ML20214P508
| ML20214P508 | |
| Person / Time | |
|---|---|
| Issue date: | 11/26/1986 |
| From: | Speis T Office of Nuclear Reactor Regulation |
| To: | Buhl A INTERNATIONAL TECHNOLOGY CORP. |
| References | |
| NUDOCS 8612040241 | |
| Download: ML20214P508 (19) | |
Text
.. :
g%
UNITED STATES 8
o,%
NUCLEAR REGULATORY COMMISSION g
t WASHINGTON, D. C. 20555
\\...../
NOV 2 s agg Dr. Anthony Buhl International lechnology 575 Oak Ridge Turnpike Oak Ridge, IN 37830
Dear Tony:
During the NRC/IDCOR technical exchange meetings concerning integrated containment analyses, it Decame apparent that several major modelling differences between NMC and IDCOR play a significant role in the plant analyses.
Key areas ot disagreement were subsequently identified ana consolidated into the 18 NRC/IDCOR Technical Issues.
It was recognized that further study of these issues was necessary before proceeding with the plant reviews.
Over the past several months draft NRR position papers were prepared for each of the NRC/IDCOR Technical Issues by DSR0 staff in coordination with our Office of Research. These papers summarize our assessment of the status of the issue and how remaining differences between IDCOR and NRC wilI be accounted for in the plant reviews and the individual plant examination methodology. The draft papers were circulated to our contractors and to IDCOR for comment.
Comments receivea were taken into consideration in the preparation of the final version of each paper.
Position papers are now available in final form for the following issues and are enclosed:
Issue 4 - Fission Product & Aerosol Deposition in Primary System assue 12 - Fission Product & Aerosol Deposition in Containment Issue 9 - Ex-vessel Fission Product Release Issue 10 - Ex-Vessel Heat Transfer Moaels from Molten Core issue 138 - Retention of tission Products in Ice Bed Issue 15 - Containment verformance Issue 16 - Secondary Containment Pertormance I
b[t og 41 861 26 NRRC PDR l
l L
NOV 2 6 M 2
Dr.~ Anthony Buhl r
The position papers for the remaining issues, Direct Containment Heating and Essential Equipment Performance, will be forwarded to you when they become This wiil complete our resolution of the IDCOR/NRC Technical available.
Issues.
Sincerely, 5
h 1N Themis P. Speis, Director Division or datety Review and Oversight Office of Nuclear Reactor Regulation
Enclosures:
As stated cc:
B. Morris G. Arlotto D. Houston, ACRS Y
q
- l
-w
-, ~,. -
'*;?.
ISSUES 4 & 12 FISSION PRODUCT AND AEROSOL DEPOSITION IN RCS AND IN CONTAINMENT 1.0 ISSUE DEFINITION The staff relies upon computer simulations to estimate depletion of fission products from the gas phase in both the reactor coolant system and the containment..IDCOR, on the other hand, uses empirical corre-lations which have been fit to the computer simulations and to selected experiments.
IDCOR contends that its correlations are a convenient-
+
and inexpensive method of obtaining. virtually the same results as the staff's computations, while the staff and its contractors do not trust the correlations to be relied upon-in all applications.
2.0 APPROACH TO RESOLUTION IDCOR and NRC agreed upon the following path to resolution.
1.
IDCOR will continue to explore the appropriate range of aerosol model decay constants required to test the applicability of the model for RCS analyses.
2.
IDCOR will assess the empirical model with a detailed (mechanistic) aerosol model.
3.
IDCOR will continue to compare the empirical model against applicable experimental data.
In its draft technical report "FAI Aerosol Correlation" (August 1984), and at IDCOR/NRC Technical Exchange Meeting 3B, IDCOR compared its correlations to the observed pressurizer deposition in the fifth series of Marviken tests, sodium pool five aerosol experiments, and the Containment System Experiment results of BNL-1457.
In the Technical Report 85.2 (July 1985), IDCOR rendered the derivations of its correlations explicit and supplied comparisons between its empirical model and selected experiments and mechanistic computer models. These last comparisons were against an unnamed Fauske and Associates, Inc., aerosol computer code, which was in turn compared to the CONTAIN code.
3.0 STAFF ASSESSMENT OF IDCOR AND NRC CONTRACTOR'S MODELS a
Neither model includes the fonnation and growth of condensation aerosols or accounts for the latent heat of condensation in aerosol formation, 4
and both depend upon subjective estimates of the aerosol properties. The IDCOR correlations, in addition, are limited by the assumption that a mass transfer coefficient is proportional to a power of aerosol particle 4
volume, and they cannot be applied to depletions in which more than two mechanisms are simultaneously causing significant changes. The 4
correlations are intended only for mature aerosols, i.e., those in
r.
