ML20207Q231

Forwards Comments on BNL Draft Rept Re Emergency Planning Sensitivity Study.Comments Substantially Unchanged from Comments During 870114 Meeting W/Bnl & NRC in Bethesda,Md. Responses to Listed Items Revised Per 870114 Meeting

ML20207Q231
Person / Time
Site:	Seabrook
Issue date:	01/20/1987
From:	George Thomas PUBLIC SERVICE CO. OF NEW HAMPSHIRE
To:	Nerses V Office of Nuclear Reactor Regulation
References
CON-FIN-A-3852 NYN-87-002, NUDOCS 8701270047
Download: ML20207Q231 (45)

Text

,,

i 4

'~

George S. Thomas Vice President-Nuclear Production January 20, 1987 h Service of New Hampshire NYN-87-002 New Hampshire Yankee Division United States Nuclear Regulatory Commission Washington, DC 20555 Attention:

Mr. Victor Nerses, Project Manager PWR Project Directorate No. 5 Division of PWR Licensing-A

References:

(a) Facility Operating License NPF-56, Construction Permit CFrR-136, Docket No.s 50-443 and 50-444 (b) USNRC Letter, dated December 8,1986, "Trans-mission of Brookhaven National Laboratory Draf t Report of The Seabrook Emergency Planning Sensitivity Study", V. Nerses to R. J. Harrison

Subject:

Comments on Draft Report

Dear Sir:

Enclosed please find New Hampshire Yankee's (NHY's) comments con-cerning the Brookhaven National Laboratory (BNL) Draft Report A-3852

[ Reference (b)].

These comments are substantially the same on those provided to the Staff and BNL at the January 14, 1987, meeting held in Bethesda to discuss this draf t report. The only changes of substance (i.e.,

non-typographical corrections) are contained in the responses to items 1.1, 1.3, 2.7, 3.5, 4.0 and 5.0.

NHY modified these comments as a result of the discussions at this meeting.

If the Staf f requires any further clarification of these comments, do not hesitate to contact the NHY Bethesda Licensing Of fice (Mr. R. E.

Sweeney) at (301) 656-6100.

Very truly yours, 12{0047870120 DOCK 0500 3

ch M

e S. Thomas Enclosure cc:

Atomic Safety and Licensing Board Service List (Of f site EP) 00 (l

P.O. Box 300. Seabrook, NH 03874.' Telephone (603) 474-9574

l o

Q ASLB Service List (Off-Site EP)

Helen Hoyt, Chairperson Carol S. Sneider, Esquire Atomic Safety and Licensing Board Panel Assistant Attorney General U.S. Nuclear Regulatory Commission Department of the Attorney General East West Towera Building One Ashburton Place,19th Floor 4350 East West Highway Boston, MA 02108 Bethesda, MD 20814 Senator Gordon J. Humphrey*

Dr. Emmeth A. Luebke U.S. Senate Atomic Safety and Licensing Board Panel Washington, DC 20510 U.S. Nuclear Regulatory Commission (ATTN: Tom Burack)

East West Towers Building 4350 East West Highway Richard A. Hampe, Esq.

Bethesda, MD 20814 Hampe and McNicholas 35 Pleasant Street Dr. Jerry Harbour Concord, NH 03301 Atomic Safety and Licensing Board Panel U.S. Nuclear Regulatory Commission Thomas F. Powers, III East West Towers Building Town Manager 4350 East West Highway Town of Exeter Bethesda, MD 20814 10 Front Street Exeter, NH 03833 Diane Curran, Esquire Harmon & Weiss Brentwood Board of Selectmen 2001 S. Street, N.W.

RFD Dalton Road Suite 430 Brentwood, NH 03833 Washington, D.C.

20009 Peter J. Mathews, Mayor Sherwin E. Turk, Esq.

City Hall Office of the Executive Legal Director Newburyport, MA 01950 U.S. Nuclear Regulatory Commission Tenth Floor l

7735 Old Georgetown Road Calvin A. Canney Bethesda, MD 20814 City Manager l

City Hall Robert A. Backus, Esquire 126 Daniel Street 116 Lowell Street Portsmouth, NH 03801 P.O. Box 516 Manchester, NH 03105 Stephen E. Merrill, Esquire Attorney General Philip Ahrens, Esquire George Dana Bisbee, Esquire Assistant Attorney General Assistant Attorney General Department of The Attorney General Office of the Attorney General Statehouse Station #6 25 Capitol Street Augusta, ME 04333 Concord, NH 03301-6397 Mrs. Sandra Gavutis Mr. J. P. Nadeau Chairman, Board of Selectmen Selectmen's Office l

RFD 1 - Box 1154 10 Central Road l

Kennsington, NH 03827 Rye, NH 03870

• Letter of Transmittal Only l

4 ASLB S2rvica Lint (Off-Sits EP)

(Continued)

Mr. Angie Machiros Chairman of the Board of Selectmen Town of Newbury Newbury, MA 01950

'Mr. William S. Lord Board of Selectmen Town Hall - Friend Street Amesbury, MA 01913 Senator Gordon J. Humphrey*

1 Pillsbury Street Concord, NH 03301 (ATTN: Herb Boynton)

H. Joseph Flynn, Esquire Office of General Counsel Federal Emergency Management Agency 500 C Street, SW Washington, DC 20472 Paul McEachern, Esquire Matthew T. Brock, Esquire Shaines & McEachern 25 Maplewood Avenue P.O. Box 360 Portsmouth, NH 03801 Gary W. Holmes, Esq.

Holmes & Ells 47 Winnacunnet Road Hampton, NH 03842 Mr. Ed Thomas FEMA Region I 442 John W. McCormack PO & Courthouse Boston, MA 02109 Robert Carrigg Town Office Atlantic Avenue North Hampton, NH 03862 Judith H. Mizner*

Silvergate, Gertner, Baker, Fine, Good & Mizner 88 Broad Street Boston, MA 02110

• Letter of Transmittal Only

'l

' l ENCLOSURE TO NYN-87-002 CO)MENTS TO BNL DRAFT REPORT A-3852 SEABROOK STATION L

e g{J Our Comments on the BNL Draf t Report A-3852 are summarized in this report.

The comments are organized by the following major sections:

1.

General Comments 2.

Interfacing Systems LOCA 3.

Shutdown Events 4.

Steam Generator Tube Rupture 5.

Direct Heating 6.

Containment Structure 1.0

GENERAL COMMENT

S The general comments in this section are aimed primarily at the BNL PREFACE and

SUMMARY

but also apply to the whole report, in general.

Table 1-1 calculates the Safety Goal for Seabrook Station using BNL's stacked conservative sensitivities. As shown the Safety Goal is met with a 1 mile evacuation model.

1.1 Inconclusiveness of BNL report The pref ace and summary tend to understate BNLs actual review.

These sections should be written to be more consistent with the actual contents of the BNL Report to more accurately reflect the scope of the review ef fort.

The implication is that only PLG-0465 was reviewed. As demon-l strated throughout the BNL report, PLG-0432 and the relevant parts of the original SSPSA were reviewed and evaluated.

PLG-0465 is nothing more than a sensitivity study showing the impact un results and conclusion using the conservative WASH-1400 methodology source terms. BNL used the PLG-0465 source terms and used PLG-0465 for sensitivity studies (risk comparisons).

The report leaves the impression that additional validation may be necessary. We believe that adequate confidence for decision making can be achieved by emphasizing review of those areas that affect early releases. They are the following:

1.

Containment Bypass such as interfacing LOCAs 2.

External Events with Containment Damage 3.

Containment Isolation System 4.

Containuent Structural Integrity WASH-1400 plus the NRC's 11 years of PRA experience and know-ledge and the Seabrook evaluation of these areas should provide sufficient confidence in our results. In addition, the following is considered important with regards to confidence and margins in the Seabrook results:

i 1

~

s 1.

Thara era ccquenesc in thm SSPSA thr.t irpact early relscass j_

. that were not explicity ' included in WASH-1400. This includes Airplane Crashes, Turbine Missile, aad Earthquakes which dom-inste public risk, fires, floods and reactor coolant pump seal LOCAS.

2.

The WASH-1400 PWR _ results strictly apply only to Surry at a so-called composite site. The Seabrook PSA results in PLG-0300, PLG-0432 and PLG-0465 are specific to the Seabrook plant, contain-ment and site. It should be clear upon comparison of these studies. that WASH-1400 results are not applicable to Seabrook and the Seabrook study more accurately reflects the risks and accident f reque ncies.

3.

A one-mile EPZ can be justified even when a conservative one percent strain criteria for containment failure (requested by BNL) and conservative WASH-1400 methodology source term are used in a stacked sensitivity case. Although single sensitivity studies are useful, continual stacking of these upper bounds is not an appropriate way to address uncertainties for decision making purposes. Also, the 10 mile EPZ is based on best estimates (50% confidence level / WASH-1400). Even af ter stacking all the sensitivity cases, the BNL results still justify 2 miles when compared with the NUREG-0396 results. We believe that this goes well beyond what is needed to establish the robustness of the conclusions of the sensitivity study.

4.

This is the ~ second review by BNL of the Original SSPSA Containment Analysis. Also, the LLNL' review of the plant model identified no-significant safety issues and their consents on areas that ef fect early release were either positive or identified conservatisms in our model. These areas include:

o Aircraft Crashes (LLNL Page 4.3-2) o Turbine Missiles (LLNL Page 4.8-2) j-o Containment Isolation (LLNL Page 3.4-96) o V Sequence (LLNL Page 3.9-4 through 3.9-6 and Page 3-21) o Seismic (LLNL Page 4.1-74)

The overall SSPSA methodology is similar to other PRAs reviewed by the NRC (Zion, Indian Point). These reviews were factored into the original SSPSA.

5.

There are a number of _ potential conservatisms in PLG-0432 that i

should be recognized.

_S.6, Conservative fragilities No credit for recovery af ter earthquake S.7_

j MOV double disc treated like CV i

Test f req uency Factor of 2 in frequency model Probability of pipe break me as S7 plus no operator action Unit 2 Turbine Missile ef fects on Unit I S2, No credit for recovery after earthquake 2

l

l

,5 1.2 Characterization of PLC-0465 as a Risk Assessment One issue that needs to be addressed in the final draft of the BNL review report, A-3852 is the difference between a risk assessment and a sensitivity study. PLG-0432 is a risk assessment and PLG-0465 is a sensitivity study. The former attempts to quantify all significant contributors to risk and to quantify all identifiable sources of uncertainty. In addition, the calculations in a risk assessment are performed very carefully to distinguish between two different kinds of probability:

the relative frequency of a random variable type and a subjective assessment of a state of knowledge or expert opinion type. The best risk assessment currently available for Seabrook is presented in PLG-0300 and updated in PLG-0432.

