding tanperature of surrotoding fuel assemblies to the ignition point. In that manner, the oxidation process may. propagate through the entire fuel pool. A recent NRC-sponsored experimental and theoretical study has concluded .O that a zirconium-air reaction can propagate through the pool in that way (pISANO, E7F AL.,1984] . It is clear that a release from a spent fuel pool could significantly add
- O to the source tenn for a reactor accident. Further effort is required to assess the importance of this phenmenon.
9.6 Conclusions .O This chapter has briefly addressed several important phenmena, none of which are adequately accounted for in BCL's source teun estimations. Each of these has the potential to substantially alter the nature of a severe accident
- O sequence.
Considerable uncertainty surrounds each phenomenon. In some cases, further analysis could greatly reduce that uncertainty. In other cases, the 10 base of experimental knowledge needs to be substantially extended. of particular note here is the importance of scale - in see cases it is not possible to accurately predict behavior at reactor-scale based on anall-scale experiments. Also, for same phenomena there will always renain a residual ' to uncertainty deriving from inherently randam processes.
O 9-26 Until all these phenomena are accotmted for in the source term computer c des, r sh wn t be unimportant, source tenn estimations such as those made O by BCL are incomplete. @ose estimations may exhibit the state of the art in source term modeling - but they do not represent a realistic assessnent of the source terms from a wide range of actual reactor accidents. .,0 An important task for the NRC, and for other bodies involved in the source term issue, is to carefully catalog all the significant phenmena. We nature of the analytic and experimental effort necessary to reduce uncertainty about each phenanenon should be estimated. If a particular effort is O prohibitively expensive, or if large residual uncertainties renain after further research, then subsequent source term estimations should reflect a corresponding level of conservatisu. .O 'O O O O 'O
- O
O :: 9-27 s 9.7 End Notes for Chapter Nine 'O ,, 9.1 BCL's work was carried out for - the Nuclear Regulatory Camission (NRC) , whose staff specified the accident sequences to bd addressed. Chapters Ebur, Five, and Six of our report discuss. sme ' implications of, confining source-term estimations to this limited set of seNtees. 9.2 For stamaries of the valirious positions, see (BERMAN, SWENSON & WICKETT, 1984; GIT'ms, ET AL.,19821 " . a 9.3 In a Monte Carlo computation, a computer code is . run many times. For each rtm, the .value of each variable parameter is chosen by sapling rand =ly across the range of that parameter, appropriately weighted by. O the parmeter's assumed probability distribution. 9.4 A critique of various approaches to this problen is 'provided by (BERMAN, 1984]. 9.5 In this context, "high pressure" means several hundred psi or greater. g Note that FWRs typically operate at about .2300 psia, 'and BWRs at about 1100 psia. . 9.6 "High velocity" here means speeds of several tens of reters per second. 9.7 Core tenperatures will rise to 2000 degrees C. or higher duriry the melt O process. Zircaloy fuel cladding will melt at about 1700 degrees C., and the U0., fuel pellets will begin to dissolve in molten zircaloy at about 1900 degrees C. Typical BCS materials will melt at about 1400 degrees C. and lose much of their strength at tenperatures several hundred degrees lower (RIVARD, ET AL.,1984) . O- 9.8 he configuration of BWR contaiments is such that significant amounts of water will not acetnulate in their reactor cavities. By contrast, PWR reactor cavities can hold large ' amounts of water, and will do so in a manner peculiar to each plant and accident sequence. 9.9 Note that finely divided zirconita particles will also react vigorously O and exothermically with stean or water, yielding hydrogen. 9.10 Other experiments have shown multi-modal size distributions, with the anallest mode being about 0.5 micreaters (POWERS, ET AL.,1983] . 9.11 Releases of radioactive material into a breached containment -- .O especially releases from melt-concrete interactions -- could be significant in many accident sequences. However, a violent HPME event might lead to such dispersal of melt within the contairnent that melt-concrete interactions became insignificant. 9.12 mis apparatus is a horizontal channel 100 ft long, 8 ft high, and 6 ft O wide, along diich flames can be propagated. . 9.13 SNL's experiments are being conducted in a heated tube 17 inches in : diameter and 43 ft long. Earlier experiments intended to establish the O : 1 J