-e.
J 2
which the size distribution is not changing by rapid agglomeration, either because of a steady-state dynamic equilibrium between an aerosol source'and a depletion mechanism.or because of the aging and decay of the aerosol.
The correlations employ the device of introducing dimensionless groups as variables, thereby permitting very diverse aerosols.(as calculated by the Fauske. aerosol code or observed in experiments) to be displayed on the i
same graph.' From families of such graphs, empirical functions of the dimensionless graphs are fit. As with most empirical equations, there is no discernable physical significance to the form of the equation (see, for example, the critique of the IDCOR graphic interpolation algorithm in T. S. Kress, " Review of the FAI Aerosol Correlation," March 20, 1986.
A drawback to the correlations is the lack of direct dependence between details of volumes and gas flows and the variables used.
For example, in computing impact depletion, the geometry of the aerosol. flow appears only through Stoke's number. Since factors which do not appear in the correlations cannot affect the computed depletion, these factors will not be correctly accounted for by the correlations.
The weakest aspects of the aerosol codes developed by NRC contractors is in the coupling of aerosol behavior to the thermal hydraulics of the supporting gas, aerosol behavior in pipes, and the aerosol formation and growth processes. To correct these weaknesses, ORNL, as an NRC contractor, is participating in the LWR Aerosol Containment Experiments (LACE) being conducted by Westinghouse Hanford Co. under EPRI sponsorship. The LACE program includes aerosol transport through a piping train and other geo-metries.
4.0 CONCLUSION
The comparisons between the IDCOR correlations and the FAI sectional ccde, combined with those between the FAI sectional code and CONTAIN given in
. Technical Report 85.2 would suggest that the corre at ons will generally l
i predict CONTAIN calculations within a factor of two or three in suspended This is the approximate order of agreement noted by the APS Study mass.
Group in its comparisons of aerosol codes and experiments.
For situations in which an aerosol has a comparatively long residence time in a stationary gas phase, or in which a steady-state is established between the introduction of an aerosol into and depletion out of a gas phase, the IDCOR correlation can be reasonably expected to yield roughly the.same depletion estimates as detailed mechanistic codes.
Such situations would include aerosol depletion within containments which survive until very late. This includes the bulk of the core melt sequences identified by IDCOR.
For accident situations which are dominated by variations in parameters that are not explicit in the correlations' reduced variables, there is little evidence to suggest that the correlation results can be relied upon.
For
-accident situations involving long periods of aerosol depletion within
~,
3 containment, on the other hand, the correlations appear to be adequate.
ihere is no direct evidence that the aerosol correlations used in the MAAP code are too inaccurate for use in sequences in which aerosols are contained for sufficiently long periods of time that they either approach steady state or mature by agglomeration or settling. The staff accepts the correlations for use in such applications.
For sequences in which aerosols are rapidly released to the environment, or are subject to conditions that are not described by the correlation parameters, e.g. sequences with early containment failure, the correlations cannot be relied upon.
,x, ISSUE 9-EX-YESSEL FISSION PRODUCT RELEASE ISSUE 10-EX-VESSEL HEAT TRANSFER MODELS FROM M0LTEN CORE TO CONCRETE / CONTAINMENT 1.0- ISSUE DEFINITION In those accident scenarios in which the reactor vessel fails, high-temperature core debris falls into the reactor cavity where it may interact with structural concrete. The consequences of the thermal and chemical core-concrete interactions may significantly impact containment thermal and pressure loadings as well as the radiological source term released to the environment.
Both IDCOR and the NRC have. developed computer codes to analyze core-concrete interactions occurring during severe accidents. IDCOR analyzes these events using the DECOMP subroutine of the MAAP code (1) and the EQUUS code (2).
DECOMP calculates concrete erosion rate, heat transfer, and gas evolution.
EQUUS is used to calculate chemical reactions in the reactor cavity including rates and heats of-reaction using idealized chemical equilibrium equations.
EQUUS also calculates fission product release. The major document that IDCOR references for their core-concrete interactions is given in Reference 3. The NRC also uses two codes to analyze these effects. The CORCON MOD 2 code (4) is used to compute concrete penetration rates, mass and energy transfer, and the production of flammable and non-condensible gases. The VANESA code (5) is used to compute the generation of radioactive and inert aerosols at the surface of the molten debris pool. These aerosols are then released to the containment atmosphere. VANESA also accounts for the mitigating effect of the scrubbing of aerosols if an overlying pool of water is present over the core debris.