By contrast, PLG-0465 is a sensitivity study designed to invest-igate the singular importance of one and only one aspect of the uncertainty. That is the uncertainty in the estimation of radioactive release source terms. The curves in PLG-0465 were developed, using best estimate assumptions on all other parameters and median accident frequencies. This was done to avoid the " stacking" of uncertainties and to provide the highest degree of comparability with the results of NUREG-0396 and WASH-1400. The NUREC-0396 curves were based on median accident frequencies from WASH-1400 and risk curves that were generated using the best estimate or central estimate assumptions on source terms and consequence model parameters. Even with this approach, the direct comparisons between PLG-0465 and NUREG-0396 are conservative because of a more complete treatment of dependent events and component failure data in PLG-0465. For example, NUREG-0396 does not include the seismic events, fires, floods, other external even'ts, common cause failure, reactor-coolant pump seal LOCAs and other events that were included in PLG-0465.

Starting from the proper characterization of PLG-0465 as being based on conservative source terms and best estimate assumptions on everything else, it should not be at all suprising that the curves will increase as additional conservative assumptions are added. The BNL Report summary states that "...the risk estimates quoted from PLG-0465 at the beginning of this summary do appear to be influenced by the various sensitivity studies performed at BNL".

They are influenced, because they are generated by stacking new conservative assumptions (e.g. that there is a chance of early containment failure due to direct heating or that there is a chance of thermal failure of steam generator tubes during high pressure melts etc.) on top of a conservative source term methodology and fixing everything else.

5

D bl.

2

- A final comment we have with these sensitivity studies is the mixing together of the two dif ferent types of probabilty.

The curves in NUREG-0396, WASH-1400 and PLG-0465 only include the relative frequency type of probabilty.. All the uncertainity types of probabilty.is, isolated from the risk curves, i.e. the curves are computed as ~ conditional on a set of assumptions regarding all uncertain paramenters. By contrast, the A-3852 curvesLreflect both. For example, the curves for early containment f ailure include expert opinion _ type probabilities that uncertain phenomena could result in early containment failure. How one then compares these curves is not clear and their usefulness in decision making is questioned.

In the final report of A-3852, the proper role of. PLG-0465 and the BNL results as sensitivity analysis should be made clear.

1.3. Nature of the BNL Review Since the publication of the Reactor Safety Study, it has been generally accepted that risk assessments should be based on the most realistic set. of assumptions that can be technically justified.

This is in contrast with the traditional approach employed in deter-ministic safety analyses (e.g., the Chapter 15 FSAR analyses) which are based on a set of " stacked" (i.e., concurrent) conservative assump-tions. l An important reason for basing PRA results on realistic assumptions is to avoid misleading conclusions about risk levels and

- the principal contributors to risk.

The use of conservative assumptions in deterministic safety analyses is a way' to deal with uncertainities in the models and data used for-the calculations. An unsatisfactory characteristic of this approach is that _ the degree of overall conservatism in the final results is not quantified, nor easily controlled.

The overall degree of conservatism depends not only on the.

degree of conservatism associated with each assumption, which is seldom quantified anyway, but also on the number of such assumptions that the particular calculation happens to depend on.

For example, if a radiation dose calculation, Z, happens to depend on say, 4 variables that are uncertain, i.e., Z = Z( A, B, C, D) and Z is calculated using conservative values'of each variable such that each conservative value has a probability of.99 of being on the conservative side of the true value, j'

then the calculated Z value will have a.99999999 probability of being on the conservative side of-the true value. For this reason we find the approach of combining all sensitivi ty l-curves in Figure S.7 to establish an upper envelope risk curves to be lacking in technical meaning and an inappropriate basis for decision making for emergency planning. Such stacking of conservative assumptions was not performed in the PRA analyses

(

that provided the foundation for the decision to establish a 10 mile EPZ for all plants.

l I

4 1.

l ff There are two. approaches used in PRAs to address uncertainties:

The approach of propagating uncertainty probability distributions through the calculations to esteblish uncertainty probability distributions on the results, and the approach of performing sensitivity calculations. The former approach is a systematic way to combine all the uncertainties into the final results with the probabilistic " weight" of each point in the rangc of i

final results. This approach was followed in PLG-0432. The latter approach. is only useful to examine specific issues such as the case with PLG-0465. There is no scientific basis for combining or adding two or more sensitivity ' cases. In fact, this tends to provide misleading indications of the significance of any and all underlying issues.

f The curves used as decision criteria in PLG-0432 and reproduced as _ Figures S-2 through S-6 in the BNL report were taken from the Reactor Safety Study and NUREG-0396. Although the _ curves from these studies have ranges of unce"tainty associated with and even quantified for them, the curves themselves are based on median accident frequencies and best estimate assumptions on everything.else, including what was then the best estimate assumptions on such factors as source terms, containment strength and the progression of interf acing LOCA sequences.

It is very important to understand that WASH 1400 was based on what

'~

were then viewed as realistic assumptions on source terms and everything else and this was the basis for decision making, i.e., the decision to set the EPZ to 10 miles for all plants.

With this background, it is important to properly characterize

]

what was done by the BNL review and to contrast their approach to what was done in the Peer Review group put together by NHY to review PLG-0432 and PLG-0465. The BNL review can be accurately described as a qualitative critique of the two reports and a collection of sensitivity calculations used to explore the significance of selected issues. The BNL review appears to look at only those assumptions that could potentially drive the risk results higher. There is no evidence that identified the many conservatisms in the "best estimate" i

assumptions on NUREG-0465, which if found would support the bottom line conclusion that these studies provide a better justification for a 1-mile EPZ at Seabrook Station, than NUREG-0396 provided for 10 miles at all U.S.

sities. One such conservatism is that the PLG-0465 results include many risk contributions not included in NUREG-0396, including:

o reactor coolant pump seal LOCAs o seismic events and other external events o fires, floods and other internal plant hazards o common cause failures due to design, constuction errors, operator errors or environmental stresses o support system f aults such as loss of component cooling It is quite clear from the SSPSA results that omission of the above contrib-utors would have resulted in significant reductions in the PLG-0465 risk levels. An other f actor that must be taken into account is the fact that 5

O 5

L

/*

all the issues addressed in the BNL sensitivity curves; such as how to estimate the frequency of the V-sequence, the concern regarding possible thermal failure of SG tubes during high pressure scenarios, the issue of direct containment heating, and the question of the importance of shutdown loss of cooling events were not addressed in WASH-1400 nor NUREG-0396.

A final observation is the decision by BNL to use the PLG CRACIT results instead of the corresponding BNL CRAC2 and MACCS results in the presentation of their sensitivity results. In aldition, they added additional conservatism in the assignment of accident sequences to release categories (S1 vs S6). The only reason we can determine for this approach is that for a consistent treatment of multiple source terms BNL consistently found lower risk curves than those presented in PLG-0465. Apparently, BNL put together the worst possible combination of the PLG and BNL results to see how high the risk curves could be driven up.

Even with_all of this, the NRC safety goals are still met at between 1 and 2 miles. BNL apparently is not taking a position on where the risk curves should be, or on which side of the PLG curves they lie. This is in sharp contrast with the NHY sponsored peer review team which took a rather strong position that not only are the PLG-0465 curves conservative but so are the PLG-0432 "best estimate" curves.

i l

l l

l

[

MRC PLO-0465 (MEDIAN)

EVAC =0 EVAC =1 EVAC =2 RC FREQ 1.5 M1 POP ACUTE 30 ACUTE 30 ACUTE 30 St 1.50E-09 6.47E+03 9.26E+00 2.15E-12 9.26E+00 2.15E-12 9.26E+00 2.15E-12 S2 7.50E-M 6.47E+03 1.21E+02 1.40E-07 8.33E+00 9.65E-09 0.00E+00 0.00E+00

$6 1.50E-08 6.47E+03 3.85E+02 8.92E-10 1.93E+02 4.46E-10 2.4tE+00 5.60E-12 1.41E-07 1.01E-08 7.74E-12 PLO-0465 (MEAN)

EVAC =0 EVAC =1 EVAC-2 RC FREQ l.5 M1 POP ACUTE SO ACUTE SO ACUTE SO S1 4.00E-09 6.47E+03 9.26E+00 5.72E-12 9.26E+00 5.72E-12 9.26E+00 5.72E-12 S2 2.10E-05 6.47E+03 1.2tE+02 3.93E-07 8.33E+00 2.70E-08 0.00E+00 0.00E+00 36 6.50E-07 6.47E+03 3.85E+02 3.87E-08 1.93E+02 1.93E-06 2.41E+00 2.42E-10 4.31E-07 4.64E-08 2.48E-10 PLO-0465 +BNI.(MEAN TO PESSIMISTIC)

EVAC =0 EVAC =1 EVAC =2 RC FREQ 1.5 M1 POP ACUTE SG ACUTE SO ACUTE SO S1 5.20E-06 6.47E+03 9.26E+00 7.44E-09 9.26E+00 7.44E-09 9.26E+00 7.44E-09 S2 2.10E-05 6.47E+03 1.21E+02 3.93E-07 8.33E+00 2.70E-08 0.00E+00 0.00E+00 S6 5.60E-06 6.47E+03 3.85E+02 3.33E-07 1.93E+02 1.67E-07 2.41E+00 2.09E-09 7.33E-07 2.01E-07 9.53E-09 St = 4.0E-9(Pl.0-0465) 2.4E-6(BNI. 38 TUBE PESSIMISTIC) 1.4E-7(BNI. V SEQUENCE) 2.7E-6(BNI. DIRECT HEATING-2.?E-4*.01 )

=5.2E-6 S6 = 6.5E-7(PL6-0465) 5.0E-6(RAI 21 SHUTDOWN EVENTS-CONSERVATIVE)

=5.6E-6 l

TABLE 1-1 SAFETY 004. SPREAD SEET 7

t

~

2.0 INTERFACING' SYSTEMS LOCA BNL has a number of comments on the PLG-0432 treatment of check valve failure frequencies. These comments are presented in BNL Section 2.1 followed by a BNL reanalysis of V-sequences based on their reassessment of check valve failure frequencies. - Our responses are provided in the following subsections.

2.1 NPE Data Base BNL states that " PLG selected a particular subset of those events listed in the NPE ' data base, namely, events involving check valves at the RCS-ECCS interf ace".

This is not t rue. As stated on p. 3-17, paragraph 1 of PLG-0432:

"...only those (valves) associated with PWR, ECCS and RCS were considered the most relevant for the valves considered here, which are initially seated and testable". The data search that was performed in NPE included all the valves in these systems, not just those at the RCS-ECCS interf ace. Of the 21 leakage events reported in Table 3-8 of RMEPS (accounting for multiple events identified in several reports), the following types of failure events were indicated.