3 V 9-28 j hydrogen concentration limits for detonations were conducted in smaller i tubes. It may be that bomdary layer effects were significant for these smaller. tubes. { 3.- 9.14 For. a Large Dry IWR contalment, the period for natural vibration involving syumetrical expansion of the structure is of the order of 50 milliseconds [SHERMAN, EP AL.,1980] . Detonations in such a contalment [ are expected to generate pressure pulses at the contaiment wall with a pulse duration of the order of 10 milliseconds (DELICHATSIOS, Er AL. , 1982]. mis indicates fairly efficient coupling of the detonation wave i with .the containment's vibrations, particularly with the l higher-frequency modes of vibration. Detailed sttr!!ca will be required, t however, to predict the interaction of particular pressure pulses and contalment structures. O 9.15 Note that the steady-state contairment pressure after a detonation will be the see as after a deflagration with the same initial conditions [ j (see Figure 9-5) . 9.16 @e asstned accident follows a station blackout Sequence, featuring total failure of AC power and of feedwater to the steam generators. g Hydrogen ignition can occur without igniters. Indeed, a " spontaneous" 9.17 hydrogen deflagration (perhaps initiated by an electrical spark or a hot surface) occurred in the containment of tree Mile Island thit 2 (a Large Dry Containment PWR) during its 1979 accident. O. 9.18 For illustration, the Ginna plant (a 1520 MWt Westinghouse IWR) operates with an RCS pressure of 2200 psig and a secondary-side pressure of 755 i psig. %us the pressure on the primary side of the steam generator tubes is 1445 psi greater than on the secondary side. We four secondary-side safety valves per steam generator -- which vent directly to the atmosphere -- are set to open at 1985 psig or 1140 psig [NRC, .O 1982). s' 8 9.19 We Ginna plant has 3260 U-shaped tubes in epch of its two steam generators. We interfacing area is 44,000 ft per stem generator. Each tube has an outside diameter of about 22 nm and a wall thickness of about i um [NRC,1982] . O 9.20 If an ICS breach downstream of the stem generator is not present, convective circulation could be inhibited by the presence of water in the cold leg (the " loop seal" ef fect) . There are, however, opportunities for such a breach -- for instance a failed RCS ptnp seal. O 9.21 were will of course be a limited time between the carmenceent of core , melt and depressurization of the RCS by failure of the vessel's bottom i head. For a station blackout sequence at the Surry plant, this interval
- has been estimated to be 35-39 minutes [GIESEKE, Er AL.,1984d] .
9.22 Recent experience at Millstone Unit 2 is instructive. After a new O ch mical process had been used to remove sludge from the steam generators, thinning of tube walls by 40 percent or more was observed. r l. I i f it
0-O
@ 9-29 8 \'
I. j Eddy current testing prior to the cleaning had failed to detect much of this danage [RYAN,1985] . O'
- l. 9.23 In a recent letter from an NRC official to this author, the relevant current NRC progran was described. %e effects of convective and l fission product heating on the integrity of the BCS boundary have been under investigation since early 1984. Calculations are under way to
[ evaluate fission product heating on RCS surfaces (including steam p generator tubes) for the Seabrook plant [SILBERBERG,1985] . 9.24 %e ' calculations assuned a station blackout accident at the Surry PWR. l 9.25 For slow boiling, dissolved and suspended materials would mostly renain i behind as the water evaporated. In the present case, violent boiling would occur, and droplets of water would be entrained by the rapid flow {O j of steam and air towards the contairrnent breach. .j . I O O 1 O e O 1 i
- O-1 l
\
O !O
i i l h: 9-30