1.1 ISSUE 9 DESCRIPTION Based on NRC and IDCOR technical discussions and the results of limited code comparisons, significant differences were identified between the IDCOR and NRC computer models related to the release of refractory fission products during the core-concrete interaction process. The NRC models have previously included more chemical species in the various reactions than were included in the IDCOR treatment. Also, there are questions related to high temperature chemistry.
Steam and carbon dioxide gases released from the concrete may react with metallic components in the core debris pool such as zirconium and steel as these gases pass up through the pool. These and other high temperature reactions may occur to varying degrees in the debris pool and the uncertainties associated with these reactions could impact significantly on the final outcome of the accident. This is an issue for the NRC analysis methods as well as for IDCOR's. Uncertainties in certain in-vessel phenomena also impact on the analysis of ex-vessel core-concrete interactions. The initial temperature, chemical makeup and mass of the core debris released from the vessel are particularly important input parameters in the calculations of core-concrete interactions, and each of these parameters may vary widely depending on in-vessel behavior.
r 1.2 ISSUE 10 DESCRIPTION This issue is associated with the magnitudes and mechanisns of energy transfer from the molten core debris to the concrete, and to the containment atmosphere and surrounding structures or to an overlying pool of water if one is present.
This issue can potentially impact the mode and timing of containment failure and the chemical forms of the released fission products. This is an area in which major differences exist in the modeling approaches used in the NRC and IDCOR analyses. In general, the NRC analyses involve more heat going into concrete attack with an associated increased production of non-condensible gases and a more rapid pressurization of the containment - but lower containment atmosphere temperatures, particularly in BWR analyses. These differences are partly the result of assumptions made regarding debris dispersal and partly differences in the core-concrete attack models. When water is present in the cavity, the IDCOR approach is to use a critical heat flux correlation accounting for some uncertainty to determine if the debris bed is in a coolable state. In the NRC treatment, heat transfer from the debris bed to the water pool is calculated explicitly and the resulting heat balance determines whether the debris cools or heats up further, 2.0 STAFF ASSESSMENT Based on NRC/IDCOR meetings held in 1985(6,7,8), IDCOR agreed to revise its fission product models to include more chemical species and to compare its DECOMP heat transfer predictions with available test results. Since that time, IDCOR has revised DECOMP and EQUUS to include more species and calculations have been performed demonstrating the use of the revised models. These activities have been reported in various national and international technical meetings (9,10,11). Also, IDCOR has performed calculations comparing the predictions using DECOMP / EQUUS and a version of CORCON-MOD 2 with experimental data (12). An additional comparison with DECOMP and EQUUS is given in Reference 13.
The staff has reviewed the available documentation describing the DECOMP and EQUUS codes (1,9,10,11) and the comparisons between these codes and experimental data (12,13).
While considerable effort has been expended by IDCOR and the NRC and its contractors in developing these detailed, mechanistic codes used for analyzing core / concrete interactions, there are several problems that affect our confidence in using these codes to predict severe accident behavior in real plants.
First, there is a sparcity of data that accurately duplicates the large spectrum of conditions that could be expected during severe accidents. Such experiments are costly and technically difficult to perform, however, there is promise that improved data is forthcoming. The SURC (sustained urania/ concrete) experiments that are being performed at SNL during 1986 and 1987 should provide the most prototypical information to date. The SURC tests
-~
cover a broad range of conditions and are expected to result in data that will be useful for both verification of certain aspects of the present models and for identifying further model improvements.
We do not believe that adequate justification has been provided by IDCOR regarding its criteria for determining coolability of the core debris when there is water present. Future code comparisons with the SURC data will be used to help resolve this issue.
Another problem involves our inability to review the IDCOR DECOMP and EQUUS codes in any detail. These codes have not been released for our study, and the available descriptive documentation (1,9,10,11) is sparse. The comparisons betweer the code predictions and the available experimental data have provided some means for judging the code capabilities in a general manner. In these comparisons (12, 13), overall concrete erosion, gas evolution and temperatures compare fairly well with the measured data although it is clear that further improvements should be pursued and additional comparisons should be made as new experimental data becomes available.
In summary, the staff has performed a limited review based on the available documentation describing the IDCOR models DECOMP and EQUUS and the code comparisons between these codes and experimental data. We have reached a tentative position regarding the present acceptability of the IDCOR models and IDCOR's commitment to upgrade these models after improved data is acquired.
3.0 CONCLUSION
S From our review, we conclude that the IDCOR model revisions to include more chemical species are acceptable. The overall heat transfer predictions appear to be in reasonable agreement with the available data. If the SURC data and the comparisons of the IDCOR codes with that data indicate that the IDCOR treatment of core debris coolability is not supported, we will require modifications to the codes.