Valve Type No. of Events Accumulator Check Valves 17 ECCS/RCS Interf ace Valves 4

Total 21 It may appear' on the surf ace that only accumulator and ECCS/RCS interface valves were included in the data search. This is not the case. All reported events in the ECCS and RCS were included however in the search. The fact that other types of valves did not - produce events in the final list of screened events is in itself statistical evidence that should be accounted ror in estimating the f ailure rates.

It should be noted that the events of interest in the analysis are those that produce in excess of 1800 gpm of leakage through 1

ECCS/RCS interface valves.

In order to achieve such a leak rate, it is necessary to postulate a major rupture of the valve disc. There has never been such a failure occur in U.s reactor operating exper-ience and we are not aware of such an even,

.a any non-nuclear energy systems.

Af ter reviewing the experience with all reported check valves events on all systems in NPE we decided to focus our data analysis on all normally seated check valves in PWR ECCS systems including accumulator check valves, those at the interf ace between the ECCS and RCS and those that separate the ECCS system f rom the RWST, containment spray and containment susp.

In cases where inter-8

,.s.

facing check valves _are numbered as belonging to a different system (e.g.,. containment spray) they were excluded. All these valves are

~

-normally seated and while _ they have different loading conditions, they have ' correspondingly different design criteria. Except for variations in pressure differentials across the valve discs and internal pressures, the environmental conditions of all PWR ECCS check valves are very similar (e.g. temperature, boric acid, testing requirements). Since all the closed valves in the ECCS are designed to the ASME code and are safety grade there are. inherent margins of safety for the structural.

integrity of each valve to remain intact in its normal and even ' severe abnormal environments. Therefore, we are not. talking about events in which the valve discs rupture due to excessive. loads, we're talking about spontaneous ruptures of the material due to unforseen and undetected failure mechanisms (e.g. design flaws, material flaws, corrosion, fatigue, etc.) which reduce the capacity of the valve discs to survive their normal environment. Many believe that such failures are essentially impossible.

In view of these _ considerations, we maintain that the full-PWR ECCS check valve population is an appropriate data base for this analysis. However, under no circumstances do we think it's appro-priate to limit the data analysis to accumulator check valves. If any subset of the EECS check valves should be used, it would be the ones being analyzed:

the ECCS/RCS interface valves.

In the BNL analysis described on page 2-7 other valve types appear to be included in the BNL count of failure events but exposure time is limited to the data on accumulator check valves. The BNL survey

~

identified about twice as many leak events (35 vs 17) and the exposure time is being limited to accumulator check valves only, which was about one forth of that used in PLG-0432 (2.3 x 107 hrs vs.- 1.1 x 108 hrs). All other aspects of the PLG analysis were held fixed. As a result BNL derived a failure frequency vs leak size curve roughly 8 times greater than analyzed in PLG-0432.

In our opinion, BNL has come up with a more appropriate small leak frequency for accumulator valves, but has not contributed to a greater understanding of the frequency of interfacing check valve disc ruptures beyond that associated with providing a more complete survey of check valve leak events. The BNL analysis fails to take into account a very crucial-part of our evidence about interfacing valve _ ruptures-the fact that we have never experienced more than relatively minor problems with 0-rings and imperfect valve disc seating on any ECCS valves, including the interfacing check valves in nearly 500 reactor-years' of PWR experience. This is a hard statistical fact that must be accounted for in any valid assessment of interfacing LOCA frequency.

To examine whether the BNL results for accumulator check valves provide reasonable reselts for interfacing check valves, consider the event of a valve disc rupture. Neither PLG nor BNL has identified any valve disc ruptures. The most severe event identified was one accumulator check valve leak of about 200 gpm due to 0-ring (synthetic rubber material soaked in weak boric acid) deterioration, not a structural problem with the stainless steel. Fence, there has been zero valve disc ruptures in PWR experience thus far.

On this point, the PLG and BNL analysis is in agreement. What is at issue is how to estimate the exposure time.

9

~

.I The PLG estimate of exposure time included PWR experience thru November 30, 1984, which was about 424 reactor-years and 1.1 x 108 component hours for all ECCS check valves. BUL extended the data survey through December 31, 1985 for a total of 451 reactor years. There has been roughly an additional 50 reactor-years of PWR experience through the end of 1986 for a total of 500 reactor-years without any known valve disc ruptures, and certainly no interfacing check valve disk ruptures. The average number of ECCS check valves per plant in the PLG analysis of exposure data is about 30.

Hence, using all the ECCS valves, the total valve exposure time I

thru December 31, 1986 is:

(500 PWR years) x (30 valves per plant plant) (8766 hrs / year) = 1.3 x 108 hrs.

The above exposure time includes accumulator check valves, ECCS/RCS interface check valves and other check valves such as pump discharge check valves.

To get an idea what fraction of the ECCS check valves are interface check valves (i.e. those designed to isolate the reactor coolant system from the ECCS but excluding the accumulator check valves), consider the distribution of ECCS check valves at Seabrook shown on Figure 4-8 in PLG-0432. This figure includes the following ECCS check valves.

Number of Location checkvalves 1.

RHR/RCS cold leg interface 8

2.

RHR/RCS hot leg interface 4

3.

SI/RCS cold leg interface 4

4.

SI/RCS hot leg interface 4

1 5.

CHRG/RCS cold leg interface 6

6.

Accumulator cold leg interface 4

7.

ECCS pump discharge 6*

Total 36

• Two RHR pump discharge check valves are not shown in Figure 4-8 of PLG-0432. The discharge check valves on the charging and 2 SI pumps, are, however, shown in the Figure.

The above count does not include check valves shown in Figure 4-8 located in lines between the ECCS pumps suction paths and RWST and containment sump and indicated as belonging to the containment spray system. Of the above 36 check valves, 26 (items.1 thru 5) are ECCS/RCS interface valves, 4 are accumulator valves and 6 are pump discharge valves. There are 12 of i

the interface check valves whose discharge side communicates directly with the reactor coolant system. All but 1 of the remaining interface check valves and the accumulator check valves have one normally closed check valve between it and the reactor coolant system. If the first check valve is leaking the second check valve experiences the RCS pressure load. Hence, the Seabrook ECCS check valve configuration has a total of i

i 10 l

12 + 14 - 1 + 4 - 29 check valves that are in the same configuration as the interfacing LOCA check valves being analyzed, namely communicating with the RCS or in series next to one that does. Therefore a fraction of 29/36 =.80 of the ECCS valves at Seabrook are directly applicable to the situation being modeled in the interfacing LOCA analysis. This fraction should be representative of PWRs in general. Hence, thru 12/31/86 we have experienced (.80)(1.3 x 108 hrs) =.1 x 108 hrs of interfacing check valve experience without ex eriencing a disc rupture. This neglects to consider the roughly 3 x 10 hrs of experience associated with ECCS valves that had been included in the valve count of PLG 0432. While arrived at in a different way, the 1 x108 hrs used in PLC 0432 seems to be well sup-ported in view of additional experience that has accumulated since 11/30/84 and in consideration of some ECCS valves that possibly could be excluded as not interfacing with the RCS.

If one conservatively equates a disc rupture with a leak in excess of 1800 gpm, we can compare the above experience with the PLG-0432 and BNL analyses of check valve failure frequencies. If we had in fact experienced a check valve leak or rupture having a etze of 1800gpm, or nine times larger than the largest observed, and followed the same methodology as in the PLG and BNL analyses but with the above exposure time, then the median failure frequency at 1800gpm would have been estimated as (1/1 x 108 hrs) * (8766 hrs / year) = 8.8 x 10-5/ year. The fact of the matter is we did not observe such a leak or rupture, and therefore 8.8 x 10-5/ year is not a median but an upperbound. This result tends to support the analysis in PLG-0432 and strongly indicates that the accumulator valve only approach of BNL overestimates this frequency. The median of tte BNL analysis without experiencing a rupture is greater than this analysis would indicate even if a 1800 gpm rupture is assumed.

It needs to be emphasized that the check valve failure rate estimates in PLG-0432 were assessed as having high degree of uncertainty. At 1800 gpm, the PLG-0432 estimates of failure frequency ranges from 1.4 x 10-6 at the 5th percentile to 2.8 x 10-4 at the 95th percentile. The PLG-0432 upper bound is more than 3 times higher than the upper bound estimate obtained i

directly from the above analysis of exposure hours. The assessment of uncertainity in PLG-0432 took into account the sparcity of the data and the extra uncertainity associated with extrapolating the observed data to larger leak sizes.

In view of this, we do not feel it is valid for BNL to simply scale up this distribution using the same lognormal range factors, after reassessing the central tendency of the distribution.

I Another perspective on the reasonableness of the PLG-0432 check valve failure frequency estimate can be gained upon comparison of the 1800 gpm values of the failure frequency with the failure rate used in the SSPSA V-sequence analysis. That failure rate, in turn was originally developed j

in the Indian Point PSA and is compared with the PLG-0432 and BNL failure rates for 150 and 1800 gpm in Figure 2-1.

The SSPSA distribution was developed using a different methodology and interfacing ECCS check valve data at 46 plants from start of commercial operation thru 12/31/82.

That analysis observed no disc ruptures in 1.15 x 107 valve hrs of oper-l l

11 l

ation. The methodology that'was used to generate the SSPSA distribution was to break up the data into separate parts by plant as shown in Table 2-1.

Then, a. plant-to plant variability distribution was developed using.

the first stage of the 2-stage Bayesean update approach described in Reference (1). Because of.the limited amount of data and zaro failures, the SSPSA results are heavily influenced by the initial' prior assumed i

in the application of. Bayes Theorem.

In this methodology a uniform distribution is assumed over a grid made up of assumed possible discreet values ofjg and gr", the logarithmic mean and standard deviation of.the lognormal distribution. The latter distrib.ition is used to characterize plant-to plant variability in the f ailure frequency. While the SSPSA analysis was only weakly influenced by the amount of evidence available,

. it is of interest to note it's excellent. agreement with the PLG-0432 re-sults at 1800gpm. The SSPSA distribution brackets both ends of the PLG-0432 distribution.

A. final check on the reasonableness of the PLG and BNL analyses is ob-tained by using the BNL' results at 1800gpm in the SSPSA V-Sequence method-ology.

In the SSPSA, no credit was taken 'for piping integrity.or operator i

actions to prevent melt or isolate the leak. If we use the BNL results at 1800gpm, recompute the VI and VS initiator models and use the SSPSA plant model assumption, a mean core melt frequency, Ref (1), resulting from interf acing LOCA on the order of 2 x 10-5 per reactor year is obtained.

j-This is comparable to the frequency of the dominant core melt sequences currently assessed for Seabrook. This is not a reasonable result, even without taking credit for piping integrity and operator actions.

d In summary, we have the following response:

o BNL has provided a more complete survey of check valve leakage events for accumulator check valves.