\
l I 9.8 References for Chapter Nine BENJAMIN, EP AL.,1979 A.S. Benjanin, D.J. McCloskey, D.A. Powers, and S.A. Dupree, Spent Fuel Heatup Following Loss of Water During Storage, .Sandia National Laboratories (Albuquerque, New Nxico) , NUREG/CR-0649, March 1979. BERGERON & WILLIAMS,1984 O- K.D. Bergeron, . and D.C. Williams, "Contain Calculations of Containment ' Ioading of Dry WRs," Sandia National Iaboratories (Albuquerque,. New Mexico) , in Proceedings of the Second workshop on Contairment Integrity, U.S. Nuclear Regulatory Carmission (Washirgton, D.C.) , NUREG/CP-0056, August 1984, pp. 115-128. BERMAN, SWENSON & WICKETT,1984 M. Berman, D.V. Swenson, and A.J. Wickett, An Uncertainty Study of PWR Steam Explosions, Sandia National Laboratories ( Albuquergm, New Mexico) , NUREG/CR-3369, May 1984. BERMAN, 1984 O M. Berman, "miten-Core Coolant Interactions Program," Sandia National Laboratories (Albuquerque, New Mexico) , paper presented at the NRC's 12th Water Reactor Safety Research Information Meeting, Gaithersburg, Maryland, 22-26 0:tober 1984. CAMP, ET AL., 1983 O A.L. Camp, J.C. Cunmings, M.P. Sherman, C.F. Kupiec, R.J. Healy, J.S. Caplan, J.R. Sandhop, and J.H. Saunders, Light Water Reactor Hydrogen Manual, Sandia National Laboratories (Albuquerque, New Mexico) , NUREG/CR-2726, August 1983. D E CHATSIOS, ET AL., 1982 O M.A.. Delichatsios, M.B. Genadry, and M.N. Fardis, Calculations of Hydrogen Detonations in Nuclear Containments by the Random Choice Method, Massachusetts Institute of Technology (Cambridge, Massachusetts) , NL' REG /CR-2897, September 1982. GIESEKE, EP AL., 1984a-e O J. A. Gieseke, P. Cybulskis, R.S. Denning, M.R. Kuhlman, K.W. Ime, and H. Chen, Radionuclide Release Under Specific LWR Accident Conditions, Battelle Colunbus Laboratories (Colunbus, Ohio), BMI-2104, Vols. II-VI, draft, July 1984. GITTUS, ET AL., 1982 J.H. Gittus, et al., PWR Degraded Core Analysis, Springfields Nuclear Power Development Iaboratories, UK Atanic Diergy Authority, ND-R-610 (S) , April 1982. J HASKIN, BDIR & CAMP,1984 F.E. Haskin, V.L. Behr, and A.L. Camp, "HECTR Results for Ice-Condenser O Contairment Standard Problen," Sandia National Iaboratories ( Albuquerque, New Mex ico) , in Proceedings of the Second Workshop on Containment O
O 9-31 { Integrity, U.S. Nuclear Regulatory Commission (Washington, D.C .) , g NUREG/CP-0056, August 1984, pp. 145-161. HASKIN, ET AL., 1984
- F.E. Haskin, V.L. Behr, and L.N. Snith, "embustion-Induced toads in Large Dry IMR Containnents ," Sandia National Laboratories (Albuquerque, New Mexico) , in Proceedings of the Second Workshop on Containtent Integrity, U.S. Nuclear Regulatory Conmission (Washington, D.C.) , NUREG/CP-0056, g' August 1984, pp. 129-143.
LIPINSKI, ET AL. ,1985 R.J. Lipinski, D.R. Bradley, J.E. Brockmann, J.M. Griesmeyer, C.D. Leigh, K.K. Murata, D.A. Ebwers, J.B. Rivard, A.R. Taig, J. Tills, and D.C. Williams, Uncertainty in Radionuclide Release under Specific LWR Accident g
' Conditions, Sandia National Laboratories (Albuquerque, New Mexico) ,
SAND 84-0410, Volune II, draft, February 1985. NOURBAKHSH, ET AL.,1984 H.P. Nourbakhsh, Chien-Hsiung Lee, and T.G. Theofanous, Natural Circulation Phenomena and Primary System Failure in Station Blackout q'- Accidents, Purdue University (West Lafayette, Indiana) , paper presented at the 6th Information Exchange Meeting on Debris Coolability, UCLA, 7-9 Novenber 1984. NRC, 1982 U.S. Nuclear Regulatory Comnission, NRC Report on the January 25, 1982 O Steam Generator Tube Rupture at R.E. Ginna Nuclear Power Plant, U.S. Nuclear Regulatory Comnission (Washington, D.C.) , NUREG-0909, April 1982. PISANO, ET AL., 1984 N.A. Pisano, F. Best, A.S. Benjanin, and K.T. Stalker, The Potential for Propagation of a Self-Sustaining Zirconium Oxidation Following Loss of O Water in a Spent Fuel Storage Pool, Sandia National Laboratories (Albuquerque, New mexico) and Massachusetts Institute of Technology (Cambridge, Massachusetts) , pre-publication draft report prepared for the U.S. Nuclear Regulatory Ccmnission, January 1984. PASNY & CONSOLIDATED EDISON, 1982
- O Power Authority of the State of New York, and Consolidated Blison Co. of New York, Indian Point Probabilistic Safety Study, New York, New York, March 1982, Appendix 3.4.6.