This resolution of Issues 9 and 10 must be viewed as conditional and is based on our current understanding of core-concrete phenomenology. This approval is contingent on comparisons being made with new experimental data when such data becomes available. It is our understanding that IDCOR has agreed to this approach. If significant discrepancies are identified through these comparisons, we will require that further model revisions be made to address these discrepancies. A significant step forward in acquiring this data is expected to be made with the completion of the SURC tests later this year and in 1987. In addition, there is a strong international effort involved in research and code development in the core-concrete interaction area attesting to the recognized importance of this subject (14). It is expected that this international interest will ensure that the ongoing improvements in our capability to predict the risk contribution from core-concrete interactions will continue.
l
q
=.
4.0 REFERENCES
1.
MAAP User's Manual, Final Report for IDCOR Tasks 16.2 and 16.3, Fauske &
Associates, Inc.
Ex-Vessel Fission Product Release Modeling, Trans. Am. Nucl. Soc. 50, 2.
~~~
- p. 319, November 1985.
3.
Technical Report 85.2, Technical Support for Issue Resolution, IDCOR Program Report, July 1985.
4.
NUREG/CR-3920,CORCON-MOD 2: A Computer Program for Analyses of Molten-Core Concrete Interactions, Cole, R.K., et al., August 1984.
5.
NUREG/CR-4308, VANESA: A Mechanistic Model of Radionuclide Release and Aerosol Generation During Core Debris Interactions With Concrete, Powers, D.A., et al., July 1986.
6.
NRC/IDCOR Meeting on Outstanding Technical Issues for Severe Accidents, note from T. Speis to distribution, March 7, 1985.
7.
Minutes of the NRC/IDCOR Meeting on March 26, 1985, memo from Z. Rosztoczy to T. Speis, May 9, 1985, 8.
Minutes of the NRC/IDCOR Meeting on April 30, 1985, memo from Z. Rosztoczy to T. Speis, June 21, 1985.
Influence of Containment Thermal-Hydraulics on Source Term Compositions,
- 9..M.G. Plys, P.G. Ellison, R.E. Henry, Third International Meeting on Thermal Hydraulics, October 15-18, 1985.
- 10. Ex-Vessel Source Term Contribution for a BWR Mark I, M.G. Plys, J.R.
Gabor, R.E. Henry, ANS/ ENS International Meeting on Thermal Reactor Safety, San Diego, California, February 3-6, 1986.
- 11. Ex-Vessel Fission Product Release Modeling, M.G. Plys, R.E. Henry, ANS Transactions for the Winter Meeting in San Francisco, Vol. 50, pp.
319-322, November 10-14, 1985.
- 12. FAI/86-2 DECOMP Benchmark Calculatioins for the Core-Concrete Code Comparison Exercise, January 1986.
- 13. Experimental Confirmation of the MAAP Debris / Concrete Interaction Model (DECOMP), R. Sherry, E.L. Fuller, CSNI Specialists Meeting on Core 3-5, 1986.
Debris / Concrete Interactions, Palo Alto, California, September
- 14. Report on a Specialists Meeting on Core Debris / Concrete Interactions, EPRI, Palo Alto, California, September 3-5, 1986, OECD Report, October 1, 1986.
m ISSUE 13B - FISSION PRODUCT REMOVAL IN ICE CONDENSERS 1.0 ISSUE DEFINITION The NRC source term code package (STCP) contains a program, ICEDF, which models the depletion of aerosols within ice condensers. The American Physical Society concluded, however, that the ICEDF code was poorly supported by experimental evidence. The IDCOR MAAP code has not been peer-reviewed, and also is not yet supported by experimental evidence.
The IDCOR MAAP code uses the same computations for ice condenser aerosol depletion as it u:es for depletion in the primary system and in the containment (see Issues 4 and 12), on the grounds that deposition processes are the same in the ice condenser as elsewhere, and differ only in magnitude. This issue differs from Issue 12 only in that ICEDF is not regarded with the same confidence as the NAUA-4 code, while practical considerations have led the staff to use a separate code for ice condensers other than that used for other internal plant volumes.
2.0 APPROACH TO RESOLUTION It was agreed that IDCOR would pursue resolution by three actions:
1.
reporting the contributions of depletion process in the ice condenser calculated by MAAP for the Sequoyah sequences.
2.
discussing any evidence for the specific application of the IDCOR aerosol correlations to ice condensers.
3.
comparing MAAP calculations with the data from future ice condenser experiments being conducted at Pacific Northwest Laboratories to correct the failings noted by the American Physical Society.