)

o The BNL analysis of f ailure frequency using accumulator I

valve data provides unreasonably high values at 1800 gpm or greater leak rates for interf acing check valves.

o The BNL analysis does not take into account the most relevant evidence of all the absence of disc ruptures in check valves that interf ace between the ECCS and RCS in 500 reactor years of PWR experience.

o The PLG-0432 uncertainty d'stribution at 1800gpm is in ex-cellent agreement with the SSPSA disc rupture distribution that was developed using a different methodology.

o The BNL results at 1800 gpm would produce unreasonable results for core damage frequency if no credit were ta' ken for piping integrity or operator recovery actions.

Reference (1) S. Kaplan, "On a ' Two Stage' Bayesian Procedure for Determining Failure Rates from Experiental Data, " IEEE Transactions.

on Power Apparatus and Systems, Vol. PAS-102 No.1 January 1983.

12

P I

- The_PLG-0432 results at 1800 gpm including the uncertainty o

distributions still represent the best statement of the possible range of values of check valve ruptures that freely account for applicable statistical evidence.

2.2 Check Valve Population BNL comments that " to estimate. the total number of check valve hours, we used the total population of check valves at the interfaces. This resulted in substantial overestimation of check valve hours".

It is correct that the total ECCS check valve population was used. However, as noted in Section 2.1 above most of these valves would be classified as interfacing check valves when defined as those between the RCS and ECCS, or one check valve removed from these valves. At Seabrook, 80% of the ECCS valves are " interfacing". As noted in Section 2.1 above the PLG failure rate distribution at 1800 gpm is in excellent agreement with the observation of no disc ruptures in 500 reactor years of PWR experience with about 108 hrs of interfacing valve experience.

2.3 Estimating Leak Rates BNL states that "when estimating leak rate for accumulator check valves from accumulator inleakages, it must be recognized that the reduced leak rates relate to two check valves in series, rather than leakage through a single check valve".

It is agreed that there is, in principle, an element of uncertainty in estimation of leak rates associated with the possibility that there may be multiple leakage between two series check valves. If this is in fact the case, as noted in the comment, what would be observed would be the limiting (smaller) of the leakages. However, unless the rate of leakage is very high, it would be unrealistic to expect this problem to impact the results in a statistically significant way.

For independent events, the probability of observing mulitple check valve leakages is vanishingly small. Even considering common cause failures, even a very high value of the beta factor of

<.10 would lead to a high confidence that any particular leakage is coming from a single valve.

It also should be noted that the leak rates in PLG-0432 were assigned very conservatively. We believe this conservatism more than offsets the effects of rare multiple valve leakage.

13

e e

It should be noted that each individual check valve in the ECCS, particularly the accumulator and interfacing valves must be periodically and independently tested. Hence if there were multiple leakages, both would eventually be detected.

2.4 Leakage Frequencies BNL states that " the leak failure frequencies versus leak rate curve presented in the study (reproduced in Figure 2.1) is only a first approx-imation for a more precise leak failure frequency versus relative leak rate curve.

In particular, this curve pooled data involving a variety of check valve sizes. A more sophisticated treatment would require knowledge of the size population of check valves at the interfacing pathways".

As noted in 2.1 above most of the ECCS valves are interfacing valves.

2.5 Linear Extrapolation BNL states that "the largest leak rate in Figure 2.1 is of the order of 200 gpm, whereas the arena of interest ranges to 65,000 gpm.

The " linear" extrapolation to higher rates is not necessarily j ustified.

If the shape of the distribution is Pareto, the linear extrapolation is in order. However, if it follows a Rayleigh distri-bution, the extrapolation is not correct (but conservative). Seabrook specific consideration (valve sizes, designs) are rot made in the analysis".

The results of the V-sequence analyses in PLG-0465 do not depend at all on values of the curve beyond 1800 gpm.

The plant response to valve ruptures at a greater than 1800 gpm was assumed to be equivalent, i.e., the RHR system pressurizes to 2250 psia.

Hence values on the curve beyond 1800 gpm have no impact. Overall we have extrapolated the curve based on 2 decades of leak sizes by about I decade.

It is also important to note that in RMEPS, the analysis did not hinge on a point value of valve failure frequency. Very large ranges of un-certainty were assigned including a specific allowance for the added un-certainty associated with extrapolation.

We also disagree with the comment that design specific features of Seabrook were ignored in the analysis. There are three ways in which specific features can be incorporated into a PRA type system analysis. These include:

o use of design and procedure specific model of the system failure modes via the boolean equation, block diagram, or fault tree o use of plant specific failure data o screening of generic event data for applicabilty to the plant in light of plant specific features.

14

1 e

f The first and third of these were incorporated into the Seabrook analysis. The second is impossible for a new plant. Hence we find no justification for this comment.

2.6 Leak' Tests BNL states that "the initiator models implicity assume that the leak test of the valves " discover" all failures and valves behave as new af ter each test. The study does not describe the relevant test processes and the expected "real" ef ficiency of these test".

The initiator models are state of the art models and are consistent with the way which testing was treated in WASH-1400 and NUREG-0396.

Also, the addition of the d term tends to pickup previously unde-tected failures. The Seabrook test procedures were reviewed very caref ully to justify credit for testing. Even so, the testing intervals are conservatively assigned (one year for MOV's, 1 1/2 years for CV's) in view of the fact that procedures call for testing of the interf acing valves each time the plant goes to cold shutdown.

It is very conservative to assume Seabrook will only go to cold shutdown once a year.

2.7 Common Cause BNL states that "the report does not consider common cause failures.

Such failures indeed happen due to boron deposition, improper maintenance such as installation of improper components (gaskets, seats, or valve disks) which may fail almost immediately or at a later time".

It is true that the initiator models did not include a term explicitly identified as accounting for common cause failures. However, such a term is not really necessary for the type of component configuration encountered here. In the term,l}d, the3 represents a random failure of the first valve which can be associated with a normally operating component for selection of the appropriate model. The 3 d term represents the demand failure of the second valve which can be likened with a standby component. Anytime you put together a combination of an operating and standby component, any failures of both components will occur at the same time because of the way I

the system is designed, regardless of whether the failures are due to in-dependent or common causes. Hence it is not necessary to introduce a separate term for common cause failures. It is only necessary that the value selected for3 d accounts for both types of causes. We believe that the value selected for)\\d in PLG-0465 not only accounts for common causes but does so in a conservative manner. Also, the valve configuration of the interf acing valves must be tested to confirm initial leak f ree seat-ing prior to each plant startup. Hence, boron deposition while it may occur is not applicable to the initiation of subsequent leakage or ru pt ure. Also it is not clear how gasket and seat problems could

+

j create a disc rupture.

2.8 Plant Walkdown BNL identified the following concerns:

1.

Ability of RHR pump leakage to be detected in the Control Room -

concern lies with vault compartmentation design with equipment vault sump not receiving leakage promptly thereby delaying level detection input in t'ne Control Room.

15

P 2.

Ability of RHR pump relief discharge into the PRT to be distinguishable in the Control Room from the pressurizer relief and safety valve relief -

discharge - concern with the latter relief and safety valve discharge tailpipe temperature.

The connecting door from the RHR pump room to the CBS pump room is not necessary to meet any SB design requirements and can be removed. This will facilitate the flow of any water leakage from the RHR pump in the RHR pump room to the CBS pump room and thus the level indication.

Existing instrumentation provides adequate capability to distinguish RHR relief valve discharge from either PORV or pressurizer code safety dis-charge to the pressurizer relief tank during a V-seauence event.

The pressurizer PORVs and safety valves have redundant monitoring available, i.e.; the tailpipe temperature monitoring, and the tailpipe acoustic monitoring. Pressurizer discharge to the PRT is characterized by high RCS pressure, position indication of the PORV's and PORV block valves, high tailpipe discharge temperature and increasing PRT level, pressure, and temperature.

2.9 Operator Actions The authors of PLG-0432 and the V-sequence operator actions analysis are unfamiliar with the TEEM approach to assessment of human reliability. We note that the reference provided for TEEM is dated some nine months af ter PLG-0432 was published. We find it inappropriate to criticize the PLG-0432 analysis for its failure to anticipate a subsequently published contribution to PRA methodology by BNL. With regard to the quantification of 3 different operator actions using the same number, we disagree that the numbers must be necessarily different.

It was simply judged that the most appropriate value for each action in NUREG-1278 corresponded to the same value and that the key difference was the hardware contribution.

With regard to the need for revising ECA 1.2, Training has been incorp-orated. Changing procedure may require additional studies to ensure that optimal procedure guidance is available for the operators.

2.10 Independent Check of IDCOR Result PLG performed an independent calculation to confirm the IDCOR result that peak pressure in the RHR system would be limited to RCS pressure. This is provided in Enclosure 1.

All conclusions that were reached regarding the limiting failure modes were backed up by stress calculations most of which were not presented in the reports reviewed by BNL.

Additional stress calcul-ations were performed to show that 50% of the HX material would have to be removed to result in stress approaching yield at 2250 psia.

In addition the potential for unfilled RHR piping was subse-quently addressed, quantified and found to be insignificant. Doc u-centation to support the above except for the dynamic pressure calculation was submitted to NRC/BNL shortly before the issurance of the BNL draft report as part of RAI 75.

1 16

TT.

us-

[

'2.11 ' Treatment of Pool Scrubbing Assignment of the traditional unscrubbed V-sequence to release category SIW is'more conservative than WASH-1400, where the V-sequence was assigned.

to release category'PWR-2..This conservatism was not significant at the frequencies predicted in PLC-0465, but it is worth pointing out in the

context of the increased frequency estimated by BNL.

BNL has concluded ~that source. term mitigation by pool scrubbing in the V 1

_. sequence is justified.at least at the level credited by WASH-1400. (DF=100) for sequences where the break' location is under water. BNL raises the question of pool subcooling and on the basis of our submittal accepts the conclusion that the pool is subcooled. Additional confirmation for the pool subcooling argument can be provided independent of the analysis submitted earlier. The release of radionuclides and hot gases into the pool in the-RHR vault occurs.very slowly.

Both the stirring action by the gases and the slow release assure that the bulk fluid is isothermal at 212 degrees F, with some cooler temperatures at the walls. If temperatures higher than 212 degrees would exist, free convection mixing _would render the pool is o-thermal even in the absence of a stirring effect.

Given an isothermal pool temperature of 212 degrees, the fluid.below the surface is subcooled due to the hydrostatic. pressure of the fluid.

Based on PLG-0432, the RHR pump seals would be submerged under 21 feet of water before the release of radionuclides into the vault begins. The table below gives the subcooling between the pump seal and the surface for a range of pool depths.