POWERS, ET AL. ,1983 D. A. Powers, J.E. Brockmann, D.R. Bradley, and W.W. Tarbell, "'Ihe Role of O Ex-Vessel Interactions in Deteunining the Severe Reactor Accident Source
'Ibrm for Fission Products," Sandia National Laboratories (Albuquerque, New Mexico) , paper presented at the ANS/ ENS International Meeting on Light Water Reactor Severe Accident Evaluation, Cambridge, MA, 28 August to 1 Septenber 1983.
- O REHM, 1984 Internal NRC menorandun fran T.A. Rehn to NRC Cortmissioners,
Subject:
" Weekly Infonnation Peport (to NRC Carmissioners) , Week mding 20 April 1984," mclosure E, page 3.
,O
q v 9-32 RIVARD, ET AL. ,1984 J.B. Rivard, V.L. Behr, R.G. Easterling, J.M. Griesmeyer, F.E. Haskin, O S.W. Hatch, A.M. Kolaczkowski, R.J. Lipinski, M.P. Sherman, A.R. Taig, and A.J. Wickett, Identification of Severe Accident Uncertainties, Sandia National Laboratories (Albuquerque, New Mexico) , NUREG/CR-3440, September 1984. RYAN, 1985 O M.L. Ryan, " Millstone-2 Steam Generator Cleaning Unmasks Unexpected Defects," Nucleonics Week, 25 April 1985, p. 6. SHARMA, ET AL., 1985 Sharma, S., Y.K. Wang, and M. Reich, Ultimate Pressure Capacity of Reinf reed nd Prestressed Concrete Contairrtents, Brookhaven National O Laboratory (Upton, New York) , NUREG/CR-4149, May 1985. SHERMAN, ET AL.1980 M.P. Sherman, M. Berman, J.C. Ctzimings, G.W. Perkins, D.A. Powers, P.O. Bieniarz, and 0.R. Green, The Behavior of Hydrocen During Accidents in Light Water Reactors, Sandia National Laboratories ( Albuquerque, New O Mexico) , NUREG/CR-1561, August 1980. SHERMAN, 1984 M.P. Sherman, " Hydrogen Flame Acceleration and Detonation," Sandia National Laboratories (Albuquerque, New Mexico) , paper presented at the NRC's 12th Water Reactor Safety Research Information Meeting, O Gaithersburg, Maryland, 22-26 0:tober 1984. SILBERBERG, 1984 Letter from M. Silberberg ( Assistant Director, Accident Source Term Program Office, Office of Nuclear Regulatory Research, U.S. Nuclear Regulat ry Ccximission) to Gordon 'Ihcapson, 4 March 1985. O TARBELL, 1983 W. Tarbell, Initial Test Results - SPIT 17, Memo for the Record, K-Tech Corporation, Sandia National Laboratories (Albuquerque, New Mexico) , 15 Novenber 1983. O 7333gtt, g7 3t,, 19g4 W.W. Tarbell, M. Pilch, and J. Brockmann, " Behavior of Core Debris Ejected frcm a Pressurized Vessel Into Scaled Peactor Cavities," Sandia National Laboratories (Albuquerque, New Mexico) , paper presented at the NRC's 12th Water Reactor Safety Ibsearch Information Meeting, Gaithersburg, Maryland, 22-26 Octok r 1984. O O
- O
D-
~
3 .. 9-33 O TABLE 9-1 4, I j Estimated Ntznber of Outcomes of Steam Explosions During 10,000 Core Melts, Based on a Monte Carlo Computation 0 at SNL, for Various Pararteter Ranges outcome of ' Steam Explosion Full Range Lower Third Middle Third High Third O Failure of Vessel Bottom 2017 0 2126 10,000 Separation of Vessel Head: Velocity over 50 Ws 460 0 1 9,987 Separation of Vessel Head:
- O Velocity over 90 Ws 267 0 0 9,958 Notes and Sources:
- O (a) t e results shown are from:-Berman, Swenson and Wickett, 1985.