IDCOR has not performed either of the first two actions, while the third action must await the completion of the ICEDF code verification and validation tests (FIN B2884), scheduled to be completed during FY8/.
3.0 STAFF ASSESSMENT In Technical Report 85.2, IDCOR stated that in applications of the MAAP code to ice condensers only steam condensation and gravitational settling are assumed to deplete fission products, and that the deposition area upon which these process occur is conservatively modeled.
IDCOR therefore concludes that the effectiveness of ice condensers as fission product cleanup systems is unlikely to be overestimated by MAAP, and is not a significant factor in the calculation of source terms for sequences having delayed containment failure. The staff agrees that there is little likelihood of overestimating depletion in ice condensers using MAAP as
l 2
opposed to ICEDF, assuming that the timing and phenomena the two codes calculate in transporting the fission products to the ice condenser are similar. There is, however, no advantage to substituting intentional conservatism for a technical resolution to this issue,'unless the sub-stitution has no significant effect upon the overall calculations.
4.0 CONCLUSION
There is, at present, insufficient experimental evidence to supporti a conclusion that the NRC Source Term Code Package is acceptably accurate in modeling ice condenser fission product depletion. On the other hand, IDCOR has not supplied sufficient information to the staff to allow the IDCOR calculations of ice condenser depletion to be assessed.
IDCOR maintains that the calculated depletions are likely to be underestimated in MAAP, and are therefore not an impediment to the intended uses.of the MAAP code results. The sequences calculated by IDCOR predict little sensitivity to ice condenser depletion because of high fission product retention as occurring in the primary system calculated by MAAP, and because IDCOR claims that no containment ~ failures will occur prior to the melting of the ice. The differences between theestaff and IDCOR on ice condenser retention, therefore, are of little significance in comparison to those issues (1,2, 4, 6, and 15) upon which IDCOR's reasoning depends.
If resciutien of these other issues and eventual completion of the third item for resolution of this issue" confirms IDCOR's claims, then differences between ICEDF and MAAP will not be significant.
If, however, the resolution of these other issues or the IDCOR sensitivity (uncertainty) study using MAAP introduces Sequoyah sequences for which ice condenserdepletionisimportant(e.g.,containmentfailurepriorto ice-melt), then a comparison between MAAP and the ICEDF verification tests will be necessary to demonstrate acceptable accuracy.
Since it is virtually assured that a sensitivity study will introduce sequences with early containment failure, the three actions contained in -
the agreed " approach to resolution" should either be pursued, or IDCOR should use ICEDF results for all early containment failure sequences.
5.0 REFERENCES
1.
IDCOR Technical Report 85.2, " Technical Support for Issue Resolution," July 1985.
y e,,.-m_m---.
.._x_,__
__-.__m
---mm_..,,,-_,.-,_.-,,,.---.m-, - - - - - -., - - - - - -
.,q l
-ISSUE 15 - CONTAINMENT PERFORMANCE-1.0 ISSUt DEFINITION Risk Analyses indicate that the containment performance plays a dominant role in assessment of risk associated with severe accident. A key insight emerging from the research on accident releases is that the t+ ting of containment. failure is very important in detemining accident severity.
Early failure without other mitigating factors can result in large radio-activity releases, while delayed failure of even several hours reduces the releases significantly. Hence, the concern 6f the containment performance issue is whether the containment conservatively designed for a. postulated LOCA, will withstand the pressure and temperature associated with severe core damage accident.
If not, could the containment fail via large leak-age or due to loss of structural integrity.
If the containment integrity was intact prior _to the onset of vessel failure, knowledge of the time intervals during which the containment leak-tight capability is assured is importent because, if the time interval between reactor vessel failure and containment failure is long, substantial fission product deposition will occur within the containment.
Furthermore, the manner by which contain-ment fails, i.e., gross failure versus failure of penetration, would in-fluence the amount of radioactive' material inside the containment released to outside the containment.
Two basic models can be used in severe accident risk estimation to charac-terize the loss of containment integrity: the " threshold" model and the
" leakage before failure" model. The threshold model defines a threshold pressure, with some associated uncertainty, at which the containment will suffer a loss of its fission product retention capability with the po-tential for significant and rapid release to the environment of the contain-If ment atmosphere which may contain a 1crge amount of fission products.
the containment pressure loading is calculated to be below the threshold pressure, the containment is considerUd to be intact and the offsite con-sequences are, therefore, quite low.- Some recent analyses have pointed out that gross containment failure is not the only pathway to substantial offsite consequences and that significant leakage on the order of 100 volume percent per day (vol. %/ day) or more, may result in substantial offsite releases of radioactive fission products if it occurs sufficiently The leakage-before-failure model provides early in the accident sequence.
a means of accounting for this condition when perfoming risk assessment analyses.