~

Pool Depth Above Pool Subcooling '(deg F)

Seals (feet)

At Seal Elevation Pool Average 21 24.6 11.6 15 18.4 8.8 10 12.8 6.2 5

6.8 3.4 0

0 0

This effect was neglected in the analysis submitted previously and it provides a more significant degree of. subcooling particularly nearL the RHR pump seals where the most significant pool scrubbing occurs.

17

l l

Plant No. of Exposure Plant No. of Exposure Occurrences Time (Hours)

Occurrences Time (Hours) i 1

0 2.2 (5) 24 0

2.5 (5) 2 0

6.0 (4) 25 0

4.1 (5)

~!

3 0

1.5 (5) 26 0

3.8 (5) 4 0

2.4 (5) 27 0

2.9 (5) 5 0

1.8 (5) 28 0

2.7 (5) 6 0

4.2 (5) 29 0

1.8 (5) 7 0

2.3 (5) 30 0

2.9 (5) 8 0

1.6 (5) 31 0

2.1 (5) 9 0

9.8 (4) 32 0

2.9 (4) 10 0

1.5 (5) 33 0

3.5 (5) 11 0

3.1 (4) 34 0

3.2 (4) 12 0

2.9 (5) 35 0

1.0 (4) 13 0

2.5 (5) 36 0

1.8 (5) 14 0

7.5 (5) 37 0

3.3 (5) 15 0

2.9 (5) 38 0

3.2 (5) 16 0

2.7 (5) 39 0

1.6 (5) 17 0

1.3 (4) 40 0

2.8 (5) 18 0

1.9 (5) 41 0

3.9 (5) 19 0

1.8 (5) 42 0

3.9 (5) 20 0

6.1 (4) 43 0

5.9 (5) 21 0

2.6 (5) 44 0

4.7 (5) 22 0

2.3 (5) 45 0

4.3 (5) 23 0

2.3 (5) 46 0

3.0 (5) l 6

Tot al 0

1.1 (7)

Table 2-1 Plant Specific Data Used in IPPSA and SSPSA Check Valve Rupture Frequency Analysis.

18

~0

-5

-4

-2

~

10 10 10 10 10 I

I l

l l

th th 5

X 95 g

medly ge RMEPS (> ISO gpm)

BNL (> ISO gpm)

RECIPROCAL OF INTERFACING CHECK VALVE EXPERIENCE IN PWRS WITHOUT FAILURES RMEPS (> I 800 gpm)

BNL (> 1800 gpm)

SSPSA GROSS DISK RUPTURE

-5

~

10 10 10 10 10 FREQUENCY ( events per reactor year )

FIGURE 2-1 COMPARISON OF CHECK VALVE FAILURE FREQUENCY UNCERTAINTY DISTRIBUTION 19

(*

3,0 SHUTDOWN ~ EVENTS ~

BNL provides comments on accidents during shutdown and refueling conditions in Section 2.2 of their report. They note that. their review of the information provided on this subject had not been completed. NHY responses to the comments provided thus far are pre-sented below.

3.1 Auto Closure' Logic'will~ Isolate'Both'Suetion Paths (Single Pressure Transmitter)

BNL commented that "....another dif ference between Zion and Seabroo'k is that Zion has a single drop line and Seabrook has double drop lines. This is not expected to be a significant difference because the auto closure logic at Seabrook will isolate both suction lines when a spurious signal is generated, i.e., a single pressure trans-mitter provides input to the interlock logic of the two. inner isola-tion valves, and a separate pressure transmitter provides input to the interlock of both isolation valves...."

Per procedure, once the drop leg isolation valves are open, one valve in each line, associated with a DIFFERENT. protection channel

'is tagged in the open position. This prevents a single failure of one protection channel instrument, interlock or a spurious action of the auto-close system from isolating both trains.

3.2 Credit'for'Two RHR Suction Paths BNL made the following comment on the credit that was taken for Seabrook having two RHR suction paths "....this reduced the frequency of loss of RHR by a multiplicative f actor of 0.145 and the core damage frequency by approximately a factor of 2.

From the information available to BNL, it is not clear that this reduction is justified...."

When the RHR System is aligned for RCS overpressurization protection, two of the four (one on each drop leg) drop leg valves are open and power removed. This precludes a single failure from isolating both RHR suction lines.

When the RHR System is aligned for normal DHR, all of the drop leg isolation valves associated with the operating RHR train are open and power removed.

Therefore, no matter what failure is postulated, there will always be one RHR suction safety valve available.

The procedures for removing power f rom 2 of the 4 drop line MOV's preclude the spurioins closure of valves in both lines at the same time due to the kind of spurious events that led to the experienced loss of DHR events. Therefore either mulitple independent or common cause failures must be postulated as was assumed in the derivation of the Seabrook reduction factor in the response to RAI 21. With respect to the point raised by BNL that there would normally be one 20

operating RHR train and the other in standby or maintenance, the unfor-tuttous event in which a single drop line is isolated and the other train happens to be in maintenance was explicitly accounted for in the derivation of the reduction factors.

If a single drop line becomes isolated in the normally running RHR train, there would be a high chance of recovery to realign the available standby train and/or reopen the drop line. No credit was taken for RHR recovery in the Seabrook analysis of sequences borrowed from NSAC-84. However, if the drop line on the operating RHR train were to close, there would be 3 different alarms in the control room as noted in our response to RAI 40.

Because the reduction factor is much larger than typical non-recovery factors this scenario does not challenge the credit taken for 2 drop lines. Therefore, we maintain our position that the reduction factor is j ustified.

As noted by BNL, operator failure to restore the standby train was not explicitly included in the Seabrook analysis. To include such errors in the model, the correction factors presented in the response to RAI 21 can be re-expressed as follows.

For spurious valve closure, the correction factor becomes.

/3 MOV + (1 -p MOV) (.5) [0RHRM + (1 - ORHRM) O l R

where

/3 MOV = MOV common cause failure fraction ORHRM = RHR train maintenance unavavilability during shutdown OR = frequency of operator failure to restore RUR cooling The correction factor for errors in inverter switching becomes

.5 [QRHRM + (1 - ORHRM) O l R

Table 2.7 in the BNL report lists 130 events involving loss of DHR, all which were terminated by successful operator action therefore the generic frequency of non-recovery must be less than 10-2 In comparison to Seabrook most plants pro'cably do not have the same alarm capability to alert the operators to a loss of RHR pump suction. The NSAC-84 plant does not.

Therefore a value of 10-2 for event O is very conservative for Seabrook.

R But even if we assume this value, the spurious valve closure correction factor would only increase from.072 to.076 and the inverter switching factor from.031 to.035.

Hence the effect is insignificant, even when a conservative recovery factor is used.

3.3 Loss of DHR BNL states that dominant causes of loss of DPR are; 1.

Spurious closure of RHR suction valves and, 2.

Inadequate vessel level (RHR cavitation)

Addressing the first concern, let us look at one such scenerio for Seabrook, loss of DHR due to a failure high of a wide range RCS pressure t ran smi t te r. (Note:

this is for example purposes only, the loss could occur for any reason listed in the BNL report, and the result would be the same).

21

~

C

• e The LTOP protection system at Seabrook is much less dependent on human error, (due to automatic arming), and the backup protection is afforded us again by the use of dual drop legs for the hot leg to RHR cooling system lines. Let us further review a " typical" overpressurization event. Since this type of event has it's highest probability of happening when the RCS is " water solid", we must make some basic assump-tions:

As s umptions :

(1) Per operating procedure, the RCS isn't taken solid until Mode 5,

(<180 degrees RCS).

(2)

Both trains of RHR are aligned, one is in service. This includes, per procedure, that one of the two RHR suction valves in EACH RHR drop leg is tagged open, power removed (the tagged valves are'on opposite ELECTRICAL / PROTECTION trains). This is done to protect the system from an instrument failure, or single human error, isolating BOTH RHR suctions (and consequently, both suction relief valves) from the RCS.

(3) All equipment required by technical specifications to be made inoper-able, has been made inoperable per it's applicable procedure.

Let us assume, at this point, with the RCS at <180 degrees F and water solid, the operating RHR loop fails, due to a wide range RCS pressure transmitter failing high.

SEQUENCE OF EVENTS:

1.

Inadvertent closure signal occurs.

2.

Operating loop RHR suction valve closes, non-operating loop suction valve doesn't close because it is tagged open per procedure (see initial condition #2 above).

3.

Operator receives VAS alarms on low RHR flow, low-low RHR flow and overhead alarm for low RHR flow.

4.

RCS pressure increases rapidly due to loss of letdown with continued charging.

5.

LTOP automatically actuates at 541 psig, this is independent of operator action, LTOP at Seabrook AUTOMATICALLY arms based on auction- :

eered RCS temperature and wide range pressure.

6.

If pressure reaches 450 psig, the non-operating RHR loop's suction safety valve will open.

For pressure to get to this point, some malfunction of LTOP would have to take place. This is due to only 22

0 ONE charging pump allowed operable in this mode. The RHR safety valves lift at 450 psig, passing over 900 gpm.

The runout of one centrifugal charging pump is approximately 550 gpm.

This 550 gpm is measured thru the safety injection flowpath, the operable charging pump is aligned thru the NORMAL charging path with it's flow control valve in MANUAL. Although there is no procedural requirement for the flow control valve (FCV-121) to be in manual, it is highly unlikely it will be in automatic. This is because the FCV-121 has an electrical stop built into it that ensures RCP seal flow is provided in the event of a high pressurizer level at OPERATING pressure. This stop position, with the RCS depressurized, would provide far to much charging flow to be balanced by letdown.

All of this results in substantially less charging flow than renout,

and in any case, less than the capacity of ONE RHR suction relief valve.

Regarding the second dominant sequence, the Seabrook operating staff have provided the operator with procedure guidance for inadequate vessel level. The dominant sequence for Zion (local operator leaves manual vessel drain valve open during drain down operations) is less likely here because the drain down is:

1.

Monitored in the Control Room on RVLIS and refueling level indications.

2.

Capable of being terminated from the Control Room without the utilization of a field operator.

As evidenced by BNLs report, out of the 130 ir.dustry events, all have been terminated by timely operator action. Although some of the recovery actions have lasted longer than an hour (number not given by BNL),

the preceding statement of a loss of DHR 36 hours4.166667e-4 days <br />0.01 hours <br />5.952381e-5 weeks <br />1.3698e-5 months <br /> af ter a reactor trip while partially drained causing core uncovery is difficult to understand. Although it is within the RHR/RCS system capabilities, and within technical specifi-cations to be in Mode 5 in 36 hours4.166667e-4 days <br />0.01 hours <br />5.952381e-5 weeks <br />1.3698e-5 months <br />, it is highly unlikely that the vessel could be drained down that fast with the reactor head still installed.

