ML20129E822
| ML20129E822 | |
| Person / Time | |
|---|---|
| Site: | 05000447 |
| Issue date: | 07/11/1983 |
| From: | Rowsome F Office of Nuclear Reactor Regulation |
| To: | Miraglia F Office of Nuclear Reactor Regulation |
| Shared Package | |
| ML20127A304 | List: |
| References | |
| FOIA-84-175, FOIA-84-A-66 NUDOCS 8506060609 | |
| Download: ML20129E822 (86) | |
Text
{{#Wiki_filter:, t r s p f t y. ~ 4 / ^;%' h UNITED STATES s' 4-NUCLEAR REGULATORY COMMISSION. f..
- ('
,{g j WASHINGTON, D. C. 2C555 ev 4 %. ['[.
- dUl.11 1383 MEMORANDUM FOR:
Frank J. Miraglia, Jr., Assistant Director for Safety Assessment Division of Licensing FROM: Frank Rowsome, Assistant Director
- for Technology Division of Safety Technology
SUBJECT:
GESSAR-II PRA Q-2 SUBMITTAL A list of Q-2s relating to the GESSAR-II PRA from RRAB was transmitted to you on April 29, 1983. Attached please find a complete list of Q-2s including all contributions from RRAB, RSB, and AEB.to be submitted to the General Electric Comp.any for response. If you have any questions, please contact David Yue (x 28129). f)
- =
Frank Rowseme, Assistant Director for Technology Division of Safety Technology ~ cc: C. Thomas D. Scaletti ( / J. Meyer 7 B. Hardin i J. Mitchell ([] G. Hulman J. Rosenthal 3 B. Sheron D. Bucci W. Butler R. Mattson T. Speis M. Silberberg I. Papazoglou, (BNL) R.~Bari (BNL) W. T. Pratt (BNL) R. Silver, w/o enclosure ~~ 8506060609 841203 PDR FOIA CURRAN 84-A-66 PDR M
p - o LIST OF Q-2s RELATING TO THE GESSAR-II PRA 720.88 Based upon Chapter 16 of the SAR, many transient initiators, including turbine trip and generator load rejection, result in'a reactor
- vessel level 8 within se.coids.
The level 8 would trip the reactor feed pump turbines, thus making the transient similar to a loss of feedwater. event. Provide the basis for your grouping _of these initiators under the turbine trip rather than the isolation' event tree. 720.89 A recirculation pump trip (RPT) is required to successfully mitigate a turbine trip ATWS. Chapter 15 of the SAR shows that a-RPT results in a reactor vessel 1.evel 8 within seconds. NEDE-24222 plots of vessel water level vs. time also~shows the level 8 tri~p'within seconds. Since the level 8 trips the. reactor feed pump turbine, making all turbine trip ATWS similar to the^ loss of feedwater ATWS, why does the PRA present a turbini trip ATWS event tree? Provide the basis for your grouping of these initiators. 720.90 Provide detailed calculations used to arrive at all of the PRA LOCA initiator frequencies, specifically for' those LOCA initiators which have a reduction factor applied due to check valve or automatic isolation valve failures. 720.91 a. What is the basis for the reduction facton_of.'
- E -
In IE . Infork.ati6n Notice No. 83-22;" dated April 22, 1983, some potentially significant problems pertaining to the two-stage SRVs (Model TR-7567F) failing to reseat were described. Discuss the impact on the SRV failure frequency used in the PRA. b. What is the basis of taking I (i ~~ c. Also, in you'r answer to question 720.8, 10 transi ..per---,' year were assumed to arrive at the probability of lj.? d t a-u L.
i l . ;, ; o. 720.92 Provide the operating experience data used in arriving at the recovery of feedwater within 30 minutes for the loss of feedsater and MSIV closure events (see Table D.1.3.1 of the PRA). 720.93 a. Provide a detailed explanation of how the human failure probabilities DHU001, OHU002, and DHUOO3 in the RHR-Suppression Pool Cooling Mode system fault tree were obtained. ~ b. Also, explain why different values were assigned to DHU001 on pages 16.D.3-438 and 16.03-441. c. Provide.a copy crf reference 4 on page 15.D.3-441/15.0.3-448 of the revised PRA entitled, " Technique for Human Error Rate Prediction," by E. E. Vezey, dated July 8, 1981. 720.94 In the RHR-Suppression Pool Cooling Mode system fault tree, explain the basis for the operator error recovery factor given on page 15.D.3-368. 720.95: Provide the basis of pipe break frequencies for those pipes of equivalent diameter 1" or less, which wete not seismically qualified. 720.96 It has been stated that in the event of.a complete station blackout, the RCIC can still be manually operated without using DC power frcm the batteries. Provide the technical basis for the system availability and the operating procedures under those conditions. 720.97 In Question 720.85(c) submitted in October 1982, the electrical and mechanical integrity of containment electrical penetration assembles (CEPAs) were questioned when subjected to multi phase short circuiting under severe accident conditions. The response dated February 1, 1983 was that such prolonged short circuiting would not occur since CEPAs would be protected by redundant over-current devices as required by R. G.-l.63. There are two aspects of the question which have not been addressed: Firstly, R.G.-l.63 does not address environmental conditions beyond that of the design basis accident. Following a severe accident,' the cable insulation and circuit breakers may fail, resulting in multi phase short circuiting of duration longer than those prescribed by current manufacturer's standards. Secondly, in a CEPA module consisting of multiple feedthroughs, e.g., 3 x 350MCM, 7 x #1, etc., multiple short circuiting could occur between conductor-to-conductor and conductor-to ground. The electromagnetic forces accompanying the short circuiting could (_
range from 1,000 to 5,000 lbs. per ft., depending on the size of the conductors. Such dynamic forces could result in brittle fracture of the electrical insulations and, thereby, the breach of the containment under the severe accident conditions. Such ~ failure mode of the containment has not been considered in the applicant's PRA. Discuss the impact of such containment failure modes on the progression of accident sequences and releases of fission ~ products outside the containment. 720.98. I.1 Question 720.79 submitted in October 1982, it was stated that Unresolved Safety Issues (USIs) applicable to GESSAR-II should be' evaluated in,the PRAr The response dated February 2, 1983, stated that those USIs applicable to the 238 Nuclear Island have been assessed in Appendix 1B to GESSAR-II, which was submitted as a portion of Revision 4 to GESSAR-II. However, those responses have not addressed the issues from the probabilistic standpoint, i.e., low probability, beyond DBA conditions. Provide an evaluation of USIs from the probabilistic standpoint for inclusion in GESSAR-II PRA. 720.99. ThecurrentGESSAR-IIdesignoftheboroninjectionpathwAyof'the standby liquid control system (SLCS) is through the lower plenum of the reactor pressure. vessel (RPV). When the RPV water level is below the top of active fuel (TAF) level following an ATWS event, with feedwater pumps tripped off, the boron circulation in the fuel assemblies may be impaired due to steaming and evaporation occurring at or below the TAF level. As a result, the effectiveness of SLCS would also be impaired. Evaluate the impact on the failure frequency of SLCS and core damage probability in the GESSAR-II 238 Nuclear Island design. 720.100. In the GESSAR-II PRA, the frequency and duration have been considered in the loss of offsite power (LOOP) frequency. However, there is another property ( i.e., the degraded grid or partial failure of electrical contacts) which would cause damage to equipment while AC power is not totally lost, which have not been considered. Such failures actually took place at the Millstone plant. Eva'.aate the impact of this failure mode on the LOOP frequency for the GESSAR-II 238 Nuclear Island. 720.101. Does offsite power include the output from the main turbine generator, which is really onsite power of the preferred power kind? What is the relative reliability of the station generator output versus the incoming offsite power? Could the critical load be carried by the onsite generator or by the station service? Evaluate the impact of this distinction on the LOOP frequency for the GESSAR-II 238 Nuclear Island. me
4 720.102
- I.is t all of the possible pathways through which the suppressica pool could be bypassed during a severe accident. Provide a discussic of any pathways not included in the. questions following (Questions 720.103-720.105). The discussien should include a description of the syste=s involved, the accident cenditicas necessary to cause the bypass, the erpected inpact of the bypass on plan: risk, and any design or procedural changes that have been considered by General Ilectric to prevent c: =itigate the ef f ects of 'the bypass.
720.1C3 to the respense.to Question 720.23, have calculations been Referrin "crfor: e3 to specificaily determine the temperatures that might be expected at t'he FSIVs during a severe accicent? If so,'what were the results of these calculations and how were they performed ?(i.e., describe the simolifying assumptions that were made and th.e solution methods) Also, if credit is taken for the FSIVs remaining fully functional under such an environment, provide justification.for taking such credit. The response should include a discussion of data for FSIV leakage rate taken during surveillance testing of these valves in oper.a ing BWRs. If credit is taken for the FSIV positive leakage control system, justify taking such credit at the expected temperatures and range cf ieakage fitw rates.. - Also, in the question response, it is stated that there would be significant radienuclide removal in the FSIVs due to tortuous paths and the source acrosci cenpesitien. ?:cvide further supper: fer this stat,enen accoun:ing for the range of leakage flow rates experienced in past MSIV tes:ing and including or ref erring to experinental data. Discuss any design or procedural changes.that have been censidered by General Ilectric to citi; ate the effects of MSIV, Icakage er f ailure. If your response assunes credi f or addi:1cnal valves other than the MSIVs, identify these valves and discuss the operating environ =en: crpected during a severe accident. the vrive sen-ice qualifications for the cavironment, and the planned nain:enance progra=. 720.104 / In Appendix D.l.7 (Refer to Table D.l.7-2 on pace 15.D.3-387.) it is-stated that two vacuum breake_r,s_.irtserjes,.one.5echanical and one power h [b oper,gedw r i i i . __ \\ 6
- a. Further justification of these probabilities is needed. In particular, t
address the effect of the hydrogen phenomena (local or global detona-i tions or severe burns) described in Appendix 1 of the PRA on the I probability of the vacuum breakers to stick open.
- b. Provide the number, location, and potential flow area of these breakers
( shculd they, remain stuck.open due either to mechanical failure or b ecause c; debris entrained in the atmosphere foliosing a 1.0CA or vessel melt-through. t
r _5_ 72 f c:nt.
- c. Provide a discussion of the potential for significant leakage through the vacuum breakers under severe accident conditions, include reference to available leak rate test data.
- d. Discuss any design or procedural changes that have been considered by General Electric to
=1tigate the effects of vacuu: breaker failure or leakage.- k. j 720.105 Discuss the probability that any of the following lines become a bypass plin.
- a. hydrogen sample linfs
- b. hydrogen mixing system lines
- c. drywell purge lines
~ .v Discuss any design or procedural changes that have been censidered by General Electric to =itigate the ef fects of bypass through,these lines. 720.106 Justify the following assumptions made with regard to the SARCH model . used for predicting the cuantity of hydrogen produced during in-vessel core heat-up and slumping: (a ) The input values used for parameters ISTM and IMWA in NLSOIL, which appear to have been selected to minimize H2 production during core heat-up prior to slumping. (b) The core debris fragments into particles 1.0 h in diameter. (c) The debris i,s assumed to be an intimate mixture of U0, Zr02 and 2 Zr metal. .These modeling assumptions minimize the metal-water r'eaction during core heat-up and virtually eliminate it during the core.-lumping phase. 720.107 Question 720.106 above requires that CE justify certain codeling assu=ptions used in their tiGCH analysis. l'.inor changes in input assu:ptions can signifi-c:ntly influence the predicted generation of hydrogen during the in-vessel heat-up and slu= ping of the reactor core. Hov veuld the probabilities of the various hydrogen phenc=ena discussed in the CE response to Question 720.40 be influenced by uncertainties reSarding itGCH odeling of hydrogen generation during in-vessel core heat-up and slunping? L
i.. [ '720.106 l. l The.GE response to question 720.36 has not addressed our basic concern, namely that H2 phenomena other than the four assumed categories may The probabilistic relationship between these secuences may be cccur. significantly different than those outlined in the FRA. In addition, combinations of burns and detonations may well produce pressures and temperatures in excess of those precicted by considering the four cate-gories separately. GE should further discuss the potential for these integrated phenomena and assess ho,< they may. influence the probabilities in the containment e. vent trees. 720.100 'n'ith reference to GE responses to question 720.37 anc'720.39. Although the_ expression for peak pressure in an H2 detonation occurring in a closed ; system may be verified, it probably does not take into considera-tion the presence of complex geometries which may result in reflective and focusing phenomena. Peak local pressures due to such phenomena may exceed those predicted by the simple relationship used in Appendix I of the PRA. Supply a reference which would justify application of this ex-pression to the GESSAR containment system geometry. 720.110 In the Third Technology Update Meeting, the CORRAL volume model was de-scribed. Unfortunately, this description does.not exactly correspond to 'the sample CORRAL problem provided by GE. Please provide a more de-tailed description 'of the actual CORRAL model used. In addition, also provide a description of how the thermodynamic input was determined for those volumes modeled in CORRAL but not in the MARCH code. 720.111 The response to question 720.40 and 'the containment event tree quantifi-cation given in Appendix D.1.7 are of great help when attempting to un-derstand the containment event trees in Appendix C.16., However, further discussion is needed on the derivation of the event tree probabilities, Speci.fically, provide more information on the following: q /. F.eferring to Table D.l.7-2: g g
- a. Justify thel
/ ' ~(Refer to P' on page 15.D 3-384.) ~ * ~ ~ u. j
- b. Jxp1ain why!/
i I i 1 i i
- c. Justify the failure probabilities for-l 1
) T 'E.- 'o'n~pa 9 e 15. 0. 3-35 6 ). .s, } ~ 9 - 720.111 con:. Referrine to' the response to Question 720.40:
- a. Justify the value assumed for the
]
- b. Explain the inconsistency appearing between the event trees for Cases 3,5 'and 6 and,the conditional probabilities in Table I.4-1.
~ 720.112 For ce.-tain sequences, kater will be present in the pedestal cavity anc the potent'ial for an ex-vessel' steam explosien would exist (refer to Ap-b'hile these steam explosions will pendix H.S of the GESSAR-Il PRA). probably not directly fail containment, they cocid result in enhanced Please dis-oxidation of Ruthenium and Tellurium for these sequences. cuss this possibility and its potential effect. ^ 720.113 Augmented decay heat removal may be helpful in reducing severe accident risk. The Grmans are considering separate dedicated suppression pool heat _ removal systems. Provide a discussion of the potent'jal use of such ~ systems in the GESSAR-II design including descriptions of the systems that have been considered by E and the expected impact of these systems ' on plant risk. Also include a discussion of the potential that augmented heat removal systems may have for removing the limitations of core retention devices discussed in your response to Part 3 of Question 720.83. 720.114 From an inspection of NUREG-0772, it is clear that there is no experi-mental data for fission product release beyond 28000. L' hat release rate constants were used to compute fission product release from core mate. rial at temperatures nigher than 2800C? 720.115 The total quantity of fission products released during the so-called " MELT" release phase is presumably a function of the time at which the core is assumed to slump into the bottom of the reactor vessel. Have, any sensitivity studies been performed related to the assumed point of core slumping-- If so, what is the impact on the predicted release fractions?
g. - _g_ r N 720.116L N, ' - In. ' the ' ES'SAR-II PFA,' up tol of the fission products are predicted to f'l be heldlup in the primary systed.. What is the potential for re-emission j of these fission! products? In prticular, address the possibility of these ]- fission products being released from 'the primary system af.ter vessel failure to the drydell. and_-~subsecuently_ ' om the drysell-via a breach in inithe dryeell wall.-such that the suppression pool is bypa'ssed. Include in -your. assessment the role-of decay heat from the fission products in the re-emission process. The release of fission oroducts ducir.e ir.-vessel core heat-up is a strong fun: tion of the temperatus history in space and time of Ine 'renc.or '. core,: whicn _is predicted by tha -map.CH code. The te perature ' history can be strongly influenc~ed by LURCH input parameters (refer to question'106)yJustify your choice of MARCH input parameters con-trolling core. heat-up. In addition, please describe any sensitivity ._ studies perfotmed related go core heac-up ind fission product release. '720.113-Please provide a mechanistic description of the -in. vessel vaporization . release postulated to occur when the core debris-is in the bottom of the reactor: vessel. This' respons'e should be consistent with the response related to the potential for H2 production during this phase of the core melt.down (refer to question ' 106). Finally, the. CORRAL sample ' problem -_p'r.ovided.by _GE does.not ' appear to include this release. Please explain this apparent discrepancy. 720.119- ~
- Question 720.43
- (Scaling of Tests) asked how the pool scrubbing tests were. scaled with, respect to 'the prototype.
The respense does not fully. ~ 1 resolve the question. Among the unresolved areas are the following.
- 1..Geomstric similarity was not entirely maintained.
This is a-first requirement in conducting tests that can be scaled to the prototype. A specific example of dimensions not scaled is inlet submergence. A second example is bubble. size.
- 2. The size. of large bubbles formed pri'or.to detachment was
-app 4rently _ not correlated with all relevant ' physical parameters. LIf the large bubbles break up in %1 diameter, as GE suggests, - the breakup distance would vary with initial bubble size. Available information does not allow a sufficiently confident prediction of effective scrubbing height to be made for all cases.
- 3. The rise velocity.of bubble swarms is probably a function of swarm size; which in turn would vary with model size.
Specific information which addresses the effect of size scale on swam . rise velocity has not been provided. k_
W ' := -g-t . 720.119 conc.
- 4. Interactions between multiple discharge pathways may.not be
-negiigible. No information on such interactions has been L presented.
- 5. The justification for the applicability of Froude scaling has not been presented.
-Discuss these specific points, and provide the tases for relevent , assumptions. 720.120 ~ . Question. 720.44 (Effect of Iodine Fom on DF) as' ed the effect on k predicted DF if some elemental iodine or. organic iodide is assumed. Also, the potential for formation of organic iodides in the drywsil was questioned. A first response is that the DF for elemental iodine would be comparable to that for. iodide. A GE document (NEDD-22216) is referenced to support this statement. Neither the cited document nor supporting technical discussions are provided, and the response cannot [ be accepted because: o for particles, DF depends critically on particle size, o' for'I, DF would depend importantly.on the -partition,c,oefficient, 2 which in turn would depend on water chemistry. ~ -Thus, the DF for the two assumed forms might be equal for a given set of conditions, but it is' highly unlikely that a general parity would exist. --The existing data base on pool scrubbing of I supports the view-that a substantial DF would be achieved if the fom s! required.to estimate numerical values for specifib. Analyses would be cases. -A second GE response is that 0.03% of core iodine is assumed to pass through the pool in every sequence, limiting DF to 3333. It is agreed that conditions within the RCS do not favor organic iodide formation, and for sequences where the pathway to the pool is through the 'q'uenqhers, little organic ' iodide could be present in inlet gas. The third GE response suggests that conditions in the drywell .(steam / hydrogen, ' low residence time) do not favor organic iodides, but that the-0.03% cited in NUREG-077 is assumed. Two comments are offered. 1. The absence of oxygen in the drywell atmosphere may favor radiolytic formation of organic iodides because oxygen is a radical scavenger that ,__ competes with iodine. It is not clear that dryaeli conditions would limit organic iodides to extremely low values. 2. The 0.03% femation fraction cited in HUREG-0772 has a-questionable technical basis and falls far below many experimental measurements of organic iodide fractions. While it is agreed that organic iodides will always
m: e i720.120 cont. r:present a'small fraction of total iodide (<1%) it is not ev' dent that i the 0.03% value is a realistic estimate, especially when it is the effective limit of iodine emission from the plant, in GE's analysis. Provide justification for the value used and state how organic iodine is' usedtin the CFAC analyses. 'T ~ 720.121 Ouestion 720.45 (Particle Shape Factors for Eu,0 ) ' asked what shape 2 f ac ors apply to tne Eu.0, test particles. Tnt Fesponse'is that the shace f actors are unity'bEcruse efectron micrecraphs (which are provided with question 720.47) indicate the powder to be dispersed was rounded . grains. The larger particles shown are indeed rounded grains, for which --shape factors close to unity would apply. Smaller particles appear to L ' be agglomerates and, hence, could have shape factors different from unity. - The conditions under which the sample were taken are not given. .What is needed is a sample of particles as they'are injected into the pool. .The experimental set-up used by GE invol.ved.a Ldeposition/re-entrainment processes, and it is expected that smaller,
- particles would be agglomerated by this process. The
. photographs shown may or may not depict the particl5s,refore,,the entering the pool. More infomation on where and how the samples were collected is necessary to help resolve questions of how well the pictures'shown re-present the aerosol that actually entered the pool. Provide the einfomation ' requested.