There are, however, significant. uncertainties associated with predictions of leakage from containments.
Introduction of a more realistic leakage model into predictions of containment perfomance essentially requires the development of a new technology. Some elements of this new approach are further along in their development than others.
x, 2
A.
CAPABILITY PRESSURE CALCULATION Attempts.to predict the actual time in an accident sequence at which a containment will fail requires an ability to predict defomation'of the structure,cin::luding its penetrations, well into the-inelastic range and' In the absence of experimental data associating leakage with defomation.
against which to check predictions, the first attempts to assess a contain-ment's likely' performance under severe accident conditions utilized an approach based on the analytical models used for design.
Thh result of the calculation is a pressure, at which it is felt there will be no significant leakage; this has come to be called " capability It is, of course, possible to calculate more than one cap-pressure."
ability pressure for a given containment by utilizing different criteria for terminating the calculation.
If the criteria set fcrth in Section 4' of this paper are used, there is a tendency toward agreement in calcu-lations for a given containment by different parties. This approach was used in IDCOR Report 10.1, " Containment Structural Capability of Light Water Nuclear Power Plants."
A test to failure of a 1/8 size. model, at Sandia National Laboratories, indicated that state-of-the-art analytical methods can predict the onset of failure of steel containments under pressure loadings with good accuracy.
However, another result was that, no significant leakage was detected prior to rupture but when the same techniques were applied for represen-1 tations of the range of equipment hatch designs present in steel contain-ments, some would have led to a leakage failure and some would not have.
' Additional work is needed to predict leakage rates under severe accident Once leakage criteria for all pressure and temperature conditions.
penetrations are developed and verified, performance of steel containment In the meanwhile, it is the during severe accidents will be possible.
staff position that the threshold model be used to characterize containment failure by overpressure.
It should be noted that the response of steel containment to severe acci-dent sequence in which complete core melt is assumed might be vulnerable to temperature-induced failure.
In particular, if core debris reaches the This is an area steel shell, a melt through could occur in a few minutes.
of great phenomenological uncertainty with little experimental evidence.
Therefore, discussion of this failure mode vill await the results of NUREG-1150 sensitivity studies on this % et.
There is much less confidence in the cab 'Ity of analytical models to Loads are resisted predict large defomations of concM.e m.oinments.
All techniques by concrete, reinforcement (or tendons), and steel liner.
that have some potential focus on the development of high local strains in If these strains can be predicted accurately, the onset of i
the liner.
leakage should be predictable. Current thinking is that failure will occur due to a local liner tearing and subsequent leakage. A 1/6 size model of a lined concrete containment will be tested to failure under pressure in late 1986 and early 1987 at Sandia National Laboratories to check the accuracy of different predictive methods.
i l
3 f
B..
RANGES AND DEGREES OF BELIEF There is'high confidence that capability pressure calculations. place a For steel containments, lower bound on containment failure pressure.
there is good reason to believe that analytical methods can give reliable estimates of failure pressure and location, provided that accident temper-For concrete contain-atures are below those at which steel degrades.
ments, there is hope that some of the methods currently being considered
-may be shown to be adequate to predict failure pressure a
. temperature are beyond the current state-of-the-art.
2.0 IDCOR APPROACH IDCOR considers the dominant containment failure mode to be leak-bef failure which occurs as a result of a large strain of the containment i
IDCOR analyses of large dry containments have stated that penet-boundary.
rations generally failed before the reinforcement or tendons.
Eight of the ten ice condenser plants are free standing steel containmen surrounded by a shield building.
ized failure at either weld seams or piping penetrations would be suffi-IDCOR cited the SNL 1/8 model cient to alleviate the pressurization.
steel experiments which experienced strains and displacement increase IDCOR concluded that these designs would likely ex-prior to failure.perience sufficient strain to contact the shield building before fa This localized interaction would lead to localized failures.
For BWR containments,. IDCOR indicated that if venting is initiated, contain-ment failure due to strain of the containment boundary is of secondary For Mark I sequences in which containment venting is not initiated, the failure mode is assumed by IDCOR to be due to elevated impor-tance.
i For temperature in the drywell at a long time after the vessel fai and therefore, it was concluded by IDCOR that high temperature in the drywell is unlikely. Finally, for the Mark III containment, IDC fission product, containment failure does not have any significant in-fluence on the environmental releases.
IDCOR stated that the spectrum of containment failure modes for various containment designs and the influence on environmental releases should be, and has been, considered along with uncertainty analyses with larger con-tainment failure sizes, as part of the IDCOR analyses for the reference IDCOR has also agreed to form a small group with NRC to discuss l
the influences of various containment failure modes as well as SNL experi-plants.
ments.