The reactor head, per technical specifications, cannot be removed for 100 hours0.00116 days <br />0.0278 hours <br />1.653439e-4 weeks <br />3.805e-5 months <br /> after last critically (i.e., 100 hours0.00116 days <br />0.0278 hours <br />1.653439e-4 weeks <br />3.805e-5 months <br /> from trip). Even if it were possible to draia down the RCS under these conditions with the head in place, why would the core uncover? Assuming a " typical" loss of DHR, the RHR suction valves closing, there would be no mass loss, in fact, the RCS mass would increase due to the loss of RHR letdown and continued charging from the operating charging pump.

3.4 Relief Valves Ineffective /2 PORV's or Both Relief BNL states "....in addition to the PORV's the relief valves in the RPR systen may be unavailable to relieve pressure.

Each RPR suction line has a relief valve with 900 gpm capacity at 450 psig. However, these relief valves may be made ineffective if the RHR suction valves close automatically when the setpoint of 660 psig is reached, as was the case in the Turkey Point 4 Events. Actually, the Seabrook technical specifications only require either both PORV's of both RHR relief valves to be available".

Seabrook features reduce the frequency of occarence of this LTOP event at Seabrook.

23

e -

it h

The RHR suction isolation valves are opened and power removec to the " Cross

• 1

. Train" RHR suction isolation valves during the time that the RHR system is in operation.

This provides assurance that the failure of-t,LTOPS pressure transmitter will not isolate both trains of the RHR system thus blocking access to the RHR suction relief valves.

The autoclosure setpoint for the Seabrook Station RHR suction isolation valves is 660 psig. This is above the RHR suction relief valve set pressure and the total accumulation during relief valve lif t.

The Turkey Point Event overpressure source was the charging pumps. The Seabrook Station RHR suction relief valves are sized to pass the flow from two operating charging pumps at the relief valve set pressure. Two charging pumps at the Seabrook Station are inoperable by Technical Specifications.

Thus, there is added assurance for Seabrook that the RHR system will not be isolated and that the RHR suction relief valve will be available to terminate the transient.

3.5 Applicability of LOCA's a

BNL commented that "NSAC-52 reviewed operation experience within 5 calendar years up to the end of 1981, and identified 10 loss of coolant events at PWRs. They were caused by the follwing causes:

1.

Inadvertent manual initiation of RHRs supplied containment spray.

1 2.

Inadvertent loss of inventory to the containment building sump and/or automatic initiation of recirculation mode of low pressure safety injection.

3.

Inadverent loss of inventory via the RHRs relief valves.

4.

Inadverent loss of inventory via mispositioned crossconnect or drain valves.

5.

RHRs valve packing gland removal during plant pressurization, dislodging the valve packing gland.

6.

Gross valve packing leak".

With regard to item I, the Seabrook Station design is not the RHR supplied containment spray type. Multiple mispositioned valve alignments (motor and manual valves) would have to occur for these sequences at Seabrook.

This would occur only by misaligning the refueling canal drain line in the discharge of the RHR pumps to the suction of the containment building spray pumps and subsequently to the containment building spray header.

l t

24

7 For item 2, loss of inventory to the containment building sump as a result of sump isolation valve mispositioning, testing, or automatic initiation of the recirculation mode of low pressure safety injection would require the failure of a check valve to close. Thus the check valve in the Seabrook Station RHR pump suction line from the sump pre-vents loss of inventory to the sump if the sump isolation valve is inadvertently opened during RHRs operation.

The last four are applicable to the Seabrook Station system designs.

3.6 LTOP/ Turkey Point Events BNL commented that

"....The NSAC-84 analysis of low temperature overpress-urization may be too optimistic. Events such as those Turkey Point 4 indicate that the frequency with which a rapid pressurization occurs with the RHR system isolated and the PORV's unavailable is higher than 10-3 per year. The operators have only a few minutes to respond to the event" The IE notice referenced in the BNL Report, IE notice 82-17, Overpressur-ization of Reactor Coolant System, describes two events at Turkey Point and an inoperable LTOP system at North Anna. The North Anna event is not applicable to Seabrook since Seabrook uses solenoid PORVs rather than air with backup nitrogen accumulators.

{

I The two Turkey Point Events involve failures in the operable LTOPS while the redundant circuit was out of service for calibration.

In the first event a pressure transmitter isolation valve was found closed. In the second event a summator in the actuation circuitry failed, furthermore the summator was not included in the instrument channel surveillance.

In both occurences the letdown via the RHR system was automatically isolated by automatic closure of the RHR suction isolation valve.

Both events were terminated within two minutes by manual operator action af ter the RCS pressure exceeded the allowable values in the plant technical specifi-cations.

Two additional Seabrook features can reduce the frequency of occurence of this LTOP event for Seabrook.

The RHR suction isolation valves are opened and power removed during the time that the RHR system is in operation. This provides assurance that

{

the failure of a LTOP pressure transmitter will not isolate both trains l

of the RHR system thus blocking access to the RHR suction relief valves.

l l

The autoclosure setpoint for the Seabrook Station RHR suction isolation valves is 660 psig. This is above the RHR suction relief valves set pressure and the total accumulation during relief valve lif t.

The Turkey Point Event overpressure source was the charging pumps.

The Seabrook Station RHR suction relief valves are sized tc pass the flow from two operating charging pump at the relief valve set pressure. Only one is allowed to be operable for this mode of operation. Thus there is added assurance for Seabrook that the RHR system will not be isolated and that the RHR suction relief valve will be available to terminate the transient.

25

3.7 Assessment of Consequences In our response to RAI 21, a conservative approach was followed to char-acterize consequences lof loss of shutdown cooling events by assigning the -

resultant core damage sequences directly to release. categories for power operation events. If a core damage scenario were to develop during shut-down, the scenario would evolve more slowly for a given configuration due to a lower value of af ter-heat and the source term would be reduced by radioactive decay over the period between the time of power reduction and shutdown to the time of ~ release which could be days to weeks. The accident probability calculations in the Seabrook assessment used Zion data for estimating the frequency and length of outages. The length of the outages corresponds to the possible range of times to initiation of a core damage scenario as measured from the time of plant shutdown. The Zion data i

consisted of 12 refueling outages of average duration of 1,992 hrs. and 49 maintenance outages of average duration of 488 hours0.00565 days <br />0.136 hours <br />8.068783e-4 weeks <br />1.85684e-4 months <br />.

1 Given a system f ails over a period of time, the conditional distribution of the time of failure is roughly uniform. Hence the mean time of scenario initiation would be i

12

• 1992 + 49

• 488 392 hrs.

=

(12 + 49)

• 2 l-If it is argued that a shorter time should be used then the accident pro-abilities would be correspondingly Icwer because of a shorter time the RHR would be "at risk" to fail.

Since the response to RAI 21 was prepared some additional CRACIT runs were performed for release categories S2 and S6 with the release times delayed to simulate the delay to onset of the loss of RHR scenarios. The approach taken was to simply translate the release times for all release puff s by 48 hours5.555556e-4 days <br />0.0133 hours <br />7.936508e-5 weeks <br />1.8264e-5 months <br />. Based on the Zion data, about 94% of the scenarios would have longer delay times.

By performing this translation, the radioactive decay of the source term is accounted for, but the slower evolution of release due to lower af ter heat is not.

Hence, these delayed release calculations are still viewed as ccnservative although they are somewhat less conservative than provided in the RAI 21 response.

The revised CRACIT runs were used to provide an enhanced assessment of the contribution of shutdown loss of cooling events to 200 rem and 50 rem dose vs. distance curves. The new curves are presented in Figure 3-1 and 3-2 for the realistic and conservative treatment of equipment hatch closure. These curves supercede Figures 14 and 15 in response RAI 21.

Also, Figures 14 and 15 in response RAI 21 were plotted conservatively too high. For example the 200 rem curve in Figure 14 should have been offscale.

26 4

...-.~_. -,.

i t

i a i i i is

,,c

.,,,,,i

~

NUREG.03gg u,

SENSTTWTYSTUOY

m----====

~

POR EWaf5 ATPowet

~

sN Everts

~

L (z.m el

.per-4) 01 :

~~

h i

ss

-s

~

Og

\\

Lt

\\

\\

so mau gg

\\

\\

Wo

\\

\\

OW

\\

\\

kb 0 0' 1

~~

\\20e I tg l<

14dd I l

\\

-l h

l I.

v8 i

==+Q go 200 REM

l

o. cat t

1o 1oo 1.c00 OISTANCE(MILES)

Figure 3-1 COMPARISON OF 200 REM AND 50 REM 00SE VERSUS OlSTANCE CURVES WITH CONTRl80T10NS FROM SHUTDOWN EVENTS kJefd THf( Arleg333 dr y ' fit hrs, 27 1 ;. ' '.

3 I

. iiei:l

> a i i eg i

~

[

NURgg.03gg as NN.

~

ATIIOWER sMurnowwm WW No cacDff wi N EQuipu err HA D RECOVERY 0I g<

si

=-

s

~

\\.

w

\\

k ';e t.,,a 50 REM us n\\

88

\\\\

Q 0.0:,-

3 %@

w M

8"<

~

=I k.

lg l 's S.

l t

~

IL l

~,

2OOREM I

>1.

o.co

,,i 1

10 100 1,000 OISTANCE(MILES)

Figure 3-2 COMPARISON OF 200 REM AND 50 REM COSE VERSUS DISTANCE CURVES FOR CONSERVATNE ASSUMPf10N OF NO CREDIT FOR 0 MTOR RECOVE OF 0 EQUIPMENTHATCH Ado ShitAssam Ve84Sc3 de 4sMrs.

28 2l-d.

1

... - - - - - - - - - - - ~ - - - - - - - - ~ - - -

i?'

e 4.0 STEAM GENERATOR TUBE RUPTURE BNL's comments on the steam generator tube rupture scenario is pro-vided in BNL Section 2.3 and an enclosure to the BNL Report.

Our detailed plant specific evaluation of steam generator tube creep failure (see RAI 47 response) indicates that this failure mode is very unlikely. There is no comperable evaluation that justifies the existence of.this failure mode at Seabrook Station. In fact we are not aware of any analyses that. shows tube f ailure occurs. Regardless,-

this is apparently a generic research issue and BNL evaluation conservatively assesses the f requency of this event without including probabilistic1y the Seabrook Analysis. If BNL uses their best estimate analysis consistent with WASH-1400 this issue becomes insignifi cant. Also, this issue will eventually be resolved as insignificant because of the many ways to reduce its frequency.

The frequency of this event is made up of 3 components:

F=fE*fo*fSG where :

F = the annual f requency of steam generator tube f ailure fE = the annual frequency of a high pressure core melt with dry steam generators f = the conditional frequency of operator f ailure to-o depressurized primary system given fE f g = the conditional probability that the Steam Generator 3

tubes f ail given fE and fo The following is BNL's Pessimistic calculation of F.