- For inertial deposition and sedimentation, shape factor is not an
-important issue in interpreting the experiments because, in principle, .the.. cascade impactors give aerodynamic sizes which can be put into the model directly. -However, for diffusional deposition one needs to know
- the effective density in order to estimate-the actual size.
Uncertainties' in shape factors for smaller particles introduce ~ uncertaintyLinto predictions of diffusional capture. Provide a . discussion, and bases therefor,'of the effective density and > uncertainties in shape factors. 720.122 1 'Queslion720.46(ParticleSize) asked for clarification of which of several. stated particle sizes was used by GE in the theoretical . predictions. GE's reply clarified the picture by stating that none of .the ranges or averages was used, but that impactor data were used for ~ each-experiment. Unfortunately the impactor data are not given. It is necessary in assessing the.models and' experiments to obtain the cascade impactor data. Data needed are: (1)Impactorflowconditions,(2) Calibration procedure, (3) mass of material.on each stage. Since a soecific size is input to the analysis of each experiment, provide the =requ'ested.information for'each experiment. L;
720.123 -Question 720.48 (Deposition and Reentrainment) was ' directed at particle deposition and reentrainment within the recirculating flow system used to keep Eu 0 par The GE response deals with processes inside the p5ol; ticles airborne. 22 the. question appar'ently was not understood by GE. The -crux of the question deals with the possibility that the particles entering the pool were larger than measured either by deposition or 2agglomeration of small particles downstream of the sampling point. The question.is important-because particle size is a critical parameter in scrubbing efficiencies. Relatively small errors in~ particle size could. . lead to' erroneous interpretations of scrubbinc efficiency. Provide a ciscussion of how the upstream sampling methodology realistically Jacceented for pctential pirticle aggicmeration by 1depcsition/reantrainment processes. 720.124' ~ ' Question 720.49 (Dilu'ter Function) dealt with the use of diluters to obtain impactor samples. The GE explanation for the outlet. sample is' adequate. However, the diluter used for the input impactor is not adequa tely ' described. Specifically, provide a destniption of-how was the flow rate of dilution air measured and controlled-to maintain isokinetic flow in the sampler tube. A schematic diagram with a narrative that describes the experimental procedure'would be helpful. t - This. issue is important because isokinetic sampling conditions would be required to obtain a representative sample of inlet aerosol!' The single sentence describing the input isokinetic sampler.to the extent that we f may have misinterpreted what was actually done, describes a procedure which would be unlikely to produce accurate results. From the description in the response, we infer that the input sampler is not -isokinetic. 720.125 - Question 720.50-(Sampling'Line Dimensions) addressed particle deposition in sampling lines. The-issue is important because such deposition is size dependent with larger particles being more effectively deposited. Such deposition would tend to make the impactor rich in small particles, which would in turn -lead to erroneous interpretations of the scrubbing efficiency. Deposition in the outlet sample was measured by GE to be -less than 1%, a satisfactory answcr. This result is not unexpected because only _the most penetrating particles would escape from the pool. Unfortunately losses in the inlet tube were neither measured nor theoretically analyzed. Inlet particles would be larger than those in the outlet and, therefore,: susceptible to sampling line loss. The lack -of data on inlet line losses introduces uncertainty into the particle size measurements; a technical analysis to quantify this uncertainty is requested. ,,,-n. .n
.y_
-,,,.-.--w
C.L. '720.'126 Question 720.52 (Recycle Flow) asked about fTow in the recycle stream. 'The;GE ' answer ~ is adequate and points out that an insignificant amount of material remained in the loop after 10 experiments. ~ The GE_ finding of r Liittle' residual. material supports the postulate that particles were deposited and1reentrained in the loop. Such processes usually cause Lparticle' agglomeration.which must be quantified if the size of particles -enteringithe pool is 'to be accurately known. Information supplied by GE has not-directly addressed the question of aggicmeration in the aerosol circuit. 1The recycle flow complicates interpretation of the aerosol size fdistribution d Leffect of1the'ynamics. However. it would be productive to analyze the loop and its pump on the aerosol size distribution. Provide such.an analysis.- T720.127: ~ - Question 7'20.53 (Cascade impactor Calibration) asked for a definition of . data columns of Tables 1-1 and I-2. The GE answer adequately describes the listed' particle diameters-as calculated cut diam!ters for "each impactor stage. A detailed look.at the calculations r'aises a question regarding ~the gas flow rates used in the impactors.. The manufacturer's operating manual suggests a flow rate range up to 0.75 cfm, but GE used . flow' rates of 1.0 and 2.0 cfm. The use of flow rates -higher than the Enormal range may introduce uncertainty due to calibration shifts and
- reentrainment.
A technical. discussion that quantified impactor ' calibrations and reentrainment at. higher-than-design flow is necessary, cThe -issue is imp 6rtant because the effects listed could cause a bias towards :smail particles leading to erroneous interpretation of scrubbing ' 2 efficiency versus particle size. 2 720.128 -Question 720.55 (Material Balance) dealt with the possibility of making a material. balance. The GE answer indicates the measJrements do not permit this. The GE answer also alludes to differences between the inlet 'sampi.e and.the bubble formation orifice. More information on the " isokinetic upstream sample (filter or impactor) is necessary to assess the degree to which the sample was representative of aerosols entering .the pool..If the orifice and the sample location are different then serious questions are raised concerning the applicability of anythino measured by.the' inlet sampler. 9 0
e 720.129 Question 720.56 (Entrance Effect) was directed at an entrance effect (a ~ DF due to removal processes occurring during bubble formation) that was included in the theoretical model for the single bubble tests. It is agreed that removal processes would occur during bubble formation. Three questions remain: 1. Did an integration over the size distribution yeild a DF of 10.0 for all of the experiments,-or is 10 merely an average assigned by technical judgment?
- 2. Why was an entrance effect '
rot includec in tne edei used to predict the DF in the verification test?
- 3. W.,a; particle size dis:ribution was used for scrubbing predicticr.s for the rising bubble, e.g. was the depletion of large particles by the entrance effect specifically accounted for, or was the factor of 10 DF applied to the entire aerosol mass?
~ ~ 720.130 Question 720.57 (Particle Size Spectra) dealt with size distributions measured and used in the predictions. GE notes that actual. measurements,. were used. The example given for Test 12/2 shows that 0.1 of the mass is smaller than %2.5 pm, whereas in Table 15 DA.1-5 some 0.4 of the mass is smaller than 2.5 pm. This variation indicates that the particle size spectrum varied significantly f rom test to test. Since artiy one type of - aerosol was used, this variation.is not expected. Futhermore, without a - measured distribution for each test, no checks by the staff are possible. See also the comment to question 720.45. Provide an explanation of what mechanisms cause this large variation. 720.131 Question 720.58 (Particle Size for Cunningham Factor and Diffusivity) dealt with particle sizes an.' densities used to predict the Cunningham slip factor (CSF). The GE answer asserts that the Cunningham factor depends on particle size, mean free path, and partici.e density, that ' ave; age values were used, and inat use of the average CSF may result in an average increase in predicted DF by %20%. Two coments are offered. (1) CSF does not depend explicitly on particle density as GE has suggested. Tne problem is that the use of cascade impactor data for computing CSF and diffusivity requires one to choose a particle density in order to compute effective particle size. Predicted particle diameters (from impactor data) would vary by factors 1 to 1/'t.8.
- Thus, predicted diffusivities could be in error by as much as a factor of 2.8,.
much as exp (2.8)!g to predicted errors in DF due to diffusion of as potentially leadin = 5.3. (2) The use of average values of CSF for the whole particle size spectrum introduces unnecessary errors into DF predictions. Therefore, CSF should be computed for each discrete particle size range. The problem is that particles reaching the pool surf ace are not those for which inertial depcsition dominates the m:cwo t
r -14. 720.132 Qv stion 720.59 (Predicted DF Versus Particle Diameter) asked for data to support a GE statement that experimental data supported the predicted trend of DF versus particle size. GE has submitted a graph of DF versus particle size. Inspection of the " theoretical model" curve of the figure shows that it has a minimum DF of %20 at a particle diameter of so.2 m. The bubble scrubbing models described by GE do not predict a minimum DF this high,and it is evident that the " entrance effect" discussed earlier was used. It appears that a constant DF of 10, applicable to all particle sizes was used. The introduction of a constant DF of 10, if this was done, is highly questionable on technical grounds. Therefore, provide detailed description of how the theoretical ~ mcGel line was computed.. 720.133 Q'uestion 720.61 (Particle Growth Due to Superh'eated Steam) asked for information to support a GE statement that superheated steam could cause particle growth. In GE's response, it is pointed out that bubbles quickly reach thermal equilibrium with the pool so the ste'am becomes saturated. Saturated steam can then condense on particles, as illustrated in the Figure supplied by GE. It is agreed that soluble aerosols can grow in humid atmospheres. F.ealistic predictions of particle growth require knowledge of particle solubility and of bubble humidity, factors which are not particularly easy to' evaluate. Two aspects of particle growth not addressed in the GE response are: (1) soluble particles can grow in humidities less than 100% whereas insoluble particles require a humidity >100% to nucleate water. (2) The degree of growth depends critically on the actual humidity inside a bubble. Provide a discussion of your views on both aspects. 720.134 Question 720.62 (Thermal Effects on Scrubbing) asked why diffusiophoresis can be neglected compa. red to thermophoresis, and how the low temperature tests can be considered conservative from a temperature effect. The GE answer states that diffusiophoresis (which could impede particle capture) we'.'id cease af ter %0.1 seconds, hence would not be a significant factor in limiting scr'ubbing. While it is agreed that bubbles are expected to reach thermal equilibrium quickly, this does not stop the inward flow of steam. Vapor-liquid equilibria dictate that steam will evaporate into rising bubbles because total pressure is ' decreasing with height. The magnitude of the evaporation velocity depends strongly on water temperature, and for a saturated pool -the effect is maximum. Numerical calculations show the evaporation can be a significant impediment to the capture of particles in the intermediate size range (0.01 to 3 um). Therefore, provirJe an analysis of' steam evaporation into rising bubbles and its effect on particle capture. We consider any "first principle" model deficient without specific inclusion of this effect. h..
I _w. s t 720.135 Question 720.63 (Run to Run Variations in DF) asked for an explanation of DF variations of a f actor of 4 for tests done under similar conditions. GE responds that the tests had different particle concentrations and particle size distributions and further that typographical errors were p.msent in the original table. The attached table shows that for the runs done on 12/11 and 12/14 the particle 3 concentrations were 4.34 and 1.38 g/m respectively. The measured DF's were quite.similar (2945 vs 2270) indicating (that particle concentration, had little effect on DF. For the other run 12/15), concentration was not measured and the DF was 928. The particle size data alluded to in the response are not given, so no. judgment with resoect to particle size effects can be made.. In the absence of particle size data it is not possible to assess what part of the variations in measured DFs is attributable to differences in experimental conditions and what is due to experimental error. See the two previous requests for all the data which is neces.sary to allow the staff to evaluate the tests, and-justify your conclusions if particle size data is not available. 720.136 Question 720.64 (Bubble Shattering Distance) was aimer at distances required for bubble shattering. The GE response notes that bubbles shatter when the bottom overtakes the top, that Froude scal.ing applies, and that breakup distance is geometrically similar'. No ment of the apparent conflicts on pages 49-C43 and 4g-C45, as requ, ion is made ested in ~ the question. From page 49-C45, it is stated that typically,12-inch diameter bubbles shatter about 18" above the vent (5.44-inch. diameter). If this result is scpled to the plant, then the height to shattering is (18/5.44) x 27.5 e 91 inches or 7.6 feet. If this shat'.er distance is correct, then the effective scrubbing height would be greatly diminished because the submergence of the horizontal vents varies from 8.4 to 13.5 feet (GE information suppli,ed with question 720.65). A discussion of how one estimates effective scrubbing heights for the horizontal vents should be provided. 720.137 Question 720.65 (Submergence of Vents) dealt with submergence of vents in tests and submergence in the plant. The response indicates that rise velocity and bubble size of stable bubbles would be the same in small and full scales, so that extrapolation to 14 feet is straight forward. This response fails to deal with the important question involved here, the distance required for the large bubble to shatter into stable bubbles. If the break-up distance is scaled linearly, as suggested by GE in response to question 720.64, then in the plant, break-up heights of the order of 8-feet may be predicted. This is a large fraction of m
i -16. 720.137 cont. I submergence (8 f t to 13.'5 f t.). If the tests had used' geometric i similarity a minimum submergence of (5.44/27.5)(8) = 1.6 ft. which is very close to the break-up distance quoted on page 49-C45. Please address the question of effective scrubbing height for all applicable discharge conditions. GE is using the model to validate the tests and the tests to validate the model in this answer. This is unacceptable. 720.13S ~ Cuestion 720.E5 (Bubble Rise Velocities) asked for clarification of rise vslocities presented in Figure lj DA.2.2. The response is that the carve through the data is an " eye bali" trend fit, nothing more. Other parts of the cuestion that dealt with the significance cf the data, and how it was used, were not addressed. The significance of the data set remains in question. The:aforeprovide a discussion of both topics. 720.139 Question 720.59 (Entrainment from Pool) asked how entrainment carryover con be neglected.. The response is that the bursting bubbles are expected to be relatively free of suspended particles _. This response apears to be based on a surf ace cleaning theory that is not well documented. Rather than be particle-free, it is possible that the interface would contain higher than average concentrations of particles. Interfacial crud and foams that contain concentrated impurities are c . commonly found in mass transfer processes. While entrainment would not be expected to be significant for low DFs (< 100), carryover could dominate if large DFs (>10') are claimed. In view of the above, provide further justification for your conclusions. O 720.140 ~ Osestion 720.70 (Csl Experiments) asked information on the tests done with Csl. The response indicates that experimental difficulties prevented valid results from being obtained, a satisfactory answer. The overall conclusion, that resu'.ts were consistent with DFs of 2.5 as would be predicted for 0.1 - 0.2 in particles, appears to be reasonable. Ths, fact that minimum calculated DFs are in the range of 2-5 is as expected, but is inconsistent with the theory line presented in response to question 720.59, which shows a minimum DF >10. Please discuss this inec sistency. 720.141 Concerning Question A (Table 7.2-1 Rev. !!), based upon co munications with D. Hankins, the answer provided to this question on the direction and magnitude ef' the changes between Rev. 2 and Rev ll on early and latent f atalities is not exactly correct. Attached are ccpies of Table I } 7.2-1 f rom Rev. 2 and Rev.11. Indicate in the appropriate places the I corrections necessary for mistyping, and discuss the directicn and expected magnitude of any changes made between the twa analysas.
GENERAL II.ECTRIC COMPANY' _PROPFl2TARY INTO??.ATION Class III Table 7.2-1 ESTIMATED CORE DAMAGE AND RISK COMPARISON Risk Assessed Frecuency Early Latent of Event Fatalities Fatalities Per Reicter Per Reactor Per Reactor Event Year Year Year I. CORE DA.". ATE [; RSS 3'nR/4 Mark I a a -5 -2 6, composite sit'e s4x10"# six10 %5x10 s ~ ~ RSS 3WR/4 Mark I -5 -6 -2 e site 16C s4x10 1.2x10 1.ix10 3WR/6 Ma.rk III C -6 -5
- 9 site f6 5x10
,0 2x10 II. U.S. NATURAL 3ACKGROUND Continuous 0 814' RADIATION t "With WASE-14 00 Methods (calculated from the reported curves). bThe total accident-caused f atalities over. the, lifetime of the exposed cooulation or the calculated excess cance: s in the same pop'littion from one year of background radiation. u Co$puted with the GE CRAC Code. f c 6 + 9 4 I s b 15.D.3-143
L Tablo 7.2-1 ESTIMATED CORE DAMAGE AND RISK COMPARISON- ~ t Assessed Frequencv g, rygn.1*
- h. h Ge5 Year) j s
- Per Reacto-
? * -ly, , Latent Year Fatalitids
- Fatalitiesb Event s
I; CORE DAMAGE, RSS EnR/4 Mark I .a a E co=cesite site 44'x10! 41x10 $c ~ ~ ~# N5x10.c 44x10 " 2.4x10 ) 2.5xiO~'. ~ F.SS 3nR/4 Xark.I -5 -u 8 site f6C N4x10 .7.Ex10 2.1x10.2 SWR /6 Mark III -6 -5 5x10 40
- 1. 75t10 6 site f6C II.
U.S. NATURAL 3ACKGROUND Continuous 0 .814'- RADIATION " Wit. h WASE-14 00 Methods (calculated from the reForted curves). ~ The total accident-caused fatalities over the lifetim6 of the exoosed ecoulation c the calculated excess cancers in the same' po[oulation from one year of background radiat'on. Computed wit 51 the GE CRAC Code. ~ j s e
- e g
4 -w e. h 15.D.3-143
r i.. 720.142 Nowhere does GE give an example of exact input values for their DF model and an example calculation using those inpu'ts. I Tc evaluate the GE model the staff needs the followingi 1) Fin'ai equation for DF with all coefficients defined. 2) A comolete listing of all' input to that equation. 3)* An evaluation for a single set of input of spherical and ellipsoidal bubble models. 4) An evaluation, including an input listing, of DF versus particle size for a size range from 0.01 pm to 10pm. 720.143 The GE response to Questions 720.77 and 720.78 regarding uncertainties in the GESSAR-II PRA is not acceptable. The staff believes that treatment of uncertainties is essential. Realistic point estimates with uncertainty bounds should be used for making L severe accident decisions based on a proper level of' conservatism. GE should prepare a plan describing a proposed uncertainty analysis for presentation to the staff in the near future. The plan should include the assessment of the major types of uncertainties affecting the accident sequences contributing significantly to core damage probabiliti%s and overall plant. - risk. Four types of uncertainty that shoold be addressed include: l (1) statistical undertainties in component and human failure rates and in other analysis input parameters, / (2) approximations in physical phenomena modeling, (3) errors in completeness in considering possible failure modes, and m (4) arithmetic errors The propegation of these errors should also be addressed. References for uncertainty analyses: Zion and Indian Point risk assessment. A l O-
r t .c > o 4 720.144 In the conceptual design for the advanced BWR's developed by an advanced engineering team comprising General Electric (United States), Toshiba and Hitachi (Japan), Asea Atom (Sweden) and Ansaldo Meccanica flucleare (Italy), an electrically (as opposed to hydraulically) operated high. speed scram CRD has been recommended. This would provide higher scram reliability and better load following by allowing unrestricted control rod operation at high powers. Please.pFovide an assessment of potential reduction in the core damage probability and overall plant risk by adopting the new CRD design as compared to the current CRD design in GESSAR-II plants. 34 e 6 e 6 e O W e 4
i i BROOKHAVEN NATIONAt. LABORATORY MEMORANDUM DATE: May 5, 1983 y I. A. Papazoglouh TO: N. kNK! nan, K.h% FROM: Shiu, R. Karol
SUBJECT:
Status of GESSAR-II PRA Review The purpose of this memorandum is to present the status of the GESSAR-II PRA review, as well as the preliminary results obtained to date. 1. Introduction One of the objectives of the GESSAR-!! PRA review (FIN A-3366-2) is to evaluate the frequency with which core donage might occur as a result of "in-ternal" accident initiating events in a GESSAR-II plant. To this end, the B'NL review includes an evaluation of the qualitative and quantitative analysis of the PRA as well as independent reassessments of several important items. The review includes an evaluation and revision of the plant modeling in the PRA. The plant modeling includes the identification of the accident initiators, the safety functions, the frontline systems, the support systems and their respec-tive success criteria. This part of the review has been completed and the major conclusions are summarized in Section 2 of this memorandun. fhegroupingoftheaccidentinitiatorsintogroupsofinitiatorsim-posing the same performance requirements (success criteria) to the various frontline systems, has been reviewed and revised. The frequencies of the various initiators have also been reassessed. Section 3 of the memorandum summarizes the results of this part of the review. The system fault trees have been examined, modified whenever necessary and requantified. This part of the review is summarized in Section 4 of the memorandum. Finally, some of the event trees have been qualitatively examined and requantified using the equivalent core damage fault trees. This part of the A y equD g-Q s
O Memo to: I. A. Papazoglou May 5, 1983 Page two review is still in progress. Prelininary results of the review are presented in Section 5 of the memorandum. 2. Plant Modeling The plant modeling part of the GESSAR-II PRA contains the identification of the initiating events that can lead to core damage, the safety functions important to preventing or mitigating core damage and the systms directly performing each of the safety functions. These systems are referred to as frontline systems. In addition, the plant modeling includes the identifica-tion of the support systems for each frontline system, i.e., the systems re-quired for the function of the frontline systes. 2.1 Safety Functions and Frontline Systems The sa'fety functions important to preventing or mitigating the conse-quences of core damage following an initiating event are given in Table 2.1. These functions can be further subdivided for the GESSAR plant into the func-tions given in Table 2.2. Each of the functions in Table 2.2 is directly perfomed by one or more frontline systems. The frontline systms for the GESSAR plant are given in Table 2.3. The support systems necessary for opera-tion of the frontline systes are given in Table 2.4. ~ 2.2 Initiating Events and Success Criteria' for the Frontline Systems The GESSAR-II PRA considers three general classes of initiating events: 1) Loss-of-coolant accidents (LOCAs).