3.0 STAFF EVALUATION OF IDCOR APPROACH The staff finds the IDCOR commitment to consider the spectrum of contain-i
~
ment failure modes for various containment designs and the influence on l
environmental releases to be acceptable.
1l c
.. _ _. _.... _ _ _. _. ~. _ _ _ _ _ _. _ _.. _ _ _. _ _.. _ _. _.... _ _ _ _.. _ _ _ _, _ _.,..
y 4
Although we conclude that the SNL 1/8 model steel experiment demonstrated that analytical methods can predict with a high degree of confidence the onset of failure of a steel containment, we can not make a similar con-In addition, since leakage criteria clusion for the concrete containment.
for penetrations have not been developed and verified by either the staff or IDCOR, we can not concur on the adequacy of the IDCOR conclusion that the dominant containment failure mode is leak before break.
It is, there-fore, the staff position that until such time-that the leakage criteria have been developed based on the results of separate effect experiments that have been conducted on electrical penetration assemblies, isolation valves and seal and gasket materials, it should be assumed in severe acci-dent analyses that the containment fails upon reaching the threshold pres-sure.
4.0 RESOLUTION OF CONTAINMENT PERFORMANCE ISSUE The capability of the containment building and the timing, mode, and lo-cation of a failure affect the consequences and risk associated with severe Accurate results rely on the ability of the analyst to ident-accidents.
ify the critical areas of the containment and the amount of detail that Another difficult problem con-must be included in an analytical model.
fronting rupture predictions is the prediction of leakage rates as the This could lead pressure, or temperature, or both continue to increase.
to misleading estimates of risk by favoring one possible failure mode (rupture)overanother(leakage).
Finally, since rupture is often caused by highly localized phenomenon that may be difficult to anticipate, analyses with large containment failure sizes (e.g., values used in NRC risk studies) must be undertaken.
For containments that are completely surrounded by an enclosure building where credit for deposition of fission product is assumed, several failure location should be considered in the The analyses to establish the most likely place for' containment failure.
rupture criterion for steel containments should be based on the uniaxial tensile strains at maximum load. This will a yield reasonable estimate of the bursting strength provideri the maximum strain in the containment is accurately predicted.
For concrete containments, if the following criteria are used:
yield of reinforcement for reinforced concrete containments, one percent tendon strain in prestressed containments, deformations will be small enough that no significant leakage would develop and the containment retaining capability is assured.
A 1/14 linear scale model of unlined post tensioned cylindrical containment that has been tested at the University of Alberta, Canada and the concrete segment experiments being performed under EPRI sponsorship are supportive of this conclusion.
[f.
Issue #16 Secondary Containment Perfonnance 1.0 Issue Definition For BWR Mark I and II containments, which are completely enclosed within secondary containment buildings, the releases of fission products to the environment can be significantly reduced by retention of aerosols within the secondary building. Early IDCOR predictions of the fractional retention of fission product aerosols in secondary containments were much greater than those calculated with the NRC Source Term Code Package for NUREG-1150. While IDCOR predicted decontamination factors (DF) well in excess of 10, NRC predictions fell in the range of 1.0 to 3.0.
The magnitude of the calculated DF for a given sequence depends on the mechanisms for aerosol retention and the residence time of aerosols in the building. The NRC and IDCOR models for aerosol retention mechanisms do not differ significantly. Although there is an open question concerning the effectiveness of deposition by impaction, this difference is only a secondary factor in explaining the discrepancies in DF.
The primary source of disagreement lies in the disparate views of three mechanisms which control the residence time of aerosols in the secondary the mode of containment failure, the rate of gas production in building:
core concrete interactions and the mode of hydrogen combustion in the secondary building. The IDCOR models predict containment failure by small leakage, such that the rate of gas release to the secondary building is only great enough to prevent further pressurization of the drywell. Because IDCOR predicts a lower gas production rate due to core-concrete interactions, the rate of gas release to the secondary building is correspondingly slower. Moreover, the IDCOR analyses have indicated that the occurrence of global hydrogen burns which might be expected to periodically expel aerosols from the secondary building, would not be anticipated.
Instead, IDCOR analyces predict that in general the oxygen would be purged from the reactor building or that the hydrogen would burn continuously at the point of exit from the drywell.
The NRC Source Term Code Package calculations for NUREG-1150 embody very different perspective on the mechanisms which control residence times.