F = 2.4 x 10-6 = 4 x 10 -5 (fe) *.2 (fo) *.3 (f sg) l If NRG accepts our evidence on Seabrook fgg, would be very small and the frequency would become very small (insignificant).

j l

I It should be noted that even if fgG was to remain pessimistic, either could be reduced significantly with plant enhancements.

fE or fo Therefore, regardless of the final resolution, this issue will become an insignificant event.

Our additional comments are provided in Sections 4.1 and 4.2 below.

4.1 Fluid Dynamics Experiments vs. Models Several comments suggest that the assumption of complete mixing in the steam generator inlet plenum is inappropriate. While dye observations in the Westinghouse experiments suggested somewhat incomplete mixing, the definitive temperature measurements indicate a high degree of mixing.

Good mixing and moderate temperatures are expected near the tubesheet 29

,e.

.g due to tha antrainnant of-coldar fluid by the hat jst af ter.it sxito the top of the hot-leg.'

In view of the la'rge margins which exist in the

~

calculations (about 7000F in the best-estimate case), an assumption of even;very poor mixing would not change the conclusions with respect to steam generator tube integrity.

It is suggested that the Westinghouse scale model data is preliminary and does not offer a good test of the model. While it is true that high pressure -

sulfur hexafluoride experiments have not yet been conducted, data apanning a wide range in densities, viscosities, and coefficients of volumetric expansion have been completed. The model predictions' agree;well with this data..Given the ranges of parameters tested, no new phenomena would be anticipated in the high pressure experiments which would change this

).

agreement.

The reviewer commented that one or more steam generators could be depressur-ized, which would increase the stresses on the tubes and.thus accelerate creep rupture. It should be noted that operator actions to depressurize r

the steam generators would occur only in the presence of adequate steam generator level, in which case failure of the steam generator tubes is not an issue. A MAAP case was run with the assumption that one of the steam generetor relief valves stuck.open. As shown on Figure 4-15 of the FAI report (RAI 47), this negligibly altered steam generator inlet plenum gas temperatures compared to the base case. Based on the analysis in Appendix C (RAI 47), some degradation in the secondary side heat transfer coef ficient at the hottest portion of the tube surface would be expected at low pressure.

However, radiation (which was relatively small compared to convection at high pressures and was therefore neglected), would partially make up for this degradation. In any event, even the primary side gas temperatures (let alone the tube temperatures) in this scenario were well below the levels required to threaten tube integrity.

The reviewer commented that the MAAP code, which was used for the analyses has not been verified and documentation was not available. As part of the IODOR prodram, an earlier version of MAAP was checked line-by-line by an outside reviewer and all comments were resolved.

In addition, the MAAP code has been extensively discussed with the NRC and its contractors over several years as part of the issue resolution process. Finally, MAAP has been benchmarked against a body of detailed code calculations, experiments, integral experiments and industrial experience including TMI-2 and LOFT FP-2.

a 30 1

~

'g 4.2 Probability Assessments BNL assigns a probability of 0.2 to f ailure of the operator to depressurize the RCS prior to the occurrance of an induced steam generator tube rupture (ISGTR). In view of the long times predicted for tube heatup this seems to be an unreasonably conservative assessment.

The conditional probability of a small leak ISGTR given no repressurization of 0.3 appears very high in view of our analyses. This would mean there is a 30% chance that the results of the FAI analysis is in error by several hundred degrees centigrade, which we consider very unlikely. A more reasonable interpretation of the FAI analysis would assign a confidence lev ('

in the range of 0.95 to 0.99 for the conclusions to be substantially correct. The corresponding probability of an ISGTR then would be in the range of 0.01 to 0.05.

This would also coincide more with our assessment of multiple tube ruptures. The probability that several tubes fail given that one does fail could be fairly high if no depressurization occurs af ter the first tube has failed. In this case we would accept the value of 0.01 for the probability of a large leak ISGTR given no depressurization.

However, given that the operator has not depressurized the RCS before the first tube fails, we believe it must be more likely that he initiates depressurization after the first tube has failed. Whether depressurization at this time can prevent the failure of additional tubes will largely depend on the failure modes for the subsequent tubes.

If these tubes fail because of a domino ef fect then it is unlikely that depressurization would prevent further f ailures but it would still reduce the source term.

If each tube fails independently then depressurization would be likely to prevent further tube failures because of the slow tube heatup rate. Oper-ator action to fill the steam generator and flood the tube break soon af ter core uncovery would also mitigate the source term substantially by a pool scrubbing ef fect similar to the V sequence. BNL assigns both the small leak ISGTR and the large leak ISGTR to release category SIW. This is a very conservative assignment at least for the small leak ISGTR, and it does not recognize the fact that the release characteristics of these two cases are significantly dif ferent. If this were not the case there would be no reason to distinguish the two.

Release category S2W would best represent the release characteristics of a small ISGTR.

31

-e

+-

-1%.,

5.0 DIRECT HEATING BNL's comments on the potential for early containment failure are provided in BNL Section.4 with references to BNL Section 3 regarding failure modes and their probability.

It should be pointed out that experiments performed by FAI conclude that this phenomena is not applicable to Seabrook when the configuration

.above the cavity is properly accounted.-

The details of the BNL calculations for direct heating nre not known. However, there appears to have been a misunderstanding with respect-- to the interpretation of the conditional containment failure probability due to direct heating.. We understood that the direct heating peak pressures calculated for Zion were used by BNL to.

determine the conditional probability of containment failure for Seabrook from the Seabrook specific probability distribution for the containment - failure pressure derived for the SSPSA (Figure 4-7 of PLG-0432). In other words we understood the conditional f ailure probabilty of 0.01 to be a Seabrook specific value but based on.

pressure transient estimates for the Zion plant. It appears from the BNL draf t report that the conditional. failure probabilty of 0.01 is entirely a Zion specific value including the lower assessed strength of the Zion containment. The response provided previously was based on the Seabrook specific interpretation and it appears to have been very conservative, because it did not account for the most significant difference, namely the higher strength of the Seabrook containment which is supported by the BNL structural review.

The conservatism in the assessment of direct heating ef fects is illustrated below in an analysis which is entirely based on the BNL assessment of the Seabroo'k containment pressure capacity. The most consistent approach to utilize the BNL assessment would be to regenerate the Seabrook specific containment f ailure probability distributian (Figure 4-7 in PLG-0432) using the BNL input.' This was not possible within the time frame available.

Instead, the following simplified approach was followed.

Under wet containment conditions, the lowest median f ailure pressure indicated by BNL for a type B or C failure is 167 psig for penetration X-8.

BNL indicated that shear failure at the base could be an important f ailure mode but the finite element analysis indicated that this would not occur before the general state of yield is reached at 157 psig.

Note that on page 4-6 of the BNL report a value of 140 psig is reported for the shear failure mode which appears to be in error. Based on our previous analyses and BNL conclusion that f ailure will not occur below 157 psig (a lower bound) we assume that the median value for this failure mode would be higher than 167 psig. Therefore, the median failure pressure for type B or C failures is controlled by penetration X-8 at 167 psig.

A statement of the uncertainties in the f ailure pressure can be derived f rom BNL's assessment of the hoop failure mode.

BNL indicated a median failure presure of 175 psig with a lower bound at the general yield state which corresponds to a pressure of 157 psig. Interpreting this lower bound as the 5th percentile and assuming a lognormal distribution 32

n.-

ig yields a standard deviation of 0.066.

The overall containment failure pressure distribution based on the BNL assessment is then approximated by a lognormal distribution with a median value of 167 psig and a standard deviation of 0.066.

PT,G believes that this is a conservative assessment for the lower tails of the distribution which are important in assessing the direct heating ef fects, because the same lower bound of 157 psig would apply to the f ailure mode of penetration X-8. ' Because of this lower median failure pressure, this combination would have resulted in a narrower distribution for the f ailure pressure and it would consequently have yielded an even lower containment failure probability due to direct heating than what is estimated below. Even though this would have been a more logical and consistent distribution, the more conservative repre-sentation was chosen to be sure that the larger uncertainties for the hoop failure and the shear f ailure mode are included.

The resulting distribution was plotted against the containment failure pro-bahility distribution for the Zion containment taken from Figure 2.5.1-2 in the Zion PRA, which is reproduced in the attached Figure 5-1.

The pressure where the Zion curve has a probabilty of 0.01 is 146 psia (131 psig).

At. this pressure the f ailure probability for the Seabrook contain-ment according to our interpretation of the BNL assessment is 0.00015, or a f actor of 67 lower. This value compares very well with the value for early containment failure of 0.0001 which was used for release category S1 in the original SSPSA and which is still embodied in the analyses for the EPZ study (PLG-0432 and PLG-0465). We conclude from this approximate analysis that the direct heating ef fects are not changing any of the results or conclusions of the RMEPS study or the EPZ sensitivity study even when the BNL assessment of the Seabrook containment capacity and the NUREG-1150 data for direct heating are combined and assigned a probability of 1.0.

This would indicate that these conclusions are reached even on the basis

-of a very conservative combination of all aspects of the analysis and assumptions raised by BNL.