- 2) Transients with successful scram.
- 3) Anticipated transients without scram (ATWS).
The LOCA initiators are further subdivided into three groups (Large, Intermediate and Small LOCAs) depending on the size of the break which in turn, detemines the success criteria for the frontline systes. The success
- criteria for the frontline systems for the LOCA initiators are given in Table 2.5.
a. ) o Memo to: I. A. Papazoglou Mcy 5,1983 Page three The transient initiators with successful scram are further subdivided i into five groups that are characterized by similar responses from the frontline systems. 1. Planned reactor shutdown. 2. Turbine trip. 3. Isolation of the reactor vessel (includes MSIV closure and loss of feedwater). 4. Loss of offsite power. 5. Inadvertent open relief valve (IORV). The success criteria for the frontline systems for the transient initi-ators are given in Table 2.6. Theanticipatedtransientswithoutscram(ATWS) are subdivided,'in the GESSAR-II PRA, into four groups corresponding to the four categories of transients (groups 2 through 5 above) for which an unsuc-cessful scram is possible. The success criteria of the frontline systems for the ATWS initiators are given in. Table 2.7. The success criteria used in the GESSAR-II PRA are determined on a " realistic" basis and do not necessarily correspond to conservative analysis of the SAR. The most important dif ference between the success criteria for the GESSAR-II systems and those used in PRAs of other BWRs (like the Reactor Safety Study BWR or the Limerick Generating Station) is the different require-ments for the Automatic Depressurization System (A05). The ADS has eight Safety Relief Valves (SRV) that can be used for automatic depressurization. In all the cases, three SRVs are enough for the depressurization of the plant. Thus the success criterion for this system is successful operation of three-out-of-eight SRVs. This criterion should be compared with the four-out-of-five SRVs for the Reactor Safety Study BWR and the two-cut-of-five for the Limerick Generating Station. In general, the frontline-system success criteria used in the GESSAR-II PRA are considered reasonable. This BNL conclusion, however, is not based on ,.p ,--.e..
o. D Nemoto: I. A. Papazoglou May 5, 1983 Page four independent analysis, but rather on information published elsewhere. In par-ticular, approximate calculations for injection requirements for LOCAs, and information in the reports NEDO-24708 for transients and NEDE-24222 for the ATWS were used. There is one particular instance in which the BNL review disagrees with a GESSAR-II PRA success criterion. Namely, in the case of an intermediate LOCA, the successful operation of three SRVs was assumed instead of two because two SRVs are sufficient to depressurize the reactor vessel for only a fraction of the breaks in this LOCA category (see Table 2.5). 3. Frequencies of Accident Initiators As it was noted in Section 2, the GESSAR-II PRA considers three ' general classes of initiating events: 1. Loss-of-coolant accidents (LOCAs). 2. Transients with successful scram. 3. Anticipated transients without scram (ATWS). The assessment of the frequencies of the initiators in each class is dis-cussed in the following three subsections. 3.1 Loss of Coolant Accidents The GESSAR-II PRA states that the frequencies for the various types of LOCA initiators were determined on the basis of the Reactor Safety Study. These frequencies were assessed for three categories differentiated by loca-tion: inside' the drywell (drywell LOCA); inside the primary containment, but outside drywell (containment LOCA); and outside containment (ex-containment LOCA). According to the GESSAR-II PRA, the frequency of each type of LOCA was calculated based on the length of the piping which has the potential of causing that type of LOCA. Containment and ex-containment LOCA initiating event frequencies include the probability of failure of the check valves and/or automatic isolation valves that are located upstream of the break. t e
a. ? Meno te: I. A. Papazoglou May 5, 1983 Page five Table 3.1 summarizes the LOCA frequencies used in the GESSAR-II PRA. Since detailed information about length of piping was not available to the BNL re-viewers, a reevaluation of the LOCA frequencies was not performed. The values used in the GESSAR-II PRA appear, however, reasonable. The effect of the potential variations in the frequencies of LOCA will be addressed in the importance and sensitivity analysis part of the review. 3.2 Transients with Successful Scram The transient initiators with successful scram have been subdivided into five groups: 1. Planned reactor shutdown. ~ 2. Turbine tr1p. 3. Jsolation (includes MSIV closure and loss of feedwater). 4. Loss of offsite power. 5. Inadvertent open relief valve (IORV). Each of the five groups represents a collecticn of transients which demand similar responses from plant systems. A summary of the GESSAR-II grouping is given in Table 3.2. In the BNL review, the assessment of the frequency for each of the five transient initiators was based on the grouping of a list of transient initiators.for BWRs ' contained in an EPRI report (NP-2301). Thirty-seven transient initiators are identified in this report (see Table 3.3). There is a substantial difference between the grouping of the various transients in the GESSAR-II PRA and the corresponding grouping in the BNL review. The major departure of the BNL grouping fram the GESSAR-II PRA lies in the treatment of events noted by an asterisk in Table 3.4. Under the GESSAR-Il grouping scheme these events are classified as turbine trip transients. The GESSAR-II SAR (Chapter 15 - Accident Analysis) indicates, however, that these transient initiators cause a reactor vessel " Level 8" signal which in turn causes a trip of the main feedwater turbine in less than ten seconds.
Memo to: I. A. Papazoglou May 5,1983 Page six This particularity of the GESSAR-II design requires the grouping of these ten transient initiators into the isolation group. In the GESSAR-II PRA, the frequencies of transient initiators were based on General Electric data modified to reflect BWR-6 standard plant design. These frequencies are given in Table 3.5. An independent' assessment was conducted to determine point values and associated distributions for the frequency of each of the transient initia-tors. A detailed discussion of the methodology employed in this evaluation. can be found in a draft BNL report (BNL-NUREG-31794). The data used in this assessment for all transients, except the loss of offsite power, were obtained from the EPRI NP-2301 report whereas the loss of offsite power data came from an NRC report by R. F. Scholl. Results of the BNL study are given in Table 3.6 where the GESSAR-II PRA initiator frequencies are also given for compari-son purposes. The values in the table reflect the BNL grouping of the tran-sient and are representative of the population of the operating BWRs. Va ri a-tions of these frequencies and their effect on core damage probability will be addressed in the sensitivity analysis part of this review. 3.3 Anticipated Transients without Scram (ATWS) If the Reactor Protection System fails to scram the reactor after a transient initiating event, then an ATWS results. Four groups of ATWS initiators were considered in the GESSAR-II PRA. 1. Turbine trip ATWS. 2. Isolation ATWS. 3. Loss of offsite power ATWS. 4 10RV ATWS. The specific transient events that were grouped in these four groups correspond to those grouped in the corresponding groups of transients with successful scram (see Table 3.2). In the BNL review the group " Turbine Trip" was eliminated for the following reason. A successful mitigation of turbine trip ATWS in GESSAR-II requires a recirculation pump trip (RPT). As a result
Memo to:
- 1. A. Papazoglou May 5, 1983 Page seven of the RPT, a reactor vessel " Level 8" trip signal is generated resulting in the tripping of the feedwater turbine (see SAR Chapter 15); NEDE-24222 also shows that the reactor vessel Level 8 is reached in i few seconds. - All the turbine trip ATWS result, there_ fore, in a feedwater isolation and hence the event, trees that delineate the accident sequences for isolation ATWS are applicable.
The ATWS initiator frequencies were derived from the corresponding frequencies of the transient with successful scram. The BNL isolation ATWS frequency is, of course, equal to the sum of the turbine trip and isolation transients. The frequencies of the ATWS used in the BNL review are given in Ta bl e 3. 6*. It is noteworthy, here again, that the ATWS frequencies are derived from data (transient with successful scrams) that are representative of the total -population of the operating BWRs., 4. System Fault Trees The system fault trees are contained in Section 0.2 of the GESSAR-PRA. The following GESSAR-II systems were analyzed by system fault trees: High pressure core spray (HPCS) " Reactor coolant isolation cooling (RCIC) Automaticdepressurizationsystem(ADS) Lowpressurecoolantinjection(LPCI) Residualheatremoval*(RHR) Low pressure core spray (LPCS) Electric power systems (EPS) Essential service water (ESW) Standbyliquidcontrolsystem(SLCS) Redundant reactivity control system (RRCS)
- This system has four fault trees associated with it.
One tree applies for shutdown cooling, one for containment sprays, one for suppression pool cool-ing, and one for steam condensing. The steam condensing function of the RHR was not considered in the GESSAR-PRA accident sequence evaluation.
m. Nemoto: I. A. Papazoglou May 5, 1983 Page eight I#,3 2-There are no system fault trees developed for the following systems: / - a) Reactorprotectionsystem-f I I b) Plant air systems (support system) c)_Feedwater[ndPCS d) Condensate GESSAR-II PRA states that the main steam and feedwater systems (beyond the nuclear boundary), the main condenser and its supporting systems, the con-densate systems, the plant air supply systems, the essential service water supply systems, and the offsite electrical supply systems are not part of the GESSAR nuclear island. Thus, the PRA assumed values for the unavailability of, these systems based upon judgement. As a result of the BNL review, additions and revisions were made to the trees based on the GESSAR. systems description and the Failure Mode and Effect Analysis (FMEA) reported in the SAR. These changes are discussed in the fol-lowing subsection. The level of resolution in the' trees is consistent with the state-of-the-art and PRA practice. The following items were not included in the analysis of the failure of a component (or system) because they are considered to be outside the scope of the PRA: a) External events b) Sabotage c) Operator errors of commission d) Location-dependent common mode failures. Qv\\ t s c .i l 4 O e
Meno to:
- 1. A. Papazoglou May 5, 1983 Page nine The intention in the GESSAR-PRA was to treat dependences within a system and between systems by using the same alphanumeric ' designator for components that appear several times in the trees.
For example, if a division of DC power supplies an initiating circuit of an ECCS pump, then the fault tree would designate this 125 V DC diYision by the same alphanumeric designator, in both the initiating circuit and the control system. This ensures that the fault tree, when properly manipulated, will account for'this dependence. The preliminary results of the BNL review indicate that this policy was not con-sistently applied when system fault trees were developed. 4.1 Summary of BNL Modifications to GESSAR-!! System Fault Trees A thorough review of each fault tree was performed using GESSAR drawings. The following is a list of major changes that were made to the trees and which significant19 changed some of the system unavailability values: High pressure Core Spray a) The failure of Division IV power (EVID and EV30) as a failure mode of channels D and H, as included. b) Failures of the HPCS pump suction supply in which automatic transfer from the CST to the suppression pool is precluded were included. c) Mechanic'al failures of HPCS valves.in which manual ~ repositioning of the failed valves is precluded were included. d) Some b.: sic component names were changed to account for commonalities with'other systems. e) Eliminated diesel cooling failure when offsite power is available. f) The designation of loss of pond water was changed to account for com-monalities with other systems. Reactor Core Isolation Cooling a) Failures of DC bus E as a failure mode of RCIC where added. b) The failure of V1 and V3 power as a failure mode of RCIC was added. k
Memo to:
- 1. A. Papazoglou May 5, 1983 Page ten c) The names of some basic components were changed to properly account for commonalities with other systems, d) Mechanical failures of RCIC* valves in which manual repositioning of the failed valves is precluded were included.
e) Failures of. the RCIC pump suction supply in which automatic transfer from the CST to the suppression pool is precluded were included. f) The pump room cooling portion of the tree was modified per SAR drawings. g) Decreased failure to run values of tne turbine and the pump to be consistent with GE data. Automatic Depressurization System (ADS) a) Failure of the operator to open non-ADS SRVs and failure of the operator to manually initiate ADS given the auto-initi,ation failure was considered as a common failure to depressurize. b) The basic components used for AC and DC power were substituted by respective sub-trees in order to properly account for the inter-dependences. c) Failures of the normal and backup air supply due to a rec 61ver rup-ture, clogged drywell penetration failure, stuck open receiver relief - valves, and failed check valves were included. d) Failure of the normal air supply owing to a clogged discharge filter or failure of the air dryer unit was included. e) The reactor low water level signal (Level 1) designators for failures of the transmitter, ACU, isolator and common mode miscalibration of the transmitter were changed to conform to those used in the low pres-sure ECCS system fault trees. This was done in order to properly account for commonalities between these systems. f) The requirement to have a high drywell pressure signal for ADS auto-initiation was deleted. ~ w--_
Memo to: I. A. Papazoglou May 5, 1983 Page eleven Low Pressure Coolant Injection a) The loss of pump auxiliaries was modified to properly. account for pump room cooling failure. b) Some component failure values were modified to be consistent with GE data. Low Press'ure Core Spray a) Failure of the system, if the minimum flow valve (normally closed) fatis to open upon system initiation while the pump is running but the injection valve is still closed, was included. b) ThI loss of pumping system auxiliaries was modiff 9d to properly account for pump room cooling failure. c) Failure of fuses in the auto-start signal was included. RHR Suppression Pool Cooling' Hode a) Mechanical failures of the system valves which would prevent the operator from taking corrective action were included. b) Component names were changed to properly account for commonalities between systems. c). Pipe rupture in crossover line was included. d) The loss of pump auxiliaries was modified to properly account for pump room cooling failure. e) Component failure values were modified to be consistent with the GE data. RHR Containment Spray a) Mechanical failures of system valves which would prevent the operator from taking corrective action were included. b) Component names were changed to properly account for commonalities between systems. c) Incorrect valve designators were corrected.
Memo to: I. A. Papazoglou May 5, 1983 Page twelve Suppression Pool Makeup a) Common mode miscalibration of the suppression pool level instrumenta-tion was included. Essential Service Water a) The common mode failure probability of all service water due to loss of pond water was changed from 10-6 to 10-7 since the PRA value of 10 6 was felt to be conservative. This change will be further examined as the review of the PRA proceeds. RHR Steam Condensing Function This tree was not evaluated because this function was not used in the GESSAR-PRA. Standby Liquid Control System (SLC) a) The failure of a pipe ca'p downstream of a valve was removed since there is no cap on this line. b) Component unavailabilities values were changed to be consistent with GE data. Electric Power ~ a) The failure of service water to the three emergency diesel generators was treated as a basic event in the PRA. These basic events were substituted by appropriate ESW sub-trees in order to properly account for commonalities. b) Failures of breakers that were omitted were included. c) The non-class IE battery chargers were included as a means of sup-plying rectified power to Division 1 through 4125V DC buses. The PRA did not take credit for the non-class IE chargers. d) Thelossof6900VACpreferredpowertobusG(Division 3)wasmodi-fled to account for the fact that there is no alternate offsite sup-ply to this Division 3 bus (according to GESSAR-!! Electric Power System drawing).
~ c. g. Kero to:
- 1. A. Papazoglou May 5, 1983 Page thirteen 4.2 Quantitativ6 Results and Comparison A copy of the GESSAR system fault-tree inputs was supplied to BNL to facilitate the review process. Quantification of the GE supplied system fault t,
trees pertinent to this review was made by BNL using both the WAM8AM and SETS computer codes and the results were in very close agreement with those re-ported in GESSAR. The BNL modifications to the system fault trees were implemented. It is noteworthy that all support systems, namely, ess'ential service water and elec-tric power systems, are explicitly included in the BNL modified system fault trees, i.e., the support system designators were substituted by respective sub-trees. This inclusion does not alter the unavailability of each system as presented in Table 4.1. However,' this is necessary in order to properly ac-count for the inter-system dependences when the accident sequences are quanti-fled. Results in Table 4.1 show that there is little difference between the GESSAR-!! and BNL values' for the high pressure injection systems while there is less than 40% increase in the system unavailabilities of the low pressure corecoolingsystemsandtheRHR(suppressioncoolingmode). The major in. crease (factor of 5) comes in the ADS unavailability. This can be attributed to the t'reatment of common mode of operator failure to open non-ADS valves and to manually initiate ADS. 5. Quantitative Analysis of Accident Sequences 5.1 GESSAR Event Trees, Event trees were used in the GESSAR-!! PRA to model the succession of failures of various syste-s which lead to core damage sequences. A total of five event trees were developed for the transient initiators with successful , scram, corresponding to the five groups defined in Section 3.2. These event trees are presented in Figures 5.1-5.12. e a 4 L
M'emo to:
- 1. A. Papazoglou May 5, 1983 Page fourteen Turbine Trip Transient For the turbine trip initiator, two event trees are given (Figures 5.1-5.2).
The first one addresses the question of successful scram. For the case where there is a successful reac' tor scram or ARI, after the onset of a transient (see Figure 5.1) the progression of the event is further investi-gated in either the second turbine trip event tree (Figure 5.2) or in the second isolation tree (Figure 5.4) if the feedwater system is not available. ' After the reactor has attained subcriticality, failure to accommodate the pressure surge caused by the transient due to failure of safety relief valves (SRVs) to open (M) is conservatively assumed to result in a large LOCA event. The success and the failure of the SRVs to reclose lead to two different se-quence paths. The top branch leads to a successful sequence due to the avail-ability of the feedwater and PCS system. The lower branch is evaluated for the high pressure system functions, viz., the HPCS or RCIC function, UHR. If the high pressure functions are successful, core damage may not occur pro-vided that the containment heat removal function is successful. If both high pressure functions fail, then the depressurization function, X, is examined. This function entails the automatic ADS depressurization function as well as the capability of the operator to depressurize the system manually on a timely basis. Failure of the containment heat removal function (W) or the low pres-sure injection or the timely ADS actuation function leads to core damage. Isolation Transient (Tp) (Figures 5.3, 5.4) - Two event trees are used to describe the isolation event as in the turbine trip transient. These types of transients lead to a more significant challenge to the plant as compared to the turbine trip transients. The isolation event tree is similar in structure i to that of the turbine trip. This is because of the similarities in both the challenges posed by the two initiators and the required response of the safety , functions of the plant to mitigate the events. The difference between the two event trees resides in the recovery of the feedwater/ power conversion system e
M'emo to:
- 1. A. Papazoglou May 5, 1983 Page fifteen for the high pressure injection function (Up).
This recovery is less likely for the isolation transients than for turbine trip transients. Inadvertent Open Relief Valve (T ) (Figures 5.5, 5.6, 5.7) -This I transient was treated separately since the operator must recognize the event and manually scram the reactor. Additionally, the containment conditions are different from other transients because of the higher total heat addition to the suppression pool at the time of plant shutdown. This higher heat addition places a' more significant demand on the containment heat removal function. ~ Of the three event trees used to describe the 10RV transient, the first tree, Figure 5.5, presents the occurrence of single or multiple 10RV and also addresses the likelihood of occurrence of an 10RV ATWS. The remaining two trees, Figures 5.6 and 5.7, describe the progression of the accident resulting from a single or a multiple 10RV transient initiation with successful scram, respectively. These two trees are similar in structure, in that they evaluate availability of high and low pressure injection, and containment heat removal. The only difference between them comes from the fact that for multiple 10RV, depressurization of the reactor is not required. Reactor Shutdown (T ) (Figure 5.8) - This event tree accounts for those M challenges to the plant which result from a planned manual shutdown. Unplan-ned reactor shutdown events are transferred to the ' turbine trip event tree. The shutdown considered is not a scram but a manual control rod insertion in a slow, orderly manner. Examples of these types of shutdowns are scheduled or forced maintenance outages and refueling outages. Operating experience indicates that because of the controlled nature of the transient, the SRVs are not challenged. Therefore, only the high pressure injection function and ADS actdation or low pressure injection function fail- ' ures would lead to core damage. Failure of the containment heat removal function also results in core damage. e M e
Memo to: I. A.* Papazoglou May 5, 1983 Page sixteen Loss of Offsite Power (T ) (Figures 5.9-5.12) - This transient provides E unique initial conditions for accident sequences because of the loss of AC power and the resulting demand for the diesel generators. The initial condi-tion of loss of AC power affects the majority of the frontline systems since AC power is needed for most plant systems. This tree has been time phased for the coolant injection and containment heat removal functions to account for recovery of AC power. The first LOOP event tree, Figure 5.9, addresses the likelihood of oc-currence of an ATWS. In the case where there is a successful scram, failure to accommodate the pressure surge (M) after the plant transient results in a transfer to the large,LOCA event tree. If the M function is successful, recovery of offsite power in two and ten hour time periods is addressed. Based on the time of recovery of offsite power, three event trees are developed in Ficures 5.10, 5.11, and 5.12 which describe the progression of the events, characterized b'y the loss of offsite power for less than 2 hours, for more than 2 hours but.less than 10 hours, and for more than 10 hours, respectively. These event trees are identical in structure and show the success or failure of the SRVs to reclose, high pressure injection, ADS or manual depressurization, low. pressure injection and containment heat removal ~ - functions. 5.2 Bhl Review of Event Trees Turbine Trio Transients ~ BNL revised the GESSAR-II turbine trip accident sequence delineation to allow for a more realistic modeling of the failure of the feedwater system. This is especially evident for the case of no SORV in which more credit is given to the recovery of the feedwater system. The unavailability of the containment heat removal function given a failure of low pressure core injection and low pressure core spray (LPCI and LPCS) was modified to reflect the fact that the LPCI has already failed. This modification is necessary because the RHR and LPCI systems share pumps and several valves. k
I I Memo to:
- 1. A. Papazoglou May 5,1983 -
Page seventeen Isolation Transients Revisions similar to those presented in the turbine trip transient were made to the GESSAR-II isolation event tree. 10RV Transients The failure probability of the scram initiation function (C ), was mod-1 ified using more realistic values for human error to manually initiate reactor ^ scram.. Reactor Shu'tdown No changes were made to the GESSAR-II reactor shutdown event tree. Loss of Offsite Power In order to properly model the details of a loss of offsite power event, BNL developed a time phased fault tree to consider the interactions'of systems and their responses for various discrete time intervals ranging' up to 24 hours following the onset of the transient. Thus, the possibility of recovering certain systems during the~ various time phases can be taken into account. Especially, in time phasing the fault tree, the following considerations are taken into account:
- 1) Cooling of the RCIC room is required after'two hours 'or system failure is assumed.