The containment failure mode allows for blowdown of the drywell, thereby releasing a gas volume roughly equal to that of the secondary containment Vigorous core concrete interactions result in gas production volume.
rates equivalent to approximately 0.5 secondary building volumes per hour.
Thus, the residence time of aerosols would be expected to be of the order of two hours. Finally, global hydrogen burns are predicted which periodically flush the aerosol inventory of the secondary containment out into the environment.
For PWR plants, the issue relates to retention of aerosols in auxiliary / safeguards buildings following interfacing system LOCAs
4 q
. '(V-sequence)orfailureofcontainmentisolation(gfailure). Early IDCOR calculations predicted DF values in the range of 10 for Zion, while the BMI-2104 calculations predicted a DF of 1.2 for Surry (These two calculations are not directly comparable because the volume of the Zion auxiliary building is about 30 times greater than that of the Surry~
safeguards building). Residence time is still the deciding factor in The differences in residence time are due in part to the calculating DF.
fact that the NRC defines the sequence as a pipe rupture with loss of injection, while IDCOR assumes a more benign pump seal LOCA with slow core uncovery over a 20 hour2.314815e-4 days <br />0.00556 hours <br />3.306878e-5 weeks <br />7.61e-6 months <br /> period.
2.0 Approach to Resolution Several recent calculations have been performed to quantify the significance of various parameters to the overall perfomance of secondary Nildings. As with most issues related to severe-accident policy, there are significant uncertainties involved. However it is possible to briefly characterize the results as follows:
1.
CONTAIN calculations conducted by the Sandia National Laboratory indicate that reactor buildings will remain intact, with the exception of the blowout panels, even in the presence of global e
hydrogen burns.
STCP calculations 'perfomed by Battelle Columbus Laboratories 2.
indicate aerosol DFs in the vicinity of 2 for cases in which global burns occur and 4 when no burns occur.
CONTAIN calculations performed with more detailed nodalization by 3.
Oak Ridge National Laboratory for the Peach Bottom and Browns Ferry plants led to the following insights:
Aerosol DFs are in the vicinity of 10 for cases with multiple hydrogen burns.
With the full coverage fire sprays operating, the DFs for Browns Ferry would be in the vicinity of 40 with multiple hydrogen burns.
Recent IDCOR calculations, documented in Technical Report 85.2, 4.
indicate a DF of about 8 for the case with no burns and no spray operation.
These results indicate that the magnitude of the NRC/IDCOR disagreement on.this subject is significantly less than earlie: calculations indicated.
Calculations based on more detailed and more realistic modeling of the secondary building tend to predict higher DF values. However, these results are more sensitive to assumptions about the failure location
q
~,-
. The potential for release of fission of the primary containment.
products-to the upper regions-of the secondary building, or directly to refueling bay, cautions against placing high reliance on detailed modeling.
The frequency and magnitude of hydrogen burns have an important impact on predicted DF. The NRC staff regards the occurrence of periodic hydrogen burns in the secondary building as being highly likely.
The ORNL calculations confirm the staff's belief that the operation of full coverage sprays would greatly increase the effectiveness of fission product retention in the secondary building.
3.0 Conclusions Aerosol retention in secondary containment buildings is an important mechanism for fission product retention, and should be included in all relevant severe accident analyses.
In general, Mark I secondary buildings are expected to remain intact, However, this even in the presence of periodic hydrogen burns.
conclusion should be verified on a plant specific basis.
In general, the effect of hydrogen burns should be considered in Cases in which no hydrogen burns are performing base case calculations.
assumed should be regarded as sensitivity studies.
The enhancement of DFs due to detailed modelling of the secondary building geometry should be accounted for where possible, but only insofar as due consideration of the uncertainties in primary containment failure location is included.
Calculated DFs should account for sequence-specific calculations of gas production due to core-concrete interactions. Uncertainties exist in those gas production rates due to a lack of complete understanding of core-concrete interactions. Until those phenomenological uncertainties can be resolved and general agreement on gas production rates is achieved, prudence dictates that secondary building DFs should be derived based on gas production rates toward the high end of the range.
Full coverage fire sprays are potentially a very important factor in aerosol retention calculations, and should be modelled for appropriate plants and sequences.
In summary, it should be recognized that the secondary containment building is not designed or operated with the intention of mitigating severe accidents. However, analyses have demonstrated that these buildings can Numerous sources provide significant mitigation of environmental releases.
of potential uncertainties exist, some of which have been discussed above.
Calculated DFs should account for the full range of uncertainty, with due consideration for the effect of pessimistic assumptions.
' v; 1
i,
Furthermore, calculations which predict excessively high DFs should not be used in decisionmaking, even in cases where quantitative uncertainties have been included, a
.