4 (b) We five major paraneters used in this calculation were: i g- -
- molten fraction of core t
- pour dianeter
- pour length
- conversion ratio
- Condensed-phase fraction of slug.
- 9 1
l l
- O.
- o
O 9-34 ^ TABLE- 9-2
<g Estimated Effect of High Pressure Melt Ejection on Contairstent Atmosphere of Zion Plant: Three Cases l0 Peak Contairment Peak Containn Case Pressure (psia) Temperature (gnt Q
No direct heating of atmosphere. Fast V quench of .100% of core by water in cavity. (Note c.) 87 150 Direct heating of atmosphere by , 100% of core. No oxidation. 128 600 O Direct heating of atmosphere by 90% of core. Fast quench of 10% of core by water in cavity. Oxidation in atmosphere of 90% of- netal- (Zr , Fe) in the core at O tine of RCS failure. (Notes e and d.) 202 1970 Notes and Sources: O (a) Estimates are from: Bergeron and Williams, 1984. (b) Contairment pressure and tenperature before RCS failure are estimated to be,43 psia and 120 C. (c) Local pressure pulses due to steam explosions in the cavity are not
- O modeled here.
(d) Note that 50% of the initial Zr inventory of the core is estimated to be oxidized prior to RCS failure. (e) One estimate of the failure pressure for the Zion contairment is 149 psia l .O [GIESEKE, ET AL. , 1984). Another, probably more accurate, estimate is 126 psia - [SHARMA, ET AL. , 1985] . l l l l 1 i i l .O
O 9-35 FICURE 9-1 O i,, O i8
,jsh' gRUST,SINTEFIDRlF.3LE h 7 ;7,, ;g fj C g I ., ,,
4 /8 FRACTURED FUEL,-Zr02 f,;.N N ,., , J INTACT FUEL RODS ? O -
; \f;l.: *., . .J I t ,
y *
*)
Y, ' - ,t j, r - ... , 1, s . l;f.
' - I O Q. , .. tfLT l
{fl h. . [l.'lll j,! , x, m . tiELT FLOW ___ Qyy l@ Tl Icj $X - INio WATER p y .O (j$ gs ;f :
' LOWER CORE PLATE Ns / Ns / ' DIFFUSER PLATE ' ' LOWER SUPPORT PLATE O sN WATER gs o ,
i O l 1 I i
\
O J Possible Mode of Melt Flow Into Base of Vessel l [ Source: Berman, Swenson and Wickett, 1984] O l 1 !o i I
9-36 g 4 PAOLillJ f UEL FIGURE 9-2 1 o 7 1 Q~ ?e:En-:
. -e x _
- U.'t
,- t: w.7m L4 i
4 #p'? I 'y I f \ \ s '1 0 f S.2 j\ \ (/& f % cr Jt3 D .- l j WATER FUEL WATER O' INITIAL CONDITION: STAGE 1, COARSE MIXING (1) (ii) MOLTEN FUEL AND COOLANT ARE SLOW HEAT TRANSFER, NO PRESSURE INCREASE SEPARATE O FINE MIXTURE O WATER WATER I i grryg u. . .- l 1 y .J I i p g,- r O a, 5. pig,1,,-/
....=,. ,*j 3 4 Jf y)\ if ,/6 M' %s:f.f \g r/,.e g .
y* </4 .. h)kN s t <M[ 1h!5j.M.$/ M M7 PRESSURE WAVE j O LOCAL PRESSURE FUEL FUEL (iii) STAGE 2, TRIGGER PROCESS (iv) STAGE 3, PROPAGATION I
. LOCAL PRESSURE, E.G. FROM IMPACT PRESSURE WAVE FRAGMENTS FUEL VERY RAPIDLY; OR ENTRAPMENT HEAT TRANSFER FROM FINE FRAGMENTS IS VERY O g3 pro
-O Simplified Illustration of Stages of a Steam Explosion [ Source: Gittus et al, 1982] t l 1
. -~u - - . . __ -. .., . - - _ . e4 s ,
f E
'n)+
9 F I C U'R E 9-3 p, ,v-5 M. i 6 t
) 't / . hl / 'n / / '/ Y /
i -Q'
- /
' / / / / / / . / / / / - . . . .