2) In the event that the high pressure core spray is unavailable at 1/2 hour due to failure of its own diesel (Diesel 3) and that RCIC fails at 2 hours due to lack of room cooling, credit is given for the re-covery of Diesel 3 at 2 hours.
- 3) Detailed modeling of the availability of the various divisions of power is possible for the ADS, low pressure injection and containment
~ heat removal functions. O
. ~. 4 Memo to: I. A. Papazoglou May 5, 1983 Page eighteen 5.3 Quantification of Accident Sequences 5.3.1 BNL Methodology l-To properly account for the support system dependences, the quantifica-tion method must preserve these dependences, first within the functions and next at the accident sequence level. This is accomplished by transforming the event tree into equivalent fault trees called core-damage fault trees (CDFT). The top event of the CDFT is core damage and the " basic events" of the,CDFT are.the functions of the event tree, e.g., U,V. These " basic events" are then substituted by the respective functional fault trees, e.g., U is substituted by.HPCS, RCIC where each of the systems within the functional fault tree is replaced by its system fault tree. As a result of the method used in quanti-fying the system fault trees, namely, the proper inclusion of all support systems, quantification of the CDFTs will properly preserve the dependences within functions and also at accident sequence level. An example of the CDFT is provided for the turbine trip transient in Figure 5.13; its equivalent event tree is given in Figure 5.14. 5.3.2 Results Core damage fault trees were developed for the four transients and re-actor shutdown. A summary of the results obtained by using the BNL method-ology, i.e., the CDFT, is presented in Tables 5.1 and 5.2. GESSAR-II PRA results are also presented in these tables for comparison purposes. Table 5.1 shows the conditional probability of core damage given the respective initi-ators. The probabilities of core damage given in the first column of Table 5.2 were obtained by using the results presented in Table 5.1 multiplied by the respective BNL initiators; the GESSAR-PRA values are also given for com-parison. Values presented in Tables 5.1 and 5.2 for Class II type accidents should not be interpreted as core-damage frequencies since they only represent the i loss of containment heat removal with successful coolant makeup. 1 L e v- <e, yw,,--,-ev--n.,--e w ,v-, w ---v-, ,-,,,-y---,--- -:,--e.--,-s,-+-~. --.-,----*--m m --es t
Nemo to: I. A..Papazoglou May 5, 1983 Page nineteen . Turbine Trip Given that there is a turbine trip, BNL calculated the core-damage-frequency of 5.1x10-7 for Class I type accidents as compared to. s for GESSAR-II PRA (see Table 5.1). The major contribution to the increase can be attributed to transient induced loss of offsite power which is accounted for in the BNL method. .-Q Without including the dependences within functions, a conditional j ~j i probability of core damage of 3.4x10-9 was obtained using the BNL revised 6 event tree and system unavailabilities. The difference between this result ~ and the GESSAR-II PRA value is mainly due to the higher unavailability for the ADS; this agreement in the core damage frequencies appears to indicate 'that dependences among systems were neglected in the,GESSAR-II PRA. For Class II sequences, the BNL analysis shows an increase from ,f 'GESSAR-PRA) to 3.4x10-7.(BNL). The difference in values can be ascribed to the increase in RHR and PCS unavailabiliti s, and proper evaluation of dependences. The increase in PCS unavailability is due to the e, treatment of dependences between feedwater and PCS. [ C' p r ~ Isolation Given that there is an isolation transient, BNL calculated the core-daqage frequency of 5.6x10-7 for Class I type accidents as compared to ,q ( $for GESSAR-II PRA, (see Table 5.1). The major contributors to the Ih l ~ increase can be ascribed to the transient induced loss of 'offsite power which is properly accounted for in the BNL method and to the increase in the ADS and feedwater unavailabilities. %'I For Class 11 sequences, there is an increase from to Jf ,6.4x10-6 The major contributions come from the increase in RHR and PCS J. unavailabilities.
-i 1 Memo to: I. A. Papazoglou May 5,1983 Page twenty 10RV Giventhatthereisan10RV,BNLcalculatedthecore-qmagefrequencyof y f 9.6x10-7 for Class I type accidents as compared to/ for GESSAR-II I/ 2 ' PRA. The major contributors to the increase can be' attributed to transient induced loss of offsite power which is properly accounted for in the BNL method, and to the increase in the ADS unavailability. For the Class II type sequences, there is an increase from . s to if )< 4.0x10-6 The difference is primarily due to the increases in the RHR un-1 availability and in the probability of failure to manually scram the reactor. Reactor Shutdown Given that there is a reactor shutdown initiator, BNL calculated the core damage frequency of 1.0x10-7 for Class I type sequences as compared to h g (:comes fro / y GESSAR-II PRA. The major contribution for this difference i h s f mthesupportsystemdehendenceonthepond. 3 For Class II sequences, there is an increase from! ho 1.3x10-7 This is attributable to the increase in RHR unavailability and t. ./ the proper treatment of dependences between feedwater and PCS. LOOP An increase is noted in the core damage frequency for Class I type se-quences given the loss of offsite power initiator. The BNL value was cal-culated to be 1.6x10-4 whereas the GESSAR-II PRA analysis reported a value k of3-Examination of the minimum cut set results indicates that the \\ }\\ diesel commo]n mode failures contribute significantly h frequency of Class I type sequences. It appears that in the GESSAR-II PRA, . these common mode failures were not adequately included in the quantification. The total core damage frequency for Class I sequerices as calculated by BNL is 4.0x10-5 This represents an approximate increase by a factor of a
= .- \\ Memo to: I. A.- Papazoglou May 5, 1983 Page twenty-one C / s nine over the GESSAR-II PRA value of (see Table 5.2). In the h /: u e GESSAR-II PRA, about 99% of the Class I type sequences come from the loss of y offsite power transient as compared to about 88% in the BNL review. The turbine trip and isolation sequence each constitutes about 6% of the total Class I core damage in the BNL review. O O e e G .O % o y G 9 e n* N a 99 9 g5 ~
a Table 2.1 Safety Functions Required for Initiating Events
- 1) Render reactor subcritical
- 2) -Protect reactor coolant system from overpressure failure
- 3) Remove decay and sensible heat from core
- 4) Protect containment from overpressure
- 5) Scrub radioactivity from containment atmosphere Table 2.2 Safety Functions for GESSAR-II
- 1) Render reactor subcritical.
- 2) Protect reactor coolant system from overpressure failure
- 3) High pressure injection of coolant into core
- 4) Depressurization
- 5) Low pressure injection of coolant into core
- 6) Containment heat removal and/or containment spray *
- 7) Scrub radioactivity from containment atmosphere
- Containment spray is required to prevent containment overpressurization for LOCAs which are located in the containment but outside the drywell.
9 C e -,r-
Table 2.3 Frontlin'e Systems for GESSAR-II Safety Function Frontline Systems
- 1) Reactor subcriticality
- 1) Reactor protection system
- 2) Redundant reactivity control system including a) Alternate rod insertion b) Automatic standby liquid control c) Recirculation pump trip d) Feedwater runback
- 2) ' Reactor coolant system over-
- 3) 19 Safety relief valves (SRV) pressure protection
- 3) High pressure injection
- 4) HPCS
- 5) RCIC
- 6) CRD*
- 7) Condensate and feedwater system with power conversion system
- 4) Depressurization
- 8) ' Automatic depressurization-system (8 SRV's used for this function)
- 9) Manual depressurization
- 5) Low pressure injection
- 10) LPCI
- 11) LPCS
- 12) Condensate pumps
- 6) Containment heat removal
- 13) RHR and ESW
- 14) PCS
- 15) Suppression pool
- 16) Containment sprays **
- 7) _ Scrub ~ radioactivity from
- 17) Suppression pool containment atmosphere
- 18) Containment sprays
- These systems were conservatively not considered in the PRA front end an-alysis.
- For LOCAs in the containment (outside the drywell), spray ~1s required to prevent overpressurization failure.
Table 2.4-GESSAR-II Support Systems i - o Electric Power System .Three' Diesel Generators - Three-Load Divisions Four 125V DC Class IE Buses o Essential Service Water System Three Loops o Plant Air. System .e 4 e e O ~ t e n
Table 3.2 GESSAR-II Initiating Event Groupings Initiating Event Groupings Reactor Shutdown o Planned Shutdown o Other Scrams - Inadvertent Scram - Flux or Pressure Scram - Single MSIV Closure Turbine Trip - Turbine Trip - Generator Load Rejection Isolation o Feedwater Failure - Recirculation Failure - Loss of All Feedwater - Feedwater Control Failure o Immediate Isolation - Pressure Regulator Failure - MSIV Closure - Loss of Condenser Vacuum Loss of Offsite Power Inadvertent Opening of Safety / Relief Valves O =
Table 3.3 Summary of the Categories of BWR Transients Used to Classify Operating Experience Data on Anticipated Transients
- 1.
Electric Load Rejection 2. Electric Load Rejection with Turbine Bypass Valve Failure 3. Turbine Trip 4. Turbine Trip with Turbine Bypass Valve Failure 5. Main Steam Isolation Valve Closure 6. Inadvertent Closure of One MSIV (Rest Open) 7. Partial MSIV Closure 8. Loss of Normal Condenser Vacuum 9. Pressure Regulator Fails Open 10. Pressure Regulator Fail,s Closed 11. Inadvertent Opening of a Safety / Relief Valve (Stuck)
- 12. Turbine Bypass Fails Open 13.
Turbine Bypass or Control Valves Cause Increase Pressure (Closed) 14. Recirculation Control Failure -- Increasing Flow
- 15. Recirculation Control Failure -- Decreasing Flow
- 16. Trip of One Recirculation Pump
- 17. Trip of All Recirculation Pumps
- 18. Abnormal Startup of Idle Recirculation Pump
- 19. Recirculation Pump Seizure 20.
Feedwater -- Increasing Flow at Power 21. Loss of Feedwater Heater 22. Loss of All Feedwater Flo'w 23. Trip of One Feedwater Pump (or Condensate Pump) 24. Feedwater -- Low Flow 25. Low Feedwater Flow During Startup or Shutdown 26. Hig.h Feedwater Flow During Startup or Shutdown
- 27. Rod Withdraw at Power 28.
High Flux Due to Rod Withdrawal at Startup 29. Inadvertent Insertion of Rod or Rods 30. Detected Fault in Reactor Protection System
- 31. Loss of Offsite Power 32.
Loss of Auxiliary Power (Loss of Auxiliary Transformer) 33. Inadvertent Startup of HPCI/HPCS 34 Scram due to Plant Occurrences
- 35. Spurious Trip via Instrumentation, RPS Fault
- 36. Manual Scram -- No Out-of-Tolerance Condition
- 37. Cause Unknown
- EPRI-SAI Study S
c Table 3.4 BNL Grouping of GESSAR-II Transient Initiators Item Transient Group ** 1 ~ Isolation Electric Load Rejection Turbine Trip (3*,4,13*) (1*,2) Closure of All MSIVs (5) Loss of Condenser (8) Pressure Regulator Failures (9*,10*) Recirculation Problems (15*,17*,19*) Disturbance of Feedwater (20*,22) Loss of Auxiliary Power (32*) ~ 2 Turbine Trip Partial Closure of MSIVs (6,7) Bypass Fails Open (12) Recirculation Problems (14,16,18) Disturbance of Feedwater (21,23,24,25,26) Rod Withdrawal / Insertion (27,28,29) Fault in RPS (30) Inadvertent Startup of HPCS (33) Others (34,35,36,37) 3 Loss of Offsite Power (31) 4 Inadvertent Open Relief Valve (11)
- Number in parenthesis refers to transient numbers from Table 3.3.
- These transients are normally grouped under turbine trip events, however, for the GESSAR-II, they are better characterized as isolation events (see Section 3.0).
O
Table 3.6 BNL Initiator Frequencies for GESSAR-II r y Oh' 7 GESSAR-II PRA BNL Planned Reactor Shutdown (Tg) 3.20 Turbine Trip (T ) 5.08, T . Isolation (Tp)' 4.32 LOOP (T ) 0.21 E 10RV (T ) 0.25 I TT ATWS 0 Tp ATWS 9.40 TE ATWS 0.21 Tg_ATWS l 0.25 i [ V,1,
- /
77' 4 I e e s ea, e ag e 'O 9 0
o Table 4.1 Comparison of System Fault Tree Results - GE Failure Probabilities h*9tA%fV I f Preliminary GESSAR BNL Results HPCS 3.9E ' RCIC 9.1E-2 ADS l 5.0E-5 LPCCS 3.0E-5 k RHR (Suppressiol pool cooling) 1 5.3E-4 \\ 4 m_ e 9 ' 6 e b O 4 g-y ,--r~ r. t- -<---ww >~-- -e- - - - - - - - - - -
- a---
1 Table 5.1 . Conditional Probabilities of Core Damage (Given an Accident Initiator) Class I Class II BNL GE, BNL GE p Turbine Trip 5.1E-7 3.4E-7 Iso'lation 5.6E-7 6.aE-6 10RV 9.6E-7 4.0E-6 Reactor Shutdown 1.0E-7 i 1.3E-7 Loss of Offsite 1.6E-4 NA Power J t\\ NA = Not available as of this date. ll M e * ' 4 4 4 9 9 = e g 9 e 8 9
r ~ Table 5.2 Core Damage Frequencies for Transient Initiators Class I Class II r f BNL* N.a E** BNL* GE** G.c.. - Turbine Trip 2.6E-6 1.7E-6 r, <t I' solation 2.4E-6 2.8E-6 t h - 10RV' 2.4E-7 1.0E-6 - Reactor Shutdown 3.2E-7 4.3E-7 ' Loss of Offsite Power - 3.4E-5 NA
- Values were obtained using the BNL methodology and the BNL initiators.
- Values from GESSAR-II PRA with GE initiators.
NA = Not available as of this date. e* ee O g a0 6 4* DIP em e
E 1 I - l Nomenclature for Figures 5.13 and 5.14 T. Turbine Trip T = C Scran = ' M Limit Pressure = P SRVs Reclose = Q Feedwater System = U High-Pressure Injection (HPCS or RCIC) = X ' = ADS or Manual Depressurization 4 V4 Low-Pressure Injection (LPCI or LPCS) = Vc Condensate Pump Injection = W Containment Heat Removal (one RHR or PCS) = Q/P Feedwater System Given P Failure = , q/F Feedwater System Given P Success = U/P = _High Pressure Injection Given P Failure U/F High Pressure' Injection Given P Success = W/P. Containment Heat Removal Given P Failure = W/F Containment Heat Removal Given P Success = W/F, VT = Containment Heat Removal Given P and V4 Success W/F,V4 Containment Heat Removal Given P Success and V4 Failure = G 4 O
TTCD' A [h C' M. TT T 1 I j C2 C1 . l i T Figure 5.13a - Turbine Trip Core-Damage Fault Tree 1 e 9 e N
~ s A 4V 9 AV G c AX 9 \\ X 1 ss I l a l C P / Q eer T 8 t G lu p a F e g F a / m U a D ~ -ero D C 'CT p T r i T en i br u T 5 G b 3 1 [V 4 5 eru A X g i F 4 4G P
TTCD T O G18 O _.T_ G12 p T 219 0/F 4 4 t f A G13 G136 320 G24 T- [ W g Q/P . /P. U/F X "~ gjp b g j U/F W/F GIS GIS4 g rg- - r-b( JO LVP W/P X V4 U/P W/P W/P,W V4 Vc V4 W/F,V4 Figure 5.13c - Turbine Trir '; ore-Damage Fault Tree I i
g. -r a T C M P Q U X V4 Vc W T I 1 1 3 ~ !i ) I 1 i i i Figure 5.14 - Turbine Trip Event Tr'ee l: t L
~ r y. ^ '. h .f go>g gN' (o (. BRIEFING ON THE BNL LETTER REPORT REVIEW 0F GESSAR-il PRA PRESENTED TO GE STAFF AT BROOKHAVEN NATIONAL LABORATORY. JULY 20-21, 1983 9 BROOKHAVEN NATIONAL lABORATOPYl} g)l 3 A5500ATED UNIVERSITIES, INC.(IllI $Q ^
>y.,:
- i..