l 33

8 m

1 10 5

p

/OM

~~~~

2 m

5

/

I 2

3 l 10-1 j

E

/

I l

b 5

/

O.

A'

/

k 1

l" SEA BROOK y

2 l

/

. (BA.MD ON B_, o.,

~3 l

BA'L DA7~4) 5 i

/

I r

l l

f I

2 1

f f

10-3

/

/

.r r

5

/

y

/

/

2

- f,[X /()

f 2

j I

>,iiiiiii 10-4 130 140 150 160 170 180 PRESSURE (PSIA)

Of'S/]/A' USA 't M/L t.flf i/! / / t=

77

~

i

~

/)/LkQ.'/~ #fd/7A 'C' /)A)c=f) (A' iTA Q OA/ A Fu3u.re.

ET-l 34

%)

6.0 CONTAINMENT STRUCTURE 6.1 Independent Review An independent check of the containment strength calculations has been completed. This review also includes additional documentation regarding assumptions made in the SMA analysis. The review is documented in cal-culation number SBC-198.

6.2 Shear Failure at Wall-Basemat Intersection BNL investigated failures at the above location and confirmed that '* such a failure is not expected to occur for pressures up to 157 psig".

However, BNL states that "a shear failure mode at the base may develop at a pressure slightly above 157 psig".

The basis for this statement is not provided in the report.

We have concluded that the above f ailure mode will not occur until internal pressure exceedes 300 psig, wall above the 216 psig critical failure mode.

Our independent review, utilizing more conservative assumptions has concluded that shear failure in the concrete could occur at approximat ely 197 psig but that subsequent deformation would be limited by hoop and meridional reinforcement preventing liner tearing. Thus, this failure mode, while possibly initiating below 216 psig is not critical for cato-strophic failure.

6.3 Application of One Percent Strain as the Median Failure Pressure We disagree with the ENL judgment that one percent rebar strain repre-sents the median ultimate capacity of the Seabrook containmant.

The median ulimate capacity calculated by SMA (216 psig) corresponds to an average rebar strain of about 4.7% -(S

= 109 kai). These ultimate u

values are based on actual certified material test reports for the rehar material. 14 rebar strain of 1% represents loading conditions which pro-duce stress levels well below the median ultimate strength.

In fact, 1% strain in the highly ductile (34% elongation) rebar material is typically located on the steeply ascending portion in the plastic region of the material stress-strain curve, well below the as-tested median ultimate strength of the Seabrook rebar material.

35 1

~-

• • • 1 PICKAco, LCWE AND GARRICK. INC.

9 MEMORANDUM Date:

August 22, 1986 Ref:

File:

PSNH 1016.08 Copies: DRButtemer JWRead TO:

K. N. Fleming /D C. B y DJWakefield FROM:

M. V. Frank SUBJECT. SH0CK WAVE E F 'CT F LLOWING HYP0THETICAL, CATASTROPHIC FAILURE OF ISOLATION VALVES SEPARATINGHIGH AND LOW PRESSURE PIPING IN THE RHR SYSTEM INTRODUCTION AND

SUMMARY

Reference 1 contains a characterization of the nature of the hydrodynamic shock effect following a postulated sudden catastrophic rupture of high to low pressure isolation valves in piping. Two situations were evaluated, one with piping filled with water and one with voids in the water. The scenarios were evaluated in connection with an interfacing LOCA postulated to be initiated by a sudden, complete rupture of the discs in two series motor-operated valves in the RHR hot leg suction line, or an interfacing LOCA postulated to be initiated by a sudden, complete rupture of the discs of two series check valves in one of the RHR cold leg discharge lines. The situations were evaluated to determine if they could generate pressures that would seriously threaten the integrity of RHR system piping. The evaluation in Reference 1 appears to conclude that pressures higher than the reactor coolant system pressure at the time of the valve failure cannot be generated. Therefore, other failure mechanisms, such as RHR pump seal failure, are far more likely.

I was intrigued enough by the argument to attempt an independent evaluation. My evaluation appears to verify the above conclusion.

Because of the similar design of the relevant plant features, this conclusion holds for both the Seabrook Station and the Diablo Canyon Station.

TECHNICAL EVALUATION Consider the situation shown in Figure 1.

The pipe tc the right of the valve (downstream) is initially filled with water of density pC, temperature T, and pressure P. The volume to the left of the valve C

C (upstream) is considered to be much larger than the downstream volume and is nearly filled with water of density pH, temperature T, and H

pressure P. Before the valve disappears, the following conditions H

apply:

TH>TC pH < PC PH>PC V = 0; overall fluid velocity is zero both upstream and downstream.

}

w T0:

Karl N. Fleming, Dennis C. Bley FROM:

M. V. Frank D

SUBJECT:

SHOCK WAVE EFFECT FOLLOWING HYP0THETICAL, CATASTROPHIC FAILURE OF ISOLATION VALVES SEPARATING HIGH AND LOW PRESSURE PIPING IN THE RHR SYSTEM DATE:

August 22, 1986 PAGE:

2 FIGURE 1 CONCEPTUALIZATION OF SITUATION WATER l

LEVEL g

RAREFACTION COMPRESSION e

WAVE PROPAGATION WAVE PROPAGATION T.PC,PC C

UPSTREAM DOWNSTREAM SIDE 90* ELBOW SIDE POSTULATE THAT P

VALVE DISK H

INSTANTLY RUPTURES

, TH

~TH>TC SYSTEM FILLED WITH' WATER UP TO WATER PH < PC LEVEL INITIAL STATE INITIAL CONDITIONS Pg>PC WATER RE GI EN THERE IS NO IN THE DIAGRAM

_ FLOW; V = 0 1412P082286

m

~

Y T0:

Karl N. Fleming, Dennis C. Bley o

FROM:.

M. V. Frank

~

SUBJECT:

SH0CK WAVE EFFECT FOLLOWING HYP0THETICAL, CATASTROPHIC FAILURE OF ISOLATION VALVES SEPARATING HIGH AND LOW PRESSURE PIPING IN THE RHR SYSTEM DATE:

August 22, 1986 PAGE:

3 The sudden rupture of the valve causes a sudden pressure change, which manifests itself as a pressure wave. The wave propagates at sonic velocity. A discontinuity in both pressure and fluid particle velocity exists across the wave front. In fact, for the situation depicted in Figure 1, there are two wave fronts. A rarefaction wave propagates upstream, and a compression wave propagates downstream.

Reference 2 provides the following derivation of the velocity, pressure, and density relationship across the wave front itself.

The pressure and velocity changes across an acoustic wave in one dimensional flow can be derived from the momentum equation au au 9C 3P II) g+u 3x p 3x where u is the fluid velocity, P the pressure, x the distance, and ge is the gravitational constant that is needed for consistent units when P is expressed in pounds-force and p is expressed in pounds-mass. Since the velocity change with respect to distance is relatively small in a straight pipe, the au/3x term is negligible, and Equation (1) becomes au 9c BP g=g (2)

The pressure wave travels at the velocity of sound a; therefore, it will travel a distance (at) in time t.

Let 0 = x 2 at (3) where the negative sign is required for a wave moving in the direction opposite to u.

Substituting Equation (3) into Equation (2), one obtains t a g = - 1 dPg (4) du which yields by integration P2-Pi

= 1 pa W

9Cu2-u1 The plus and minus signs result from the two opposing wave fronts. The terms P2-P1 and u2 - ul refer to the discontinuity of pressure and velocity across each wave front.

1412P082286

rt>

. Karl N. _ Fleming, Dennis C. Bicy TO:

_ *g.

FROM:

M. V. Frank

SUBJECT:

SH0CK WAVE EFFECT FOLLOWING HYP0THETICAL, CATASTROPHIC FAILURE OF ISOLATION VALVES SEPARATING HIGH AND LOW PRESSURE PIPING IN THE RHR SYSTEM DATE:

August 22, 1986 PAGE:

4 If AP = PH-PC is the total pressure drop across both the rarefaction and compression waves, then Equation (5) implies that AP = APH + APC (6) where APH and APC are the pressure drops across the rarefaction and compression waves, respectively.

We note that, for a rigid system, total fluid momentum is conserved.

Mathematically, we can express this concept as AP 9C H + P lau !

(7) au "PH C

c a

and pguH-PClauCl=0 (8) where auH is the velocity _ discontinuity across the rarefaction wave andlauCl15theabsolutevalueofthevelocitydiscontinuity(inthe opposite direction) across the compression wave. Substitution of Equation (8) into Equation (7) results in auH"pa (9) g and lau!*

(10) c a

c Equations (6), (9), and (10) define an intermediate pressure, Pw, at one boundary of each wave. Thus, APH=PH - Pw (11)

APC = Pw - PC Subtraction of Equation (10) from Equation (9) and substitution of l

Equation (11) yields i

j PH+PC Py=

2 (12) i I

1412P082286

q ks' J

s g

1

)

T0:

Karl N. Fleming, Dennis C. BIGy 0

FROM:

M. V. Frank

SUBJECT:

SH0CK WAVE EFFECT FOLLOWING HYP0THETICAL, CATASTROPHIC FAILURE i

0F ISOLATION VALVES SEPARATING HIGH AND LOW PRESSURE PIPING IN THE RHR SYSTEM DATE:

August 22, 1986 PAGE:

6 to dissolution of air near the valve. There are two bounding situations with respect to dynamic shock effects that include two components of fluid (e.g., air and water), at least one of which is compressible (Reference 3).

In the situation in which flow of the air and water is stratified, with respect to each other, the velocity of the front lies between the sonic velocities of either component. Sf nce the sonic velocity in air is lower than in water, the factor, a, in Equation (13) would, in effect, be reduced relative to the all water situation.

In the situation in which the flow of air and water may be thought of as homogeneously mixed, the effective density of the mixture is lower. This would, in effect, reduce the factor pC in Equation (13) relative to the all water situation.

In either bounding situation, the maximum pressure applied to downstream piping generated by a situation that includes an air pocket is less than the all water situation.

In other words, air pockets or voids, such as described in Reference 1, Appendix B, cushion the shock, which is a conclusion we would reach intuitively as well.

REFERENCES 1.

Fauske and Associates, " Evaluations of Containment Bypass and Failure to Isolate Sequences for the IDCOR Reference Plants," FAI/84-9, Appendices A and B, July 1984.

2.

Tong, L. S., and J. Weisman, " Thermal Analysis of Pressurized Water Reactors," Transactions of the American Nuclear Society Annual Meeting, pp. 137-133, 1970.

3.

Graham, B. W., One-Dimensional Two-Phase Flow, Chapter 6, McGraw-Hill, 1969.

I j

1412P082286 l

o G

4-T0:

Karl N. Fleming, Dennis C. Blsy

-[

FROM:

M. V. Frank-SU6 JECT: SHOCK WAVE EFFECT FULLOWING HYPOTHETICAL, CATASTROPHIC FAILURE OF ISOLATION VALVES SEPARATING HIGH AND LOW PRESSURE PIPING-

.IN THE RHR SYSTEM DATE:

August 22, 1986 PAGE:

5 and APH = APC*

The maximum possible pressure applied to a pipe, due to the dynamic shock-effect, would occur at the first bend. This effect is maximized if the first bend is 90', as shown in Figure 1.

We further maximize the effect if-we postulate that the fluid pushed by the wave front must decelerate from sonic velocity to zero, thereby giving all of its momentum to the pipe wall before it reflects in a different direction.

The total pressure, PMAX on the pipe is the sum of the dynamic pressure plus the maximum static pressure that could coexist concurrently with the dynamic pressure.

In the. situation of interest in which a compression wave propagates downstream to a 90* elbow in the RHR system, the maximum static pressure that could exist-behind the wave front is Pw. This is expressed mathematically as PMAX ".4P0YNAMIC + PSTATIC For the compression wave,

• lAuIPa C

C PMAX "

gC Substitution of Equations (10) and (12) into Equation (13) yields PMAX =

+P y

Pg-PC+Pg+PC

=

2 2

Simplifying yields PMAX = Pg (14)

Therefore, the maximu:n pressure cannot exceed the original upstream pressure, which, in this case, is the reactor coolant system pressure.

The foregoing applies to a system completely filled with water.

Reference 1 points out the possibility of formation of an air pocket due 1412P082286