j~ e,q 4 q /) . waua e F4.ac4.n.. ....s. u..eeu\,e g-(g - v; - -uny/ v u/ 2O L ~~ 'y'- uoueN
/ Cone , b M AT ERI A'.$
f h ?'sn NAf g'y n, fl j -tNST AUMENT ATION
.I . Il I inats on t rwro9 .
p 'l U..** \"l' nooonives
.:'). . V .:. . . ./ .- .] . BUSBLES F AOu g,y , ;.. . .** .*~ ORoPLETS rnou cases cou~s . . - s;... . : * . wEaEn sazAxua a c:: cc :. - . * - * "::p
- i SOLUTION . */l* . , ** ,,*.
- y , .*.
- AEROSOLS
** **** . . * */ FRou -*-. - vAvon 1 -
CONDeN S A TION lQ-4 4
- O 1
s
- O l e
t Possible Mode of High Pressure Melt Ejection 1 -9.
- (Source: Rivard et al, 1984]
4 4
- O 4
- + + c , >-, , , - - .....,,---..-.__e m., ,,,,,,,_.m.-_..,,.mw.,,.-.,,.,_,m,..% y._..-__.. ._ ..-.---w,.,-e- - ,,m,-m.y_ ~ , . . , . . - . ~ , . , , . . , , - - ,
m. .J 9-38 F. L Q U R E 9 . .' 4
- I" i a i e i a s s :
3 8WR-MARK I & 11 V = 300,000 f t 3 SWR-MARK lli, V e 1,500,d00 f t . 90 . PWR-ICE CONDENSER, V = 1,250,000 f t v ALSO 3 80 PWR-SUS- ATMOSPHERIC, V = 1.850,000 f t
=
3 -
- PWR - D R Y, V = 2,000,000 f t 70 g
i
- 3. -
z '60 - w .
- - - - - E z
3 Q V 2 -
~
0 ' 50 - E i z b 40 - E 30 - ~O . 20 -
- O -
10 - 3 PWR- DR Y, V = 3.500.000 f t f t f f 1 f f f f 60 70 80 90 100 O iO 20 30 40 s0
.O MET AL WATER RE ACTION (percent) e
~O Approximate Hydrogen Concentrations in Various Containments, for Differing Amounts of Zirconium-Water Reaction t' [ Source: Camp et al, 1983)
- O 9 9
i
~- -
FIGURE 9-5 i i i i i i i i i i i 2150 310 - 1950 D- 290 - 270 - 1750 0 m 250 - D @U - '
- o. _
i s, AO 1550 y 230 - D _ - 210 - ! @3 3 [ _ [ 5 5 go*M - 1350 m o 190 - b fNERTED ' ',300 _ S E -
/ / -
1150 O 170 - o O
~
l - 9 150 -/ O - 950 O x 130 j/ - INI IAL CONDITIONS - 750 110 -
# STEAM PARTIAL PRESSURE 0 __ ~
CORRESPONDS TO SATURATION 90 - AT TEMPERATURE T3 550 AIR PARTIAL PRESSURE 70 -
~
O _ CORRESPONDS TO 14.7 psia (101kPa) SATURATED AT 300K (540'R) - 350 50 - 4 8 12 16 20 24 28 32 36 40 44 48 52 HYDROGEN MOLE FRACTION, PERCENT 'O eressure voitowing Aatabatic, constant-volume nydrogen combustion (certagration), for Various Initini Conditions (Source: Camp et al, 1983] , O
i ! 30 , , , , , , , , , 38 i INITIAL CONDITIONS P3/P) T j = 298 K = 536*R i 24 - 36 i P 3 = 101 kPa E 14.7 psia ~ 100% RELATIVE HUMIDITY c. 34 1 20 - 9 ' REFLECTED SHOCK W AVE Q 18 ,P2 P3 - 32
~ w , - D '8 i m -
j
~
30 l 6m 16 - p2 fp1 m e 1 1 m g ; 9
- i 28 kC 14 - ---- -
i i w DETONATION WAVE I C I o m m 12 P2
*P3 - 26 i W '
1 C CL j ' ' 24 10 54 58 18 22 26 30 34 38 42 46 50 l HYDROGEN CONCENTRATION, VOLUME PERCENT Theoretical Detonation Pressures, for Various liydrogen Concentrations [ Source: Camp et al, 1983]
., 9-41
.s_)
F I C U R'E 9-7 O nc . i e i i i
- y
.n co m
4 .