- p O
. OUTLINE ~ o CONTAINMENT EVENT TREES o' CORE MELTDOWN PHENOMENA o-FISSION PRODUCT RELEASE AND TRANSPORT ~ i BROOKHAVEN NATIONAL LABORATORY l} g)l A5500ATED UNIVERSITIES, INC.(I Lll i. o
ya ,0. ,,g i ~ k QUESTION 720 104
- o. EFFECT 0F HYDROGEN PHEkOMONA ON THE PROBABILITY 0F.THE VActlUM BREAKERS TO STICK OPEN o
NUMBER, LOCATION, AND POTENTIAL' FLOW AREA 0F THOSE ~ BREAKERS SHOULD THEY REMAIN OPEN o POTENTIAL FOR SIGNIFICANT LEAKAGE THROUGH THE VACUUM BREAKERS UNDER SEVERE ACCIDENT CONDITIONS k l' f i. I-BROOKHAVEN NATIONAL LABORATORY l} g)l r A5500ATED UNIVERSITIES, INC.(Illl .r r .. -, _. -.. -, ~ _..... c.~,._ ,_.m ,_,..---~.-r,--__--
+; g O E QUESTION 720 111 o TOTAL PROBABITLITY OF EARLY LOSS OF CONTAINMENT' INTEGR1TY (Y') .o PROBABILITY OF CONTINUOUS BURN IS INCLUDED IN THE CALCULATION OF.THE PROBABILITY OF DRYWELL LEAKAGE THROUGH PIPING PENETRATIONS AND OF THE PROBABILITY 0F DISABLED INSTRUMENT LINES ( $ AND 5') o PROBABILITY OF DISABLED INSTRUMENTS LINES REQUIRES-THE BURN OF THREE OR MORE CLUSTERS 0F ASSEMBLIES (s') o PROBABILITY OF FAILURE OF THE REDUNDANT LOCA ISOLATION SIGNALS OR THE INBOARD AND OUTBOARD HVAC ' ISOLATION VALVES (c) BROOKHAVEN NATIONAL LABORATORY l)l)l A5500ATED UmVER5mES, INC(I LII
m ,~ y n} f.t{ k, y_, ;; : [ '., / C ^ ~ CUESTION 720.111 (IN RESPONSE-TO QUESTION 720 40) 9 o JUSTlFY THE VALUES ASSUMED FOR THE CONTINUOUS BURN MULTl? LIER o EXPLAIN INCONSISTENCY _.BETWEEN THE EVENT TREES FOR TASES 3, 5 AND 6 AND THE CONDITIONAL PROBABILITIES IN TABLE 1 4-1 ~P - J w . 's 's ~ 4 o. BROOKHAVEN Nail 0NAL lASORATORYl} g)l y 'y A5500ATED UNIVER$1 TIES, INC.(lll1 I , Ig t
~ s -i QUANTIFICATION OF CONTA-INMENT EVENT TREES I INDEPENDENT ASSESSMENT OF BRANCH POINT PROBABILITIES -8 AN EXAMPLE: POTENTIAL FOR FOUR " TYPES" 0F H2 PHENOMENA 0 ' MODIFY MARCH TO MODEL LOCAL H2 CONCENTRATIONS ? I TIME DEPENDENT CONCENTRATIONS 0 CALCULATE REPRESENTATIVE PROBAEILITIES OVER SELECTED TIME FRAME [ BROOKHAVEN NATIONAL LABORATORY l} g)l ASSOCIATED UNIVERSITIES, INC.(llll
.l .- >~
- z= ~i2
.:::: E
- -.c_ <
.~ c"-5 F.: c a c.-. Co e_m u.- cEE== 0 lum E m e cd < w -JW = = =,, - C:: c W cm - c= w' cc
- 4. - -.
l b L l b ] p L 4 [ $ LS ~ v 1 Sill [ 1 +t 1-c. I 1. i .g s n. Gn e.{ l YotOmE 1 N O LUrol f .' s s 1 -+- /0 Lurn E \\/otvewF 2 s q ~ _ _ _. i s i. 4 b f' [,, 34, \\ 6 PFG - 6/1/82 ll v' ~ GLOBAL H DETONATION EFi ECT CN STRUCTUR$5 2
t r 'h y, - 4.s.-.-,,,, J,* )$ A * '"* d' ?.. p yb-d -T J*%" 5 ..._. ~ 4 e (i ~' 1 g \\k l \\q s T \\ \\ ~k \\ 4 \\ \\ I g d M I m. t e q N . m 6 lli n e . y. \\ 8tu s \\ k-m - 8 'l i ~ o W . -Q A b k = 1 ll l I Q l ~ e """"-~~l l s F e s e Y s e s A .a M u k e Y N e-l
G s - e t g e
- EP e c
nN ~$ -8m 1 c -g -a } x y I i R -{ 4 b' tQ "g - 4 mg ~E% x og h. k N k E h k]4 g _a e + i, e b i (- e k UW g-Eg ~q % n n 3 t' yo k k E' kb i _g t-e 4 I l l 8 4 \\ D Q D N g N Q Q D Y So m ee e k D k M
r . m n !.Q &. 7 W 55h.*]., ?!- m. a. 'c. s ,. _2, 0 'p r .(. \\ s... 1 r n .ex 4 Q TV) DM [. - -k I f m y {g ~ 1e 4e b-t~ ( f - s + ek 8 w s _4 9m _ul 8 g i x C i i. h b 1 ~ x wD p +g 1 _L 9 \\-e 1 i 1 u_,l l N D N o N o s Y o N N g a a M a Q a-Q 1 'n
e 3" le O 4 g b n g g. g - g s h _E~ R=- ~ w 4 Po x -e. 6 _8 bg w l c %5 4i e 1 y k ( 1 8 k k I s o s o g C c+ M n N N u so 1:1 x E A,
. # M.._ _,. w
- cp y'
- ~
rfiE 1;;d 2-M ".L 2 9 Table'3.2 three mixture criteriaTime during wnich tne cc e Perico in which criterion is Satisfied ( Cell min) __ Number Combustible _ Ma rginally Detonable Fully 1-Detonable 195 20 2 - 0 125 110 3-25 360 355
- 4 225 900
~ 830 0 Total 1150 1145 230 Table 3.3 satisfy the three mixture criteriaPercentag artments Cell- __ Humber Combustible Marginally Detenable Fully l1 Detonable 0 0 2 0 12 11 3 7 83.9 82.8 4 80 0 0 l o f_ . Total 94.4 93.6 1 87 "Tne ti me F.elting. perioc is for one hour measured from the start of m
' f ;- f COMPARISON OF CONDITIONAL PROBABILITIES FOR H2. PHENOMENA GE d, BNL LGLOBAL DETONATION-f \\ IM 0 I ~ GLOBAL:-COMBUSTION l 0-l 0.93 LOCAL' DETONATION -LOCAL COMBUSTION i '/ O.07 \\ \\ / v BROOKHAVEN NAiiONAL LABORAl'JRY[][]l ASSOCIAlED UNIVERS111ES, INC.O ll1
,e u. -T': CORE MELTDOWN YHEN0MENA ~ o. HYDROGEN GENERATION-o INPUT VALUES GOVERNING HYDROGEN PRODUCTION (NLB0ll) o-CORE DEBRIS FRAGMENT SIZE o DISTRIBUTION OF MATERIAL WITHIN DEBRIS FRAGMENTS o POOL-DF o DECAY. HEAT MODELS 4-e h BROOKHAVEN NATIONAL LABORATORY l)l)l A5500ATED UNIVERSITIES, INC.(llll E
FISSIONPRODliCTRELEASE AND TRANSPORT o BASE CASE VS. ADVANCED TECHNOLOGY CASE -o CORRAL COMPARISIONS I o-IMPACT 0F RSS METHODOLOGY BASE CASE o EFFECT OF GE CORE RELEASE FRACTIONS ~ USING RSS METHODOLOGY o EFFECT OF SELECTIVE HOLD-UP AND RE-EMISSION 4 BROOKHAVEN NATIONAL LABORATORY [] g)l AS$00ATED UNIVERSITIES, INC.(Il.11
m-g: r a ' BASE CASE VS. ADVANCED TECHNOLOGY' o :NRC REQUEST FOR PARELLEL APPROACH: ~ .o. BASE CASE: o. USES RSS METHODS REGARDING SOURCE TERMS o--ADVANCED-TECHNOLOGY. o 'USES NEW METHODS GENERATED BY SOURCE TERM OFFICE BROOKHAVEN NATIONAL LABORATORY l} r } l A5500ATED UNIVERSITIES, INC.(l(ll -w
i,. s-1. ~ n 4 FRACTION OF CORE INVENTORY RELEASED (DEPARTURE FROM RSS) o TE' RELEASED IN MELT PHASE RATHER THAN-VAPORIZATION PHASE o MORE BA-SR IS RELEASED (~4 TIMES) o.; LESS Ru 'IS RELEASED (~ HALF) o ORGANIC 10 DINE RELEASE REDUCED (~22 TIMES LESS) t BROOKHAVEN NAll0NAL LABORATORY l} [][ A5500ATED UNIVERSITIES, INC.(lil1 o
FRACT10N0F CORE lNVENTORY. RELEASED NUCLIDE GRour Gap MELT ~ VaE08.12/1T_IDN WAsit-liiOO GE \\ WASil-1fl00 GE WASil-1fl00 GE \\ XE-KR 3.0(-2) 8.7(-1) s 1.0(-1) 01 1.2( li) 6.18(-3) I 7.0( fl) \\ i 1-BR 1.688(-2) ' 8.768(-1) i-9.93(-2) Cs-RB 5.0(-2)_ 7.6(-1)
- 1. 9 ('-1)
I TE 1.0( li) 1.5(-1) 8.fl99(-1) BA-SR .l.0(-6) l 1.0(-1) 1.0(-2) Ru 0.0 ) 3.0(-2) i 5.0(-2) I i -LA 0.0 \\ i 3.0(-3) 1.0(-2) NJ XFV ..\\ \\ a t'S BR00KilAVEN NATIONAL LABORATORY l} g)l ASSOCIATED UNIVERSITl[5, INC. (I L l l
COMPARIS0N OF RE liASE FRACTI0iS AS PREDICTED FOR T E;RSIONS-0F T E CORRA_ CODi ' BY GE AND BNL I l-T-L3 CATEGORY. \\ I FISSION PR0nucT Gn0urs CASES Ka,XE. OI,1,Bn CS, Ris TE BA,Sa Ru LA 2 BASED ON GE VERSION i 0F CURRAL f BASED ON BNL VERSION 'l.0 3.0E fi 1.8E-6 1.8E-6 1.8E-6 9.0E-6 1.5E-6 0F CORRAL = - ( /d W ././ BR00KilAVEN NAll0NAL IABORATORYl} g)] ~ ~ ASSOCIATED UNIVERSITl[S, INC.(IIll ~..... _. _
r COMPARISON OF I-T-E2 AND BNL BYPASS SEQUENCE o. NOBLE GAS UNCHANGED o ORGANIC 10 DINE' DIFFER BY 22 '(RATIO 0F CORE INVENTORIES RELEASED) o MELT RELEASE FRACTION SUBJECT OT A DF=100 IN BNL SEQUENCE o VAPORIZATION RELEASE FRACTION SUBJECT TO A DF=1 IN BNL SEQUENCE CONLY ATTENUATION DUE TO SETTLING'AND PLATE OUT) BROOKHAVEN NATIONAL LABORATORY l} g)l A5500ATED UNIVERSITIES, INC.(llll l-
n COMPARISON BETWEEN GE SEQUEllCE I-T-E2 RELEASE; AND BNL. ~ RELEASE BASED ON RSS METil0D0_0GY g Fission PaooucT GROUPS CASES XE-Kn 01 1 Cs-Ra Te BA-Sn Ru LA' GE l-T-E2 s / BNL BASED ON .9916 6.9Il(-3)
- 5. 2'l (-3 )
2.05(-2) 6.39(-2) 1.56(-3) 3.93(-3) 7.62( II) rf Y RSS METil0DS W' BR00KilAVEN NAll0NAL LABORATORY l}'g} l A5500ATED UNIVERSITIES, INC.(IIll
P# ~ COMPARISON OF CORE RELEASE FRACTIONS o. NOBLE GAS'- ESSENTIALLY SAME o ORGAN 1C 10 DINE IN RATIO 0F RELEASE FRACTIONS o 10 DINE SIMILAR - PLATE OUT OF VAP. RELEASE BALANCED BY MELT RELEASE-DF o -CESIUM AND TELLURIUM - HIGHER VAP. RELEASE FRACTION ACCOUNTS FOR DIFFERENCE o BARIUM - HIGHER f:ELEASE FRACTION FOR GE ACCOUNTS FOR DIFFERENCE o RUTHENIUM'- HIGHER VAP. RELEASE FRACTION ACCOUNTS FOR-DIFFERENCE o LANTHANUM -lN RATIO 0F RELEASE FRACT10NS 4 BR00r; HAVEN NAll0NAL LABORATORY l} g)l A5500ATED UNIVERSITIES, INC.(llll
s EFFECT OF CORE RELEASE FRACTIONS' BASED ON THE RSS COMPARED. 111T11 GE CALCULATED FRACTIONS \\ FISSION PRODUCT GROUPS' CASES XE-KR 0I I C_S.-Bn TE Ila-S3 Ru ~La lNL BASED ON . (SS RELEASE .9916 6.9fl(-3) 5.2'l(-3) 2.05(-2).6.39(-2) 1,56(-3) 3.93(-3) 7.62( fi) MODEL 1 AS AB0vE BUT WITil GE ASSUMPTION REGARDING FP RE- .( ' q LEASE FROM FUEL cy Vey ~.. BR00KilAV[N NATIONAL LABORATORY l} g)l A550CIAT[D UNIVERSITIES, INC. (Ill! ~ y ~
i h.L ^ HOLD-UP AND REIEMISSION o NOB.LE GAS, ORGANIC IODINE AND-IODINE N0 HOLD-UP o CESIUM 40%. HOLD-UP o TELLURIUM 65% HOLD-UP o. BARIUM, RUTHENIUM AND LANTHANUM 60% HOLD-UP o 'RE-EMISSION OF CESIUM AND TELLURIUM ONLY F' BROOKHAVEN NATIONAL LABORATORY l}l)l A5500ATED UNIVERSITIES, INC.(Illl i. b'_
- x q COMPARIS0N OF GE DASED RELEASE FRACTIONS filTil'AND llITil00T SELECTIVE Il0LD-UP AND RE-EMISSION t . FISSION PRbouCT GR0ues. CASES XE-KR 01 1 -CS-Ra 'Te BA-SR RU LA BASED ON GE ~ CORE RELEASE-FRACTIONS ( BUT WITil NO PRIMARY SYSTEM GE RELEASE FRACTIONS WITil l 'I!OLD-(IP AND .}. TilEN _LO% RE-[ LEASE *
- f-1 GE RELEASE
/ FRACTIONS WITil !!OLD-UP AND 'f TilEN 10,0% RE j ~ .,__..._..3%/ D LEASE * ... f,39, ~~~ Ter
- 10 AND 100% RELEASE REFERS TO Tile ' EVENTUAL -RE-EMISSION OF Tile CS-RB AND IE FISSION PRODUCT ' GROUPS BR00KilAV[N NATL 0NAl LABORATORY [j g)l INITI ALLY ASSUMED HELD-UP IN. Tile. PRIMARY SYSTEM.
A550CIAl[D UNIVER5lIl[5,1NC.(Illl
[pm -v= n:;. . pg;.. t,;-n -4 ~;
- ^ ~,7 +:.
-;Q;; ' ~ 'o UNITED STATES P, f k NUCLEAR REGULATORY COMMISSION l' 3 9 C WASHINGTON. D. C. 20555 %, Q f o.*** SEP 61983 MEMORANDUM FOR: Brian W. Sheron, Chief Reactor Systems Branch Division of Systems Integration FROM: George E. Lear, Chief Structural and Geotechnical Engineering Branch Division of Engineering
SUBJECT:
REVIEW 0F THE GESSAR - II INTERNAL EVENT PRA -i In response to your request for assistance in performing containment structural integrity review of the GESSAR - II internal PRA per a note from B. Hardin to D. Jeng dated August 2,1983, we have completed' the review of Appendix G to the GESSAR Standard Safety Analysis Report (SSAR). Our review questions are listed in the enclosure. The review was per-formed by S.. Chan and K. Leu of Structural Engineering Section A of the Structural and Geotechnical Engineering Branch (SGEB). / ~ sL George [j. Lear, Chief Structural and Geotechnical Engineering Branch Division of Engineering
Enclosure:
As stated cc: W. Butler C. Thomas J. Lane D. Scaletti D. Jeng J. Meyer D. Yue W Hardin S. Chan K. Leu CONTACT: K. C. Leu, SGEB x28044 BW k.
{.lI l .~& L' + q. ENCLOSURE GESSAR - 11 PRA REVIEW QUESTIONS ON' APPENDIX G - SSAR STRUCTURAL AND GE0 TECHNICAL ENGINEERING BRANCH STRUCTURAL ENGINEERING SECTION A 1. The ultimate capacity calculations by GE for the steel containment based on limit analysis using the ultimate strength of SA 516 grade 70 steel may be questionable. Please_ provide justification for such an approach. 2. The discussion oil fracture, especially in weldments is not presented in a manner that could be evaluated. The conten; ion that cracks will develop only when stresses are between yield and ultimate strengths should be justified. 3. The use of Equation G.8.25 for calculation of buckling of the knuckle region geometry should be justified. It is also not clear how thermal effects are factored into the calculation. ~.T In reviewing the pressure time curves for hydrogen detonation 4 (i.e., Fig. G.10-1) with the BNL accident analysis group, we were told that the figure may not represent the time pressure phenomenon accurately. Any impact due to the inaccuracy of the time pressure phenomenon on the containment capacity should be further assessed. 5. The use of dynamic load factors to represent the dynamic effects of H,3 detonation in the analysis of complex structural systems / uhdergoing large plastic deformations is questionable. Provide a discussion to justify such a usage. 6. PP.597:, The loss of integrity has been assumed to occur when either the ultimate tensile strength in the high stress recion is reached or cracks develop. Reasons for not using other criteria, e.g., maximum strain in the steel containment, are not given. Please discuss the basis for not using other failure criteria. 7. PP.600: ' Details of determinirg the crack size in concrete (1/4 in. widtn) are not provided. Provide the basis for the crack size determination. ~ 8. PP.607: Temperature range that was considered in the analysis has not been identified. Please identify the range and the basis there-of. 9. PP. 609: Details of stress calculations in the knuckle region (laole G.2-1) are not given. Provide details of the stress calcu-lation. 10. PP. 609: Formula for d' in the column for P should be pr/2h instead of pr/h. In th6 same column y shouid be replaced by g.
g. .g. ;.,, . ;.\\ . +. g_. ;.3p.- .. g ,j.., .. : 2;; v..m w s
- y.
L. 11. PP. 610: Is the radius of the containment really 80'ft. as shown I, in Figure G.2-1? If not, provide the correct radius of the containment. 12. PP. 61i: The ring-stiffeners and the crane-girder used in the ~ finite element model (Figure G.2-2) are not described in the report. Please describe these items in sufficient detail for use i in analysis. 13. PP. 616: The calculations for P in Region 4, and for Pg+Pb I" Regions 1, 3 and 4 need to be clErified. 14. PP. 619: It is not clear what the compatibility condition (Eq. G.4-3) represents. Provide a discussion. .y. - r. v 15. PP. 620: Details of the pressure carrying capability of,the ECCS' "7 lines are not given. Please provide a more detailed discussion. 16. PP. 655: 5 The details of the calculation for li = E056 x 10-3 '.,. ~ should be provided. ~ f .17. PP. 655 & 656: The notation used in Eq. G.9-2 is not consistent with Eq. G.9-3. In my opinion, in Eq. G.9-2, F (s) should be Z replaced by f (s) and x should be replaced by s 7 18. PP.'656: How was the expression for the ultimate moment [i.e., 1.5 Mo +.(S /Sy
- 1) M ] derived?
ult / 19. PP. 656: What do the cases A, B and C represent? Please explain. 20. PP. 650: What is the basis for X = 51 psig shown in Fig. G.9-17 Please provide a justification. ~21. PP. 6595 What is the definition of " normal deviation" in Fig. G.9-1 and. Fig. G.9-27 22. How does the Appendix G fit into PRA analysis? How are the results of Appe'ndix G utilized in PRA? What specific end products of Appendix G are required by PRA? ~ 23. Provide bases and calculations that backed up the results of pressure capabilities listed in Tabies G.1-1 and G.1-2. Also' a g d( explain _ho.w the conclus. ion,that! ,s ,l 46 24 Provide background information of the ASHSD com; uter program (Ref. G.2-1). Can this program, which is based on axisymmetrical finite element shell model, be used for non-axisymmetrical loading? If yes, please explain how it is achieved. L.
C ;. . r- ..;n .~ p -3 l 25. Indicate the significance and location where various stresses listed in Table G.2-1 occurred, such as P, P), P, Q, etc. m b 26. At the lower portion of containment shell the 8' concrete shell has been treated as "thjn" shell '(Equations G.4-2 and G.4-3). What magnitude of error has been introduced by this assumption? compare the calculated capability pressure of 74.9 psi (p.15.D.3-619) with the calculation of the stress in the steel shell when the concrete wall is treated as thick shell. 27. Typographical error on p. 15.D.3-63S? Unit of ( should be in psi and not the reciprocal of psi. 28. NE 3133 of ASi4E Section III code consists of design formulas and procedures for shells under external pressure. It is not clear what is meant by "the buckling criteria given in NE 3133." (p. = 15.D.3-638). ~ ~ ~ ' 29. It is not clear how the failure mode cf maximum shear be considered. It is mentioned in Section G.2, but not in G.8. 30. Define X in Eq. G.8.10 and P in Eq. G.8.25. Is Q in Eq. G.8.4 transverTe shear or in-plane Ehear? There are mixed-ups of notations in App. G: ( has been used for stress and standard deviations; P for pressure, load, and probability; and stress can be r, 5, or f. ) 7 31. What is the relation between : loss of integrity and ' failure 7 What is the difference between " fracture" and " crack"? What is '," plastic yielo"? Is it something different from " yield" or " elastic yield"? 32. What are the bases of making the following assumptions? a. Structural capability depends only on geometrical dimension and yield strength of material. b. Normal distribution of probability of yield and ultimate g ength at testing. 4 s G A0 d. i ~ a 4 Y
.; f ***'.- h l'Y f.*{, fhj..Y5$ f E'$ $l. { ' 7 sl. ; N . _ f-i. *, y _' ^ l .., y l.f.,.. l, .[ ;." ~ {li . 7 j g..p. y.,. 9,..c q :s.:. ..3: 9t:7.f ~g. ~ 33. What is the physical significance that DLF is less than 17 Should it be required that static load be used when DLF is less than 1 such as the case indicated in Table G.10-57 34. Describe how and why does a local detonation affect the structural response (e.g., location and distribution of the pressure pulse, shock wave propagation and refraction / reflection, thermal effects) and how are the dynamic load factors obtained. Indicate locations of potential failures in containment or at drywell and the probability corresponding to each failure. ~ 35. In assessing the response to non-condensible gas generation or to local and global hydrogen combustion, it is stated that loss of containment integrity would eventually occur.in the torispherical domelegion.-. (15.D.3-661, 662). What are the physical failure.. . l'. " ~ boundaries?.How would it affect the release of radioactive material to the environment? ~ 36. If the' pressure-carrying capability in the torispherical reg' ion is [' significantly higher than that predicted by the analysis, what would be the worst impact on steel containment due to a hydrogen detonation? eo, / e. k 4 i 9 I k
e-8 REVIEW AND EVALUATION OF THE GE55AR.!! PROBASILISTIC R!s'K'ASSE5SMENT - CONTAINMENT FAILURE H0 DES AND FISSION PRODUCT RELEASE Letter Report by Accident Analysis Group Department of Nuclear Energy Broonnaven National Lacoratory Upton, New York 11973 July 27,1983 Prepared for Division of Systems Integration Office of Nuclear Reactor Regulation U. 5. Nuclear P.egulatory Commission Wasnington, D. C. 20555 m= m,, c,, n,,ia,,( e, t. f ,w
i i 1 ABSTRACT This letter report provides the status of our continuing review of those as-l pects of the GESSAR-!! PRA that relate to the determination of potential con-tainment failure modes and the magnitude of fission product releases. The re-l view includes a technical assessment of the assumptions and methods used in the GESSAR-!! PRA. At this stage in our review we have concentrated on a re-evaluation of transient events initiated by Loss-of-Offsite Power. In parti-cular, the appropriateness of the containment event trees, core meltdown mod. eling and fission product release calculations in the PRA has been assessed and our concerns noted. Senkitivity studies are presented that show the 90-tential impact of these concerns on risk. Finally, the future direction of l our review of the GESSAR-II PRA is indicated. l 1 iii
c
- .3 (4-g.
.s, ~ / PROJECT STAFF W. T. Pratt is the principal inve,htigator for this aspect of the BNL review of the GESSAR-!! PRA. R. Gasser is reviewing the core meltdown phenomenology, l .. containment response and fission product behavior. H. Ludewig is reviewing the fission proouct behavior and offsite consequences. W. S. Yu is reviewing the3 release of fission products from the damaged core. R. Jaung has checked -thi plant specific input dat'a used in the PRA and is now assessing the loads on containment stro:tures.during H2 detonations. Two~ BNL consultants, Pro-l fessor Berlad (SUNY) and Professor Catton (UCLA), are also reviewing specific l aspects of the GESSAR-II PRA. I. Papazoglou is principal investigator for the review of the core damage fre-l 'quency assessment (under a separate NRC contract with the Division of Safety .. Technology). Two mencers (N. Hanan and K. Shiu) of his group are participa-l ting in the review of the Contr.in.er.t Event Trees (Section 2). l' R. A. Bari is responsible for coordinating all elements of the BNL review of j the GESSAR-II PRA. This aspect of the GESSAR-Il PRA review is being perfomed for the Division of Systems Integration (DSI) at the U. S. Nuclear Regulatory Commission (NRC). l . J. Meyer, B. Hardin and J. Mitchell are the NRC technical monitors for tne project a'r.d J. Rosenthal is the NRC section leader. Finally, we would like to express cur appreciation to Terri Rowlanc for her excellent typir.g of this letter. report. i i 8 't i l l l l ') 1,- i at
v Ei CONTENTS-Page . ABSTRACT............................... iii L PR OJ E C T STAFF............................. iv LIST OF FIGURES............................ vii LIST OF TABLES............................ viii 1. . INTRODUCTION........................... 1-1 1.1 Background.~......................... 1-1 1.2 Objective, Scope and Approach to Review......... 1-1 .1.3 Organi zation of Report................... 1-2 1.4 References to Section 1................... 1-3 2. CONTAI NMENT EVENT TREES..................... 2-1 2.1 Appropriateness of Containment Event Trees......... 2 2.2 Event Trees Related to H2 Phenomena...........~.. 2-4 2.3 Quantification of the Containment Event Trees........ 2-7 2.4 References to Section 2.................... 2-8 ~3. CORE MELTDOWN AND CONTAINMENT RESPONSE.............. ' 3-1 3.1 MARCH Modifications Specific to the GESSAR-II PRA Analysis.......................... 3-3 3.1.1 Multi-Compartment Containment Modifications..... 3-3 3.1.2 HPCS Pump Curve Model................ 3-4 3.1.3 Drywell to Wetwell - Bypass Model........... 3-4 3.2 MARCH Sensitivity Studies.................. 3-5 3.2.1 Comparison Between GE and BNL Versions of MARCH Code......................... 3-5 3.2.2 Effect of Primary System Water Inventory....... 3-6 3.2.3 Effect of Suppression Pool DFs............ 3-6 3.2.4 Effect of Multi-Compartment Model.......... 3-7 3.2.5 Effect of MARCH Input Parameters on H Ge ne ra t i o n............. 2 3-9 3.2.6 Effect of Time Step Control During Core / Concrete Interactions..................... 3-11 3.2.7 Effect of Containment Failure Area.......... 3-12 3.3 MARCH Analysis Used as. Input to CORRAL........... 3-12 3.4 References to Section 3................... 3-12 4. FISSION PRODUCT RELEASE AND TRANSPORT.............. 4-1 l l i e s y
CONTENTS (Cont.) Pace 4.1 Release Categories Used in the GESSAR-II PRA........ 4-3 4.2 BNL CORRAL Modifications.................. 4-6 4.3 Comparison Between GE and BNL versions of CORRAL Code............................ 4-7 4.4 Impact of RSS Source Term tiethodology............ 4-7 4.5 Sensitivity Studies..................... 4-10 4.5.1 Fission Product Release Rate Constants Based on NUREG-0772.................... 4-10 4.5.2 Pri ma ry Syst em Hol d -u p................ 4-12 4.5.3 Suppression Pool Decontamination Factors....... 4-14 4.6 CORRAL Analysis Used as Input to CRAC............ 4-14 4.7 References to Section 4................... 4-14 5. RISK ANALYSIS.......................... 5-1 5.1 Comparison Between GE and SNL Versions of the CRAC Code... 5-2 5.2 Sensitivity Studies..................... 5-2 5.3 References to Section 6................... 5-3 6.
SUMMARY
AND FURTHER WORK..................... 6-1 APPENDIX A - BUR-Mark III Suppression Pool By-Pass Model....... A-1 APPET.iDIX B - MARCH Output Data.................... B-1 vi
LIST Of FIGURES F.i gure Title Page
2.1 'CTI-Pa
Loss of' Offsite Power (LOOP) 160 minutes...... 2-10 '2.2 CTI-Pb: Loss of Offsite Power (LOOP) J. 60 minutes...... 2-11 2.3 H2 phenomena event tree for Class 1 transients with power restored within 60 minutes of core melt... 2-12 2.4 H2 phenomena event tree for Class 1 transients with power restored after 60 minutes from start of core melt............................. 2-13 3.1 Effect of pool decontamination factors on wetwell temperature.......................... 3-17 3.2-Compartmental model used in MARCH............... 3-18 3.3 02 concentration in Compartment 1............... 3-19 3.4 H2 concentration in Compartment 1............... 3-19 3.5 02 concentration in Compartment 2............... 3-20 3.6 H2 concentration in Compartment -2. 3-20 3.7 02 concentration in Compartment 3............... 3-21 3.8 H2 concentration in Compartment 3............... 3-21 3.9 02 concentration in Compartment 4............... 3-22 3.10 H2 concentration in Compartment 4............... 3-22 3.11. Time periods during which Compartment 1 is either comoustible o r d e t o n ab l e......................... 3-23 3.12 Time periods during which Compartment 2 is either comoustible or detonable......................... 3-24 3.13 Time periods during which Compartment 3 is either combustible o r det o n ao l e......................... 3-25 3.14. Time periods during which Compartment 4 is either combustible o r d et o nabl e......................... 3-26 3.15 Effect of MARCH time step control during core / concrete interactions......................... 3-27 vii
LIST OF'? ABLES Table ' Title Pace 3.1 Comparison Between Predictions of GE and BNL Versions of MARCH.1.......................... 3-14 3.2 Time During which the Compartments Satisfy the Three Mi xt ure Crite ri a....................... 3-15 3.3 Percent of Time During Which the Compartments Satisfy the Three Mixture Criteri a.................. 3-15 3.4 Impact of MARCH Input Assumptions on Clad Oxidation...... 3-16 4.1 Fraction of core inventory released.............. 4-16 4.2 GE fission product releases and release characteristics..., 4-17 4.3 Comparison of Release Fractions as Predicted by GE and BNL Versions of the CORRAL Code for the I-T-L3 Category.... 4-18 -4.4 Comparison Between GE Sequence I-T-E2 Release, and BNL Release Based on RSS Methodology............... a-19 4.5 Effect of Core Release Fractions Based on the RSS Compared with GE Calcul ated Fractions.............- 4-20 4.6 Comparison of GE Based Release Fractions With and Without Selective Hold-up and Re-emi ssion............... 4-21 5.1 Comparison Between GE and Scheme 1 Evacuation Models..... 5-4 5.2 Comparison Between Senene 1 and Schene 3 Evacuation Models............................ 5-5 5.3 Comparison of GE and BNL CRAC Code Predictions for Sequence I-T-E2............................ 5-6 5.4 Release Fractions Developed in Sensitivity Studies for t h e I - T-E 2 Ca t e g o ry...................... 5-7 5.5 Impact of Sensitivity Studies for Accident Sequence I-T-E2.. 5-8 4 t viii
1. INTRODUCTION This section describes the background to the GESSAR-Il probaoilistic risk as. sessment (PRA), indicates how the PRA is being reviewed at Brookhaven National Laboratory (BNL), and describes the way the letter report is orga'nized.
1.1 Background
GE submitted a PRA for their BWR/6 standard plant and reference MARK 111 Con-tainment on the GESSAR docket on March 19, 1982. The stated objective of the PRA was to assess the core damage probability and risk associated with the GE standard plant. GE considered that this was accomplished by evaluating the t frequency and consequences of postulated accident sequences by using state-of-the-art methods evolved "from those developed for the Reactor Safety Study (RSS).[13 The NRC placed two contracts at BNL for an in-depth review of the GESSAR-Il PRA. One contract was issueo by the Division of Safety Tecnnology (DST) to perform a review of the frequency of core damage in tne PRA. This review is being performid by the Risk Evaluation Group at BNL. A memorandumE2) de-scribing the status of this aspect of the BNL review was recently issued. Tne memorandum concentrated on a review of the most probable accident sequences, namely transients initiated by Loss-of-Offsite Power (LOOP). A separate contract was issued by the Division of systems Integration (051) 4 for a review of potential containment failure modes and the magnitude of fis-sion product releases. This review is being performed oy memoers of the Acci-dent Analysis Group at BNL ano this letter report provides *he status of tnis aspect of tne BNL review. Again, empnasis is given to tne most probaole acci-dent sequences, namely Class 1 transients. 1.2 Objective, Scope and Approach to Review The objective of the DSI contract is to assess the containment event trees, core meltdown modeling and fission product release calculation; contained in the GESSAR-II PRA. This review therefore relates specifically to the follow-ing sections of the PRA; 1-1 n ---r r-,
Section 4: Probability of Radioactive Release l Section 5: Magnitude of Radioactive Release Appendix C.18: Containment Event Trees Appendix D.1.7: Containment Event Tree Quantification Appendix 'F: Description of Computer Models and Methods Appendix H: Prienomena of Steam Explosions Appendix 1: Hydrogen Phenomena The review process was started by several meetings held between. NRC, BNL, and GE. A total of five Technical Update Meetings (refer to References 3-7) were held. These meetings provided additional information that greatly assisted the review process. First round questions (01s) were sent to GE in December, 1982. The GE[8] responses to these questions and the drafts of tne proposed modified Appendices C and D to the GESSAR-Il PRA provided much needed clarifi-cation. Second round questions (Q2s) were preparea curing May 1983. This letter report describes tne current status of our review. Tne report takes into account all of the information provided by GE in References 3 through 8. It does not, however, reflect GE responses to Q2s. Tnis letter report should therefore be considered as a working document. The results in our final report could be effected by resolution of some of the concerns ex-pressed in tne Q2s and in this letter report. An attempt is made to clearly identify areas of concern and assess how they could potentially impact risk. Further work necessary to resolve these concerns is indicated in Section 6. 1.3 Organization of Report Section 2 reviews the containment event trees used in the GESSAR-II PRA. The appropriateness of the trees are assessed and possible revisions to them sug-gested. The effect of these revisions on risk is calculated. These calcula-tions are limited to the effect of only the revisions in Section 2 and do not, at this time, reflect changes due to revisions identified in subsequent sec-tions of this letter report. Section 3 reviews core meltdown modeling and containment response. The MARCH code was used by GE and the predictions in the PRA are compared with BNL pre-dictions. MARCH code modifications, made at BNL specifically to assist in our 1-2
review of tne GE5SAR-Il PRA; are described. Sensitivity studies are also pre-sented. The review of fission product release and transport is described in Section 4. Alternative release fractions for transient sequences resulting from Loss-of-Offsite Power (LOOP) are presented. The timing of release of the fission pro-ducts from the primary system and the potential for suppression pool bypass are given particular consideration. The effects on risk of the revisions identified in Sections 3 and 4 are cal-culated in Section 5. Differences between the version of the consequence moo-el (CRAC) in use at GE and BNL are also identified. Finally in Section 6, tne results of this letter report are summarized. In addition, the future direc-tion of our review is indicated. 1.4 References to Section 1 1) Reactor Safety Study, "An Assessment of Accident Risks in U. S. Commercial Nuclear Power Plants," WASH-1400, NUREG/75-014, October 1975. 2) N.
- Hanan, K.
Shiu and R. Ka rol, " Status of GESSAR-Il PRA Review," BNL memorandum to 1. A. Papazoglou, dated May 5,1983. 3) BNL Letter Report to J. Meyer (RSB/OSI/NRC) from W. T. Pratt (BNL),datec June 18, 1982. 4) BNL Letter Report to J. Meyer (RSB/DSI/NRC) from W. T. Pratt (BNL), dateo July 15,1982. 5) BNL Letter Report to J. Meyer (RSB/OSI/NRC) from W. T. Pratt (BNL), dated August 20, 1982. 6) BNL Letter Report to J. Meyer (RSB/DSI/NRC) from W. T. Pratt (BNL), dated October 27, 1982. 7) H. Ludewig, S. Fiarman and R. Ga s ser, "GE-NRC Tecnnical Update Meeting
- 5," BNL memorandum to W. T. Pratt, dated November 1982.
8) Letter from G. G. Sherwood (GE) to D. G. Eisenhut (NRC), " Submittal of Proprietary Information in Response to Request for Additional Information Regard Severe Accident Portion of the GESSAR-II; Docket No. STN50 447," dated January 31, 1983. 1-3
2. CONTAINMENT EVENT TREES -In this section we review the containment event trees used in the GESSAR. II PRA. Tnis aspect of our review is therefore specifically related to the following sections of the PRA; Section 4: Probability of Radioactive Release Appendix C.16: Containment Event Trees Appendix D.1.7: Containment Event Tree Quantification Appendix'I: Hydrogen Phenomena _ In addition, further information on the Containment Event Trees was providea at the Second[13 GE-NRC Technical Update Meeting. Also,. the GE response to Question 720.40 was particularly useful when attempting to understand the var-ious conditional procabilities for H2 pnenomena reported in Appendix 1 of the PRA. In Section 2.1, we discuss the appropriateness of the containment event trees. In particular, we are concerned with ensuring that tne trees include all of the potentially important phenomena and failure moces. Then, in Section 2.2, we review in detail the event trees developed by GE to describe potential H2 phenomena. The quantification of the conditional probacilities associated with the various branch points in these trees is dependent on MARCH analysis (refer to Section 3). The quantification of the containment event trees is discussed, in detail, in Section 2.3. Each branch point conditional probacil-ity is reviewed and revisions suggested where appropriate. Containment event trees are, of course, used to determine the conditional probability of a given containment ouilding failure mode. Each failure moce has a representative release category and potential consequence. The impact of tne revisions to the containment event trees suggested in Sections 2.2 and 2.3 could be assessed by calculating how they changed the concitional proba-oilities of the various failure modes and hence the consequences. However, if we use the release categories and associated consequence calculations reportec in the GESSAR-II PRA, significant changes in the probabilities would have very little impact on risk. This is simply because all of the six release categor-ies associated with Class 1 transients were calculated in the PRA to have very low fission proauct release fractions and hence consequences, as shown below; 2-1 L
o The above release categories are described in detail in Section 4.
- However, it is clear that any change in the conditional probabilities of the failure modes and hence the above release categories will have very little impact on risk.