"8~ - i
!O e i L3 $_ !. m : i n : o8 L-n CL. s'. .J a p- \ '. .. E.. ~
. d i h, :., -
o i.! i .: .. - g g_ e ' - . . .,
.a- 1 Spray Train .'*.... ,' " *_-
q .......... 4 Fan Coolers y ]gQ- ----- 3 Spray Trains
' O NO 120 1h0 2k0 300 3b0 420 4bo 5d0 600 Time (min) o i
i
/
Estimated Containment Pressure During a Severe Accident at the Zion Plant, with Hydrogen Combustion Initiated by Delayed Introduction of Containment f Cooling, for 3 Cases C Notes and Sources (a) This figure from: Haskin, Behr and Smith, 1984. (b) The assumed accident follows a TMLB sequence, initiated by loss of Ac power. It is assumed that power is restored to containment cooling O systems after 6 hours. 9 9 I, O g .__
O 9-42 A P P E N D I X -T O CHAPTER NINE -O Some Steam Explosion' Incidents te incidents listed here, which exmplify the record of industrial experience of stem explosions, were described in Appendix B to Appendix VIII of the Reactor Safety Study (WASH-1400) . Our account is an edited version of that description. tree industries are represented: the metals, paper, and nuclear reactor industries. n
- (i) Metals Industry Explosive accidents are infrequent in the metals industry but when th2y
' do occur, the destrtx: tion is severe.
- O Mallory-Sharon Incident, 1954 A titanim arc-melting furnace, which was water-cooled, exploded at a plant in Ohio. Nine injuries, incitriing four fatalities, were sustained and
'O property damage was $30,000. We explosion was believed to have resulted from water entering the melting crucible. Reynolds Aluminum Incident, 1958 An altninm-water explosion occurred in Illinois, involving some 46 injuries, 6 fatalities and approximately $1,000,000 in property dmage. We ,j explosion was said to " rock a 25 mile area." Wet scrap metal was being loaded
- into a furnace when the explosion was triggered.
'O Ouebec Foundry Incident
%e accident occurred in a foundry building of approximately 18 million ft volume. One hundred pounds of molten steel fell into a shallow trough a containing about 78 gallons of water. We resulting explosion injured mill personnel (one fatally) and caused $150,000 daage to the foundry building including cracking of a 20-inch concrete floor, breaking of 6000 panes of glass, and structural daage to the walls and ceilings. Dmage was also
'O incurred by another strtx:ture separated some 75 yards frcm the foundry i building. Wis accident is one of the better doctmented incidents. L o v - - - - - - - . _- _ , _ _ _ _ _ _ _ _ _ _ _ _ _
! 9-43 !g b.
h'estern Foundries Incident,1966 hhile 3000 pounds of molten steel was being poured from an electric {n furnace into a tile-lined ladle, a cable broke and the hot steel dropped into g We result was a violent explosion that injured three2 a water filled pit. b 2 hole in the roof of a building of scme 12,000 ft h workers and tore a 600 ft 1 floor area. We explosion was heard some 3 miles from the foundry. do
') Armco Steel Incident,1967_
f An explosion occurred when molten steel fell on " damp" ground. A ladle g containing scme 30 tons of molten steel had been elevated scme 40 feet when g the ladle fell. Injuries sustained by 30inworkers incitried 6 fatalities. the porous ground to trigger g Evidently, sufficient moisture was present We small-scale explosions that showred molten steel over a wide area. f injuries were attributed primarily to burns. 5 h, East German Slag Incident An East German article which appeared in 1959 discussed a number of slag-
.g in German open-hearth steel mills. Two water explosions that had occurred k accidents were discussed in which explosions resulted frcm spraying water onto One of the explosions resulted in a fatality molten slag in open slag pits. We g and a number of other injuries.
Severe strtx:tural damage was also noted. jd Both explosions were attributed to excess m second explosion was less severe. j the slag surface passing down throtgh cracks to the hot molten h water on material below. A third explosion, which killed one man, occurred when a slag We pot was placed on a slag bed that had been previously sprayed with f water.