It would, therefore, be a ratner futile exercise to assess the impact on risk of the changes in Sections 2.2 and 2.3 using the above means. It is, however, clear from an inspection of Section 4 and 5 of this lette'r report that c'hanges made to the various release categories at BNL will result in ratner more sensitivity to the containment event tree conditional probaoili-ties. When a complete set of BNL release categories has been detenninec, tne impact on risk of the changes described in Section 2.2 anc 2.3 will be more appropriately displayed. In the ' pr~evioQs ' ~se~ction,' we noted ' t' hat this letter report conce'ntrates on ~ Class 1 transients. Consequently, this section reviews in cetail tne contain-ment event trees appropriate to Class 1 transients. We nave, however, re-viewed the other containment event trees 'anc' founc inconsi'stencies.' 'We will not incluce our concerns regar:ing tnese otner containmeat eveat trees in tns s hr c:nce-.s ave :ee ex; ressac 'r tne :2s1 aac, :ece c- 'e-ea ae:ce:. ing on GE responses, will de accropr ately accresse: in tne final report. 2.1 Aopropriateness of Containment Event Trees Class 1 transients initiated by LOOP are represented, in the GESSAR-II PRA, by two separate containment event trees. The total probability associated witn Class 1 transients is subdivided into the probability of restoring power with-in 60 minutes of the start of the core damage and the probability of restoring power after 60 minutes. The CTI-Pa containment event tree (reproduced in Fig-ure 2.1) represents sequences with power restored within 60 minutes, and the CTI-Pb tree (reproduced in Figure 2.2) represents sequences with power re-stored after 60 minutes. The containment event trees are described in 2-2
Appendix C.16 of the PRA. However, the quantification of th( branch point probabilities used in the event trees is described in Appendix D.1.7. In par-ticular, Table 0.1.7-2 of the PRA provides the basis for the probabilities used at each of the branch points in the containment event trees." The as- ' sessed probabilities are based on, input from Appendices A and C as well as from Appendix ! of the PRA. Appendices A and C are being reviewed [43 by the risk evaluation group under contract to DST /NRC.. The results[43 of their review impact some of the branch point probabilities in the containment event trees. This is discussed further in Section 2.3. Appendix I deals with H2 phenomena and is reviewed in Section 3 of this re-port. The results of Appendix I are summarized in event trees (reproduced in Figures 2.3 and 2.4) and conditional probabilities for the various H2 phe-4. nomena. The conditional probabilities in Appendix ! of the PRA (refer to Ta-ble I.4-1 in the PRA) were transferred to Taole D.1.7-1 in Appendix D.1.7 and used as inp0t to the containment event tree quantifi. cation process (refer-to Tabl D.1.7-2 of 'the PRA). The GE containment event trees appear to be comprehensive in the selection of potential containment failure modes. tie will discuss the appropriateness of the branch point split fractions in the following sections. However, the most obvious failure modes tha; have been omitted by GE are steam explosions and basemat penetration. We discuss these failure modes in Section 3. At this stage we would be reluctant to give these failure modes zero probability as was done by GE. Steam explosion induced failure of containment is now con-sidered less probaole (refer to Section 3) than in the RSS.[3] However, tne role of steam explosions in determining fission product release fractions re-mains of concern.[23 The approach taken by GE in assuming containment fail-ure via overpressurization prior to basemat penetration (and hence eliminating it as 'a potential failure mode) appears to be a conservative assumption in terms of health consequences (refer again to Section 3). Finally, sequences that bypass containment have been given very low probability in the PRA. The s failure mode (using the notation of the RSS) is considered improbable as is the potential for excessive leakage through the MSIVs. These potential fail-ure modes are currently under review and will be addressed in greater detail in the final report. 2-3
2.2 Event Trees Related to H2 Pnenomena In this section we review the event trees (refer to Figures 2.3 and 2.4) used in Appendix I of the PRA to develop the conditional probabilities incorporated into the branch point probabilities -of the containment e' vent trees (refer to Figures 2.1 and 2.2). 'The development of these H -related event trees was 2 described in detail by GE in the response, to Question 720.40. Basically, GE subdivides all H2 phenomena into four separate and non-inter-acting categories; namely: Local combustion: Combustion of H2 involving less than 50 percent of the entire volume (e.g., compartments, pockets) within a structure. Com-bustion is defined as the burning of local H2 concentrations between 4 and.18 volume percent H2 in air. Global combustion: Combustion of H2 involving more than 50 percent of the entire volume within a structure. The H2 concentration is approxi-mately uniform within the volume. Local detonation: A detonation involving less than 50 percent of the en-tire volume within a structure. The H2 concentration is not uniform within the entire volume. GE assumes that detonations can occur at H2 concentrations between 17 and 59 volume percent-H2 in air. Global detonation: A detonation involving more than 50 percent of the entire volume within a structure. The H2 concentration is approxi-mately uniform within the volume. We are concerned that all combustion phenomena is characterized in the GESSAR-II PRA in terms of the above four non-interacting categories. We can envision how a local deflagration could give rise to a local detonation, which may in-volve peak pressures ' compatible with those associated with the global detona-i tions calculated in the PRA. These concerns were expressed in the Q1s and also in the Q2s. However, for the purpose of our review of the containment event trees, we will use the four category designation utilized in the PRA. Entry to the event trees (Figures 2.3 and 2.4) requires the assumption of ig-nition and then the conditional probabilities of achieving each of the four combustion processes described above are defined. The probability of I 24 y i.-,-,,,,
achieving any of the H2 phenomena depends upon the accident sequence. and , ignition sources. The trees also include an estimate of whether or not the ' combustion process also results in a continuous burn. The potential for a ' continuous burn is important as the flame could damage penetratio'ns 'in the drywell wall and open a bypass flow path to the wetwell -(i.e, bypass the sup-pression pool). The probability of any of the four combustion categories occurring depends on the. amount and rate of H2 generation and distribution within t'he containment building. Case l, shown in Figure 2.3, corresponds to a Class I transient in which power is restored during the first hour after the start of core melt. (_ .l~ l m N l s l D l In order to check the distribution of conditional probabilities in Figure 2.3,( ~ the MARCH computer code was modified to provide a multi-compartment represen C tation of the ref.erence Mark III containment. 'In this way we were able to calculate the concentrations of H, 02 and H O vapor in the various sub-2 2 y . compartments' of the Mark III containment. The MARCH modifications are de-[ scribed in Section 3.1.1. The MARCH predictions, using the multi-compartment model, are described in Section 3.2.4. 'The potential for achieving combustion r/ or detonations in the various compartments is indicated in Table,3.3 n in-spection of Table 3.3 shows clearly that we predict local detonations ith a much higher conditional probability than assumed by GE in Figure 2.3. Local combustion events may also occur but we would given global events (combustion g or detonations) virtually zero probability. The analysis f ri Section 3.2.4 would significantly change the cps in Figure 2.3, as shown below; 2-5 ..--w... w ~4 y ..p->-m ,m.- ,,yw -r== c In Section 3.2.4, we again present our calculations of H2 concentration in the various compartments within containment for this perioo of time. Figures 3.11 througn _3.14 indicate when the various compartments reach comoustible or detonable limits and show widely dispersed time periods, it is therefore dif-ficult for us to' translate the information in Figures 3.11 and 3.14 into a . single containment event tree (CTI-Pb) as 'was done in tne GESSAR-II PRA. A simple approacn would be to ratio :the time that the various limits are ex-ceedeo ag,ainst the total time period to give an estimate of Ine conditional probaDilities. However, the proDability of restoring power curing tnis ex-tended period is not linear and this approach could De misleading. We are in the process of dividing up the time period to more appropriately reflect tne various H2 phenomena occurring in the containment building. This will De discu,ssed further in the final report. However, at this stage of our review we are still predicting local phenomena during this time period witn a mucn higher probability than assumed by GE in the PRA. It snould, however, be noted that the H2 concentrations calculated in Section 3.2.4 do not reflect the range of uncertainty regarding H2 generation noted in Section 3.2.5. In addition, the modifications made to MARCH to predict local H2 concentrations 2-6
O 4 are simpHs21c and do not take into account the ef f ects of H2 diffusion and bouyancy. These tf fects will tend to enhance H2 mixing and imply rather more uniform concentrations than predicted in the simple MARCH model. Further work is underway to calculate H2 concentrations that reflect these. ef fects. The impact of 'the above uncertainty regarding 'H2 concentrations, wnich im-pacts the conditional p'robabilities in the containment event trees, will be included in the final. report. 2.3 Quantification of the Containment Event Trees In this section we review the CTI-Pa and CTI-Pb containment event trees (refer to Figures 2.1 and 2.2) and, in particular, the basis for the probabilities used at each of the tranch points (refer to Table D.1.7-2 of the PRA). Al - though Table D.1.7-2 does provide a detailec description of how the branch point probabilities were derived in the GESSAR-II PRA, the description was not adequate for certain branches. The Q2s[23 include requests for further clarification of how these probabilities were calculated. However, this let-ter report does not have the benefit the GE responses to the Q2s. At this stage of our review, we therefore simply indicate how we would have performed the calculation. Column 1, Trees CTI-Pa and CTI-Pb The total probability for-Class 1 transients was calculated in Appendix C of the PRA to be 4.2E-6. Table D.1.7-2 of the PRA indicates how this probability was subdivided into the two time frames for recovery of power. These calcula-tions-were checked using the BNL power recovery curveE43 and althougn the total. frequency of the class was increased by a factor of approximately 9, the relative distribution between the two time frames remained similar. Column 2, Trees CTI-Pa and CTI-Pb The probability of ignition given power restoration was given a high value for w both time frames covered by the CTI-Pa and CTI-Pb trees. We would in general agree with the high probabilities allocated to these branch points. e - Column 3, Trees CTI-Pa and CTI-Pb This node addresses the potential for failure to isolate the containment. The probability (1.0E-5) derives from the failure to close the 42-inch HVAC lines.' 2-7
= GE censiders tnat the open containment prevents the possibility of global 4 ' events so _that only flame damage.to penetrations through the drywell can lead to a potential bypass of the suppression pool. These are potentially serious failure modes (in terms of release fractions) but are of very low probability in the PRA (refer' to Figures 2.1 and '2.2~)s./ We are, however, concerned about the probabilities used by GE for flame damage to drywell wall penetrations. This is discussed further under Column 8. In addition, the calculations de-scribed in Table D.1.7-2 of the PRA re. lated to f ailure of the HVAC isolation valves and to the signals to the valves need further clarification and are the subject of a Q2.[23 Columns 4 and 5, Trees CTI-Pa and CTI-Pb The branch point probabilities in these columns are taken directly from tne H2 phenomena event trees (refer to Figures 2.3 and 2.4). These trees were reviewed in Section 2.2 and the changes made would simply be inserted in these columns in the containment event trees. . Column 6, Trees CTI-Pa and CTI-Pb This branch indicates the total prooability of early loss of containment inte-grity and is the subject of a 02.[2] We would need to review GE responses-to this question before comenting further on the appropriateness of the value used in the PRA. Columns 7 and 8, Tree CTI-Pa - We are concerned that the probability of a continuous burn is included in the branch point prooability for both of these columns. By doing this we consider that GE has calculated a much lower probaoility of loss-of-drywell wall inte-grity.. than should be _ the case. This concern is also the subject of a Q2 and we would again wish to review GE responses before making a final recommenda-tion for the probabilities of these branch points. [ 2.4 References to Section 2
- 1) BNL Letter Report to J. Meyer (RSB/DSI/NRC) from W. T. Pratt (BNL), dated i
July 15, 1982. '2) BNL Letter Report to J. Meyer (RSB/DSI/NRC) from W. T. Pratt (BNL) dated May 18, 1983. 2-8
3) Reactor Safety Study, "An Assessment of Accident Riiks in U.S. Comercial Nuclear Power Plants," WASH-1400, NUREG/75-014, October 1975. 4) N.
- Hanan, K.
Shiu and R. Ka rol, " Status of GESSAR-Il PRA Review," BNL memorandum to I. A. Papazoglou, dated May 5,1983. 4 e 1 2-9
_= 3. CORE MELTDOWN AND CONTAINMENT RESPONSE In this section we review the assessment of core meltdown and containment building response in the GESSAR-II PRA. This aspect of our review is there-fore specifically related to the following sections of the PRA; Section 5: Magnitude of Radioactive Release Appendix F.2: Core Damage and Containment Response Appendix H: Phenomena of Steam Explosions Appendix I: Hydrogen Phenomena In addition, further infomation on the core meltdown modeling in the PRA wat provided at the Seconde 13 GE-NRC Technical Update Meeting. Also, the GE re-sponses to questions 720.36 through 720.42 and to question 720.84 provided further clarification of the assumptions and methodology used in this area of the PRA. In Section 1, we noted that this letter report is restricted to a reassessment of only the most probable accident sequences identified in the PRA and in the - BNL review [2] of the PRA, namely, transients initiated by LOOP. In the PRA, these sequences are denoted as Class 1 Transients (Class I ). Other sequen-T ces of lower probability are currently being evaluated and will be presented in our final review report. N was noted in Section 2 that six potential failure modes or release paths were identified in the PRA. Of these six modes, two (i.e., the Y and 5 modes) ~ / are caused by long-tenn gradual overpressurization. The other failure modes ( Y', Y ", u and p') result from various H2 related phenomena. It was as-t sessed in the PRA that approximately 14% of the total core melt frequency ~ 3( would result in failure via the Y mode, whereas approximately 86% results in failure via H2 related failure modes. Cl early, H2 phenomena are poten-tially very important contributors to containment failure in the GESSA -II PRA. Other potential failure modes such as steam explosions or penetration of \\ the basemat by the core debris are considered by GE to be not mechanistically '--possible for a BWR/6 standard plant with a reference Mark III containment. From an inspection of the reference Mark III containment, and based on our current understanding of potential core debris / concrete interactions, we are unable to confirm that the core debris would be pemanently retained in the containment building after a core meltdown accident. However, we consider 3-1
e that it. is likely that cortainment f ailure will occur either by overpressuri-zation or H -related phenomena prior to basemat penetration. In general, 2 the health consequences of containment f ailure by overpressurization are usu-ally more severe than from failure by basemat penetration. However, this con-clusion is site specific and care must be taken when applying it to a standard plant, which does not have a specific site. The rational that justified GE eliminating in-vessel and ex-vessel steam ex-plosions as potential failure modes for the BWR/6 standard plant is presented in Appendix H of the PRA. The phenomena associated with steam explosions has been under extensive investigation at Sandia National Laboratories (SNL) under the sponsorship of RES/NRC. Application of this research by NRR/NRC to an evaluation [33 of severe accidents in the Zion and Indian Point (Z/IP) facil-ities has concluded that the probability of a steam-explosion-induced failure of containment is much lower than assumed in the RSS.E43 However, both of the Z/IP facilities are pressurized water reactors (PWRs) with large dry con-tainments. A careful review of the reference Mark III containments is re-quired before i similar conclusion can be made regarding the probability of steam explosion f ailure modes. It does, however, appear that the probability will be significantly lower than assumed in the RSS. Although. we have con-cluded that steam explosion failure modes now have a lower probability than previously thnught, their contribution to the source tem remains an open is-This will be discussed in greater detail in Section 4 and is the suo-sue. ject of a Q2.[5] The remainder of this section deals with the overpressurization ( Y a. i ) and H -rel at ed ( 'l', 'f ", u and v ' ) failure modes identified in the PRA. GE used 2 core degradation and containme t de-the MARCH [6] computer code to model sponse. The MARCH code was developed at BCL and BNL has been actively in-modifying [8] the MARCH code. volved in assessing [73 and, when necessary, However, for the purpose of assuring a consistent mathematical model for use in licensing activities the MARCH 1.1 version, obtained from BCL in February 1981 has been frozen on the BNL computing system. This MARCH 1.1 version is therefore used in this report as our basic mathematical model for calculating core degradation and containment response. We are, of course, aware that the MARCH 1.1 code is being modified [93 by BCL Indeed, some of the BNL modifications [83 under the sponsorship of RES/NRC. 3-2
have been incorporated into the new version of MARCH 1.1 (namely, MARCH 2) by BCL. However, MARCH 2 is still under peer review and is not yet at a stage where it can be used directly in this review. We have, however, made a number of modifications to MARCH 1.1 at BNL that we consider pertinent to our review of the GESSAR-II PRA. These modifications are described 'in Section 3.1. BNL obtained[10] from GE a )(ARCH input deck for a Class IT sequence. In Sec-tion 3.2, we compare the predictions of the frozen MARCH 1.1. code in use at BNL with the predictions of the MARCH code in use at GE (as reported in Appen-dix F.2 of the PRA). ' The remainder of Section "3.2. presents sensitivity stud- ~ .ies designed to show how uncertainty associated with MARCH modeling assump-tions can impact the containment event tree conditional probabilities (Section
- 2) and the determination of fission product release fractions (Section 4).
The actual MARCH analyses used as input to the CORRAL code (refer to Section
- 4) is identified in Section 3.3..
3.1 MARCH Modifications Specific to the GESSAR-II PRA Analysis In order to carry out an analysis of the response of the reference Mark III containment to various core meltdown accidents, it was necessary to make some modifications to the MARCH code. Some of these modifications were general BWR related changes, while others were specifically related to modeling the Mark III containment configuration. These modifications are described in the fol-lowing sections. 3.1.1 Multi-Compartmental Containment Modifications Hydrogen production and concentration within the containment building has a strong effect on both the conditional probabilities used in the PRA (refer to Section 2) and on the phenomenology and severity of some accident sequences. It was decided to divide the containment into a system of separate communica-ting compartments in order to assess where in the containment building condi-tions might be established which could result in local or global hydrogen burns or detonations. Although the User's Manual [6] indicates that a maximum of 8 volumes may be modeled in the MARCH code, experience has indicated that without the proper code modifications, calculations which specify a number of compartments greater than 2 cannot be accommodated. A set of modifications [ll3 which accomplish this purpose and successfully extend the number of volumes to at 3-3 -..~
.=. leas?. 4, was developed at Oak Ridge National Laboratory. inea nocifscations were. obtained and installed in MARCH 1.1 at BNL for use in the present analysis. 3.1.2 HPCS Pump Curve Model. The ECC fl'ow rate as a function of vessel pressure for the High Pressure Core Spray (HPCS) system is not adequately modeled by either of the two HPCS models provided by MARCH 1.1. .Therefore, the " parabolic pump curve" model was re-moved and replaced by a correlation which more accurately describes the HPCS flow characteristics specific to BWR/6 Standard Plant Design. 3.1.3 Drywell to Wetwell Bypass Model The system of vents which allow the flow of gases from the drywell to the wetwell via the suppression pool cannot be modeled with the present version of MARCH 1.1. The vents open when a sufficient pressure differential between the drywell and the wetwell depressas the water level in a weir arrangement and progressively uncovers 3 rows of vent ports (45 vents per row). Thus, the ac-tual cross sectional flow area between the drywell and the wetwell depends on the pressure differential between the two compartments. Two options exist in the code-for calculating flow between cell s. In one model the cells are brought into pressure equilibrium at each time step. In the other, an orifice equation is employed to calculate the flow rates based on a constant cross-sectional area. The latter of the above two options (using the IVENT=12 option) was modified (refer to Appendix A).to provide two simultaneous flow paths from the drywell to the wetwell. One flow path consists of the fraction of the total vent area uncovered. The flow area for this path is calculated as a function of the pressure differential between the drywell and the wetwell. The flow through this pathway goes 'directly through the suppression pool. The second pathway is intended to model a potential -flow route which bypasses the - suppression pool. This bypass could result from a crack or hole in the drywell wall. Such a breach in the drywell wall could be caused by a failed penetration seal, failure of the wall under detonation loadings, or a vacuum breaker stuck open because of H2 burn or detonation loadings. Note that, since a differ-ential pressure of about 2.7 psid is required to depress the water level in the weir dovm to the top of the first row of vents, no flow will go through 3-4