- explosion was attributed to the heavy slag pot causing cracks in the sur ace slag bed, whereupon excess water on the surface entered these f the hot O Other explosions briefly described incit:3e one arising from rainwater cracks.
leaking throtsh an unsealed roof over a slag bed, and two instances ll of explosions resultire when molten slag was poured into dtmp cars that ha amounts of water in the bottan. gi{ k British Slag Incident,1964
- 4. .sk An explosion occurred in a British steel mill when a ladle being used to Wen tap a blast furnace was sprayed with lime water and returned to service.
/W, u .r. it was next used, the ladle exploded when it was about three-fourths ful ,.qi{!
W
4r a 9-44 h3 slag (12 to 14 tons) . Damage to the structure and injuries to personnel were reported. D (ii) Paper Industry i . me paper industry also experiences explosions, more frequently than in t Explosions occur when the metals industry, but of a less destructive nature. k paper melt (mostly fused soditra carbonne with a few percent of soditrn sulfide, sodium chloride, and minor ingredien'ts) is quenched in large containers of " green liquor". Also, explosions frequently occur when boiler k N tubes in waste-heat boilers fueled by " black liquor" fail, and water is b Wese explosions occur with injected into hot molten melt and black liquor. considerable destruction to the furnace and plant facilities. J ' (iii) Nuclear Reactor Industry O Explosive vapor formation has been observed in nuclear reactors when hot, 4' molten core materials have cme in contact with water. O^ Canadian NRX Reactor, 1952 At Chalk River , Ontario, during a low-power experiment, a nuclear fg less excursion was experienced. Althotxjh the duration of the incident was than 62 seconds, the daage was sufficient to result in contamination of the was the principal facility. A reaction between uranitzn and steam (or water) cause of daage. Borax I Reactor, 1954 O' During safety studies at the National Reactor Testing station in Idaho, the Dorax I reactor was deliberately subjected to a potentially darnagits power excursion. @is excursion lasted approximately 30 milliseconds, and produced a We power peak power of 19 Gi with a total energy release of 135 tW-sec. We reactor tank (1/2-inch steel) O'- excursion melted most of the fuel elements. I was ruptured by the pressure (probably in excess of 10,000 psi) resulting from We sound of the explosion at the g a reaction between molten metal and water. control station (1/2 mile away) was canparable to that from 1-2 pounds of 40 [. '; percent dynamite. K
9-45 3 h SPERT l-D Reactor During the final test of the destructive test program with the SPERT l-D core, damaging pressure generation was observed. Pressure transducers O recorded the generation of a pressure pulse (larger than 3000 psi) which caused the destruction of the core. We pressure pulse occurred sane 15 milliseconds af ter initiation of the power excursion. Wis excursion rapidly overheated the fuel plates, leadirg to melting of the metal fuel and the 3 cladding. Af ter the transient, much of the fuel that had been molten was fourx3 dispersed in the coolant. ' SL-1 Reactor, 1961 O A ntelear excursion occurred in the SL-1 reactor in Idaho, leading to a total energy release of approximately 130 MN-sec. Of this, 50 W-sec was produced in the outer fuel elenents of the core. %is portion of the energy release was slowly transferred to the water coolant over a 2 second period, O and no melting of these fuel elenents occurred. About 50-60 MW-sec of the total energy was released by 12 heavily damaged inner fuel elanents to the water coolant in less than 30 milliseconds. Wis prmpt energy release resulted in rapid steau formation within the core, which accelerated the water O above the core and produced a water harraner that hit the pressure vessel lid. The vessel, weighing about 30,000 lbs with its internals, sheared its connecting piping and was lifted approximately 9 feet into the air by the manentun transferred fran the water hattner. Calculations of the mechanical O deformation of the vessel indicate that about 12 percent of the prompt energy release, or 4.7 percent of the total nuclear release, was converted into mechanical energy. O General Notes In each instance, under differing circunstances, a hot molten material fell, dropped, or spewed into a mass of cooler liquid and destructive pressure O generation resulted. We canplex mechanisns triggering this type of reaction are not well understood. It may be noted that all of the above reactor tests and incidents O involved plate-type fuel elonents consisting of uranium-aluninun alloy fuel clad in aluninun. Wese are substantially different fran the uraniun oxide}}