2ne designed _ venting system until the differential pressure between the dry-well and the wetwell exceeds this value. Several cases were run with the modified MARCH 1.1 code in which two different bypass. flow areas' (1.0 ft2 and 2 0.2 ft ) were assumed. It was determined that for both of these, areas the bypass flow was sufficient to prevent the pressure differential from exceeding the 2.7 psid necessary to activate flow through the suppression pool. Hand calculations show that an opening as small as 4 inches in diameter will effectively eliminate flow through the suppres-sion pool for sequences in which operation of the ADS has resulted in primary system depressurization p. ior to vessel failure. 3.2 MARCH Sensitivity Studies In this section we compare the predictions of the GE version of MARCH (as de-scribed in Appendix F.2 of the PRA) with the BNL version of MARCH, namely, MARCH 1.1. This comparison is done in Section 3.2.1 and basically addresses the overpressurization failure mode (1). The remaining subsections presents sensitivity studies relative to the GE input deck that can potentially impact containment response during long-term core concrete interactions ( Y failure mode) and H2 generation and concentration (H2 related failure modes). 3.2.1 Comparison Between GE and BNL Versions of MARCH Code In this section we simply look at the results of running (unaltered) the MARCH input deck obtained[103 from GE on the frozen MARCH 1.1 version at BNL. The MARCH 1.1 predictions.are compared with the GE results in Table 3.1. In gen-eral, MARCH 1.1 predicts longer times to core melting, slumping and contain-ment failure. We believe that this is primarily due to differences in the de-cay heat models used in the two versions of the MARCH code. The original de-cay heat model in MARCH 1.1 was based on the 1971 ANS standard that neglected actinide decay. The latest version of MARCH 2 uses the ANSI / ANS-5.1-1979 standard and predicts an integrated decay heat, over the first hour after shutdown, about 20 percent greater than predicted by MARCH 1.1. The version of MARCH in use at GE apparently has the original MARCH 1.1 decay heat model (based on the 1971 ANS standard) replaced with a more up-to-date model. To assess the impact of actinide decay, the original decay heat model in the version of MARCH 1.1 at BNL was replaced with a decay heat correlation from 3-5
ORNL[123 and with the MARCH 2 decay heat subroutine.[9) Both of these models include actinide decay. The MARCH input deck obtained[10] from GE was run using both of the above decay heat models. The timing of major events predicted by the two models are also included in Table 3.1. The GE decay heat model appears to give similar predictions to the ORNL model. However, both ~ the ORNL and MARCH 2 models predict significantly earlier times to containment failure than the GE model. 3.2.2 Effect of Primary System tlater Inventory One of our early concerns (refer to Question 720.41) regarding GE's MARCH modeling was the elimination of 100,000 lbs of primary system water inventory. This was done to make the time of core uncovery as predicted by MARCH consis-tent with that predicted by the GE SAFE code. The GESSAR PSAR (Table 6.2-4) gives the mass of coolant in the primary system as 544,540 lbs. To assess the impact on suppression pool temperature and the timing of the event's, a case was run wherein the MARCH input parameters H0 and WDED were increased such that WMTOT, the total primary system coolant mass, was equal to tne value given in the PSAR. The results indicated an increase in the time-to-core-uncovery of about 5 minutes, and a difference in the suppression pool tempera-ture at uncovery of about 10.2'F. Subsequent cases, therefore, use the re-duced primary system inventory parameters with the additional 100,000 lbs added to the suppression pool. In addition, the initial conditions of the system are changes to reflect an initial suppression pool temperature of 100*F rather than 90'F as used in the GE input deck. This takes into account the energy associated with this quantity of coolant. 3.2.3 Effect of Sucoression Pool DFs Inspection of 'the containment building thermal response for the GE MARCH case revealed a highly superheated atmosphere in tne wetwell. This condition was traced to the decontamination factor, DCF, used in MARCH 1.1 to model the de-gree to which fission products released into the pool are transportea to the atmosphere above the pool. A value of 1.0 was used by GE for this parameter, which results in MARCH 1.1 predicting no fission product deposition in the suppression pool. The energy associated with the fission products is 3-6
therefore transmitted to the wetwell atmosphere. This is not consistent with the assumption made by GE of suppression pool decontamination f actors in the range of 600 to 10,000 (refer to Section 4). The high temperatures will af-feet pressure differentials between compartments and affect the intercompart-ment transfer rates, which feed directly into the CORRAL calculations (refer to Section 4). The effect on wetwell atmospheric temperature is clearly seen in Figure 3.1 which shows the corresponding temperatures using DCFs of 1.0 and 100. Additionally, the cell conditions predicted by MARCH (temperature, pres-sure) are used in the aerosol removal calculations in CORRAL and will, there-fore, be inaccurate. For all subsequent cases, the MARCH decontamination fac-tors were made consistent with the CORRAL analyses. 3.2.4 Effect of Multi-Comoartment Model The importance of hydrogen combustion and/or detonation on the accident pro-gression necessitated an assessment of the hydrogen concentration within the containment. To make such an assessment, it was necessary that the wetwell be divided into.a number of volumes so that the time dependent hydrogen concen-trations in each compartment could be calculated. Inspection of the reactor building (refer to Figure 3.2) suggested an appropriate scheme for subdividing the wetwell volume. Existing concrete floors at the 11'-1" and at the 45'-1" levels partially divide the annulus region into two volumes. The "open domed" region of the containment building was assumed to be separate from the higher of the two annular regions. Thus, hydrogen exiting from the suppression pool in volume 2 is transmitted upward into the second annular region (Volume 3) and then to the large domed area (Volume 4). The drywell is, of course, Vol-ume 1. The GE itARCH case was recalculated using the multi-compartment model with the heat structures appropriately redistributed among the four containment vol-umes. Hydrogen combustion was suppressed so that the predicted H2 concen-j trations would be undisturbed. It must be noted, however, that a H2 burn or l detonation in one cell, even if it did not result in burns or detonations in other compartments, would redistribute the hydrogen in a different pattern l than if ignition is suppressed. Graphs of the concentrations of the various I constituents in all of four compartments are shown in Figures 3.3 to 3.10. l 3-7 l
The concentrations of hydrogen, oxygen and water vapor determine whether the mixture is non-flammable, flammable or detonable. The water vapor concentra-tion must exceed about 0.5 mole fraction to render a mixture inert (not chem-ically reactive). The presence of the suppression pool, which condenses most of the steam,' prevents steam inerting in all four compartments. Figures 3.11 through 3.14 indicate the time periods over which the mixtures in each cell meet the conditions required for combustion, " marginal detonation," and deto-nation. Although mixtures with H2 concentrations above 0.04 are combusti-ble, concentrations in excess of 0.08 are necessary for complete combustion (all the H2 consumed in the reaction). A mole fraction of 0.1 has been taken here as the lower limit for complete combustion. The term " marginally detonable" implies H2 concentrations between 0.13 and 0.18 mole fractions. If the H2 is above a mole fraction of 0.18, the mixture is termed " fully detonanle.". In Figures 3.11 to 3.14, unity implies that the condition is. met and a zero implies that it is not met (for example, H2 concentrations above 0.1 or not above 0.1 for a comoustible concentration). The oxygen constraints are assumed to be 0.06 for both complete combustion and " marginal detonation" and 0.09 for " full detonation." The 0.09 criterion derives from a stoichio-metric reaction with H2 at a concentration of 0.18 mole fraction. An inspection of these figures will show that, in general, there is very lit-tle chance of a global detonation, that is one that encompasses the entire volume (all four compartments, or even the entire wet well) of the containment system. Neither the drywell nor the dome region of the containment building attain a fully detonable condition. This is detaileo more clearly in Taoie 3.2, which is a tabulation of the total time during which each compartment sa-tisfies the mixture criterion for each of the three reaction regimes. It is apparent, however, that most of the risk of detonation occurs in the two annu-lar regions (cells 2 and 3). The volume innediately above the suppression pool, for example, is detonaole during a narrow window (about 4 minutes) at about I hour after the beginning of the accident sequence, and again from ap-proximately 285 to approximately 320 minutes. The upper annular region, which apparently contains most of the potential ignition sources, is maintained in a detonable configuration between about I hour and 200 minutes and again between 290 minutes to 350 minutes. This volume is also marginally detonable between 50 minutes and 410 minutes. 3-8
Table 3.3 gives.the fraction (expressed as a percentage) of time during the first hour after incipient fuel damage (start of core melt) in which the three reaction criteria are satisfied. From this table it is seen that during the first hour, there is no contribution from either cells 1 or 4, but cell 2 is detonable for about 7% of the time and cell 3 can be detonated during" 80% of the period. As a consequence, provided that an ignition source exists during this period, there is nearly a 90% chance that a local detonation at high hydrogen concentration will occur in the annular region between the drywell wall and the containment shell. The way in which this information is used to quantify the conditional the probabilities in the CIT-Pa containment event tree is discussed in more detail in Section 2. 3.2.5 Effect of MARCH Input Parameters on H2 Generation The above predictions of hydrogen concentrations in the containment building were based on GE input assumptions to the MARCH code regarding the in-vessel , production of hydrogen during the core heat-up and slumping phases of the ac-cident. There are a number of input parameters to MARCH that can strongly af-fect the predicted quantity of hydrogen generated in-vessel. The values of these parameters selected by GE appear to result in hydrogen generation rates that are at the low end of the potential range. Thus, because of the sensi-tivity of H2 production to MARCH input parameters, and due to the uncertain-ties involved in these parameters, it is possible that the production of hy-drogen in-vessel has been underestimated in the GESSAR-II PRA analysis. In an effort to assess the relative sensitivity of H2 production to the modeling assumptions, a series of cases were run in which key parameters were varied. With regard to hydrogen production during core slumping, two para-meters are of particular importance; one determines the effective surface area of the debris (particle diameter, DPART), and the other determines the frac-tion of the particle that is composed of zirconium metal (fraction of zircon-ium in the central core of the particle, FZMCR). The MARCH model, which pro-vides the surface area for oxidation, consists of 3 zone spherical particles. The 002 is assumed to be in the central core while Zr02 and Zr can be dis-tributed throughout the 3 zones by user input. GE selected a combination of parameters which assumes an intimate mixture of UO, Zr02 and Zr in a 2 3-9
r-single rep on (the outer regions are eliminatec). They also selected an average particle size of I cm diameter. This model may be non-conservative regarding Zr oxidation with respect to both of these input assumptions. There may, in fact, be a separation of the metallic phase from the oxide which would tend to increase the surface area of Zr available for oxidization; also the particles may be smaller, which would likewise tend to increase oxidization. Two cases were run to determine the effects of these two modeling assumptions. In the first case, the particle zonal configuration was restructured in such a way as to place all of the oxide in the core region and all of the metallic zirconium in the outer shell region while the overall particle size remained unchanged. In the second case, the particle zonal configuration described above was retained and the particle diameter was reduced from 1 cm to 1 mm. The effect of these changes on clad reacted and hence hydrogen production are shown in Table 3.4 and compared with the original GE predictions. By restruc- .turing the particles the quantity of clad reacted during slumping increases from 4.3* to 18.0% (refer to case BNL1 in Table 3.1). The added effect of de-creased particle size, (refer to case BNL2 in Table 3.1), procuces 11 more dra-matic effect and increases the fraction of clad reacted to 56%. These two ef-fects combine to increase the total in-vessel H2 production by a factor of approximately 3. With regard to the modeling of H2 production during core heat-up, the para-meter that most strongly effects hydrogen generation in-core is the one which determines whether molten zirconium reacts with steam. The parameter in the B0ll subroutine unich controls this is IliUA. The applicant selected a value of I for this parameter, which prevents metal-wa er reactions in a melted node. A value of IfiWA=3, which allows the metal-water reaction to continue in a melted node, was utilized (refer to case BNL3) and the results are also shown in Table 3.4. The same slump parameters used in case BNL2 were also ap-plied in this case. Although hydrogen production during heat-up was increased by over 3% (to 24.2% from 21%) in this case, a more interesting observation is that H2 production during slumping was also increased by an additional 4.7%. This is probably due to a higher temperature in the core debris at the time of slumping due to the additional reaction heat. -The last sensitivity case produced over 85% of the cladding reacted, which, thus, represents 3.37 times as much hydrogen as was produced using the GE 3-10
3 s MARCH input assumptions. Although these cases may be conservative with re-- spect to predicting in-vessel hydrogen generation, they do serve to illustrate the sensitivity of Zr oxidation to relatively minor changes in MARCH input parameters. The impact of uncertainties regarding MARCH 1.1 predictions of .H2 generation will be factored into the final report. 3.2.6 Effect of Time Step control During Core / Concrete Interactions Before discussing the CORRAL analysis (Section 4), one more MARCH input para-meter must be mentioned. In terms of the CORRAL code's calculation of the disposition of radioactive fission products through, and out of, the contain-ment system, the most important parameters which CORRAL uses from the MARCH output are the cell-to-cell and cell-to-outside flow rates. As previously discussed, the MARCH calculations of flow from the drywell through the vents into the wet well makes use of an orifice flow equation. The accuracy of the orifice flow model depends on a sufficiently well controlled time' stepping procedure to prevent an artificial oscillating flow pattern from being estab-lished. This is particularly important when,two adjacent cells are at similar pressures. Subsequent to vessel failure, and after subroutine INTER has been called to calculate core-concrete interactions, the time step in MACE is ce-fined by the time step in INTER through an input parameter (NINTER). The MACE time step is calculated as: tMACE = NINTER x TINTER-The default value of 60 was used for NINTER in the GE analysis and this re-L sulted in very high amplitude flow oscillations. These can be seen clearly in the drywell pressure histories shown in Figure 3.15. By reducing NINTER to 1, I while keeping the remainder of the input unchanged, these oscillations can be l completely removed. Further studies have demonstrated that NINTER could be set as high as 10 before oscillations become a problem. It should be noted that these oscillating flows could provide an enhanced mixing process that really is not mechanistically present and which could result in conservative fission product releases from the drywell particularly if a suppression pool bypass exists. i l 3-11
- ,L,
3.2.7' lffect of Containment failure Area In the GE assessment of the Class IT accident sequences, the containment building boundary was assumed to faii ~at a specified pressure. The break was u assumed to be equivalent;to a 0.2 ft2 opening which allows the containment ' atmosphere to comriiunicate with the environment. Using an area of 0.2 ft2 results in a depressurigation rate cf about 7 psi /hr during the first 4 hours after failure.. Thus, for. thTT assumption, something in excess of 9 hours is e'uilibrium condition. In order to required to depressurize down to an q achieve a rapid blowdown following containr.ent failure, a break area of 7 ft2 was used in the BNL analysis. This break area results in complete con-ltaingent depressurization within 20 minutes (170 psi /hr). ~ 3.3 MARCH Analysis Used as Input to CORRAL The basic MARCH code configuration used in the CORRAL analysis in Section 4 reflects either directly or indirectly all of the concerns identified above in Section 3.2. Although the parameters which directly effect the in-vessel pro-duction of hydrogen discussed in Section 3.2.5 were not carried through to the CORRAL analysis, the containment was assumed to fail prior 'to vessel failure as the rer, ult of the loadings vroduced by a hydrogen detonation. A complete set of the MARCH output, data used as input: to CORRAL is included in Appendix B. The MARCH case was designated GSR17. 3.4 References tb Section 3 1) BNL Letter Report to J. Meyer (RSB/DSI/NRC) from W. T. Pratt (BNL), dated July 15,1932. 2) N. Hanan, ' K. ~Shiu and R. Karol, " Status 'of GESSAR-II PRA Review," BNL memorandum to. I. A. Papazoglou, dated May 5,1983. ' ~ 3) " Preliminary.tsessment 'cf dare Melt Accidents at the Zion and Indian Point Nuclear Power Plants and Strategies for Mitigating Their Effects - Volume 1, Analysis of Containment Building Failure Modes," USNRC Report NUREG-0850, ficvember 1981., 4) Reactor Safety Study, "An Assessment of Accident Risks in U.S. Cormiercial Nuclear Power Plants ~," WASH-1400, NUREG/75-014, October 1975. a 1 3-12
5) BNL Letter Report to J. Meyer (RSB/DSI/NRC) f roa ti. T. Pratt (BNL), dated May 18, 1983. 6) R. O. Wooton and H. I. Avci, " MARCH (Meltdown Accident Response CHarac- . teristics) User's Manual," USNRC Report NUREG/CR-1711, October 1980. 7) BNL presentation to the Class 9 Subcommittee of the ACRS, May 21-22, 1981. 8) W. T. Pratt, et al., " MARCH IB: BNL Modifications to the MARCH Computer Code," Proceedings of the International Meeting on Thermal Nuclear Reat-tor Safety, NUREG/CP-0027, Volume 2, pp. 1167-1176, February 1983. 9) Letter to W. T. Pratt (BNL) from P. Cybulskis (BCL), " MARCH 2 Code Des-cription and User's Manual," dated December 21, 1982. 10) D. A. Hankins, " Transmittal of MARCH and CORRAL Input Files to the NRC," August 25, 1982. 11) "0RNL Multi-Compartment Model," obtained through personal communication between S. Greene (0RNL) and R. Gasser (BNL). ORNL MARCH 1.1 modifica-tions are documented in BNL memorandum from R. Gasser to W. T.
- Pratt, dated July 26, 1982.
12) Letter from S. R. Greene (ORNL) to W. T. Pratt (BNL), dated August 2, 1982. 3-13
Table 3.1 Comparison between predic :en; ( GE and BNL versions of the MARCH code ~ MARCH 1.1 with Actinide Decay Timing of Events GE Version Frozen Version ORNL[123 MARCH 2[93 (Minutes) of MARCH of MARCH 1.1 Model Model Start of Core 48 53 49 47 Melt Core Slumps 92 102 88 78 RPV Ruptures 192* 215' 179 151 Loss of Contain-708* 770 668 606 ment Integrity
- Scaled from Figure F.2-2 of the GESSAR-II ?RA.
3-14 i
Table 3.2 Time during whicn the compartments satisfy the three mixture criteria Period in Which Criterion is Satisfied (min) Cell Marginally Fully Number Combustible Detonable Detonable 1 195 20 0 2 125 110 25 3 360 355 225 4 900 830 0 Total 1150 1145 230 Table 3.3 Percentage of time
- during which the compartments satisfy'the three mixture criteria Cell liarginally Fully Number-Combustible Detonabl e Detonable 1
0 0 0 -2 12 11 7 3 83.9 82.8 80 / 4 0 0 0 Total 94.4 93.6 87
- The time period is for one hour measured from the start of core melting.
i 3-15 ~
Table.3.4' Impact of MARCH input assumptions on clad oxidation 't e Percentage'of Clad Reacted Ca ses* During Core During Core Hea t-Up Slump Total GE Model 21% 4.27% 25.27% I~ BNL 1 21% 17.95 38.95% BNL 2 21% 56.22% 77.22% BNL 3 24.2 60.9% 85.1%
- Notes :
GE Model - No outer shells Zr, Ir0, UO2 nomogeneously mixed, 2 1 cm particles. BNL 1 One outer snell witn all Zr; Zr02 5 UO2 in core, I cm particles. BNL 2 One outer snell with all Zr; Zr02 & UO2 in core, 1 mm particles? BNL 3 IMWA = 3 3-16
i tm 0 w o 180 0-1500 - 34a0 O. 2g 130.0-I-- Z 12 0.0 - POOL DECONTAMINATION FACTOR = 100 11n0-Q. 2. 100 o - / O f U o0.0 - 804 CA 1004 200.0 300.0 400.0 500.0 000.0 700 0 800.0 900.0 1000.0 !!00.0 1200 0 1300.0 TIME - (MINUTE) 4000 350.0 - c:o D 3000-c: ta Q. 3 25a0-E-- t-- 2004-c POOL DECONTAMINATION FACTOR = 1 2 l g 15a0-2 8 1004-00D u CA 100.0 200.0 3004 400.0 500.0 000.0 700.0 800.0 900.0 1000.0 11004 1200.0 1300.0 TIME - (MINUTE) Figure 3.1 Effect of p001 decentamination factors On wetwell temperature. I 3-17 e/
I COMPARDENT 4 /swisto sulLDING e y,'ja" omvwELL l HEAO p
- ggg, D
/f!o h f 7 3 / %f snEt I M 0 ((h c "7AINMENT 3 y i g.;s-Si@ P.: . =ner -w +x .up 80 "'ti' s br.e$ s e~-msg
- ^"NL fl k
m g <h Ii-i h COMPARTMENT d t oj:;; "[' m ' C 4 i om a 3 3 1 Q \\M i f COMPARTMENT r I M C0:1PARTMENT 2 EN !s \\ W 2 U' CDQ suppmEss gq y -~) [ U SUP* R ES$10N L T ecol% j 1 PooLWALL IP EER 4 i Esi 9:! YET-3d / l l. ' P hDNTAINMENT COMPARTMEitT 1 ANesom4cs Figure 3.2 Compartmental model used in MARCH. 3-18
o om z tas 0 15 - O >=xo zo 0 10 - b cc En. sas 3o am-2 ODO i CD 1004 200.0 3004 400.0 SCC 0 600.0 *1000 800 0 900.0 1000 0 1100.0 1200D 1x00 TlME - (M1NUTE) Figure 3.3 0 concentration in Compartment 1. 2 oDs 040-z cisf 025-o D ~ 020-Z O. D 0.15-cc Ca. saa o.10 - r Jo2 om-oDo oD 100D 2000 lK104 4004 SOOD 0004 *100D 800 0 900 0 1000.0 1100.0 1200.0 1300.0 TIME - (MINUTE) i i Figure 3.4 H concentration in Compartment 1. 2 3-19 l
02b s 020-2 h2o >-x 015-ZO I U <m 0.10 - - k. En.1 ao2
- gas, b
CD0 no 100.0 200D 300D 400D $00.0 000.0 700 0 800.0 900.0 1000.0 1100.0 1200D 13000 TIME - (MINUTE) Figure 3.5 0 concentration in Compartment 2. 2 LO na-( Os-zo b< 0.4 - CY., Ca. Es3 2 a2-a0 CD IdOS 2dOD 3 DOS 4$00 SdOD 8$00 7$0.0 8$0.0 9dOD 10b0.0 11h04 12dno 13004 TIME - (MINUTE) Figure 3.6 H concentration in Compartment 3., 2 3-20
025 020- 7 2 to O >=x 0.15 - Z S U< Cr. 0 10 - En. to JO3 Ons-000 QO 100A 2004 3004 400.0 $000 000D 700.0 800 0 900.0 1000.0 1100.0 1200D 1300 0 TIME - (MINUTE) Figure 3.7 0 concentration in Compartment 3. ^ 2 as O - 0.5 - 8xC 04- ) ?.~. l ZO 03- / A 02 - ~ co 3 02 g. R0 R0 10a0 2000 3004 4004 $000 600.0 700 0 800 0 900 0 1000 0 1100.0 1200 0 1300 0 TIME - (MINUTE) Figure 3.8 H2 concentration in Compartment 3. 3-21
022 020-E Oas-o >-x0 0 16 - zO 0 14 - b< E 012-w 3 0 020-2 ode-CD6 s s CD 100D 200.0 300.0 400.0 500.0 000 0 700.0 800 0 900.0 1000.0 1100 0 1200 0 1300 0 TlME - (MINUTE) Figure 3.9 0 concentration in t0mpartment 4 2 025 020-8e Q 0.15 - zO U / 0 10 - Ca. w 2 h 0.06 = a00 0.0 100.0 200D 300 0 400.0 500.0 000 0 700 0 000 0 900.0 1000 0 1100 0 1200 0 13000 TIME - (MINUTE) Figure 3.10 H concentration in Compartment 4 2 3-22
n b d -g L L I e M did C -8 k 4 iE D' 5 se u h .%S g -n g
== = 9 2 F3 x C U h "3 O C-i ce u 94 -t .5 5 ' g b rv e e e 3 0 l C a \\ e N D \\ Q 'N Q O Y Y M h" e a w e a A M k D k 3-23
O "k u e CJ .g ~ b \\ = ~.~ W ~ 1 4 g= i Q \\ e e te - *3 -g % D 3; e, 5 a la 5Y= h-b )f 5 .92 t= y n =-g s 9= ~@ -- 5 il -u p N h m D % Uw hI kg .k k E x Q <w ~ = U tt D A &' c { _g e o q m io i i i i o o C N g . ts Y W Y Y w, w. m ee e k D k h 9 9 '3-24
p--- 9 h R -t x tf I eR .m Y t + t 5 w 2*t 51 ji ~ D
- n!
-% s( fi x e ~ ). n % m'u t n g 9 c. 4 D .JEi
- j g
- t-4 sE 0 -4 x h E x *t F w k k- -Rs h y es S I l 5 I I \\ D N g N g N g N ,o T C D' T d N a n a w-Q Q M A M 3-25
a t ke b l a e g s k L = 4 = m k ~k g 9 a 4 b9 E E D-Y e ,g a_ t ? =. .e i t 8
== s 9 22 &. 3 5 u. 3 O Yh P ~S E[ e-tdi E7 v I y k 38 w,; ~ b -k e b:?. L. D 'k i i i i ,o O N Q \\ Q C C W ( M or m w a A w 4 4 3-26
80 0 !.I ll y 100 - o piijf I I l%,j 800-E jjjjjf l t j // i F f z S00-If 2 l ' 40.0 - l c. E \\ u 30.0 i s NINTER = 60 [ if 20.0- ) lac 0.0 100D 200.0 300.0 400.0 5000 800.0 700.0 800 0 900.0 1000.0 1100.0 12C00 13000 TIME - (MINUTE) 80 0 y 700- 's N e0.0 - 5 l t-- l Z S0.0 - l E l 40s-c. 8 I O 30 0 - l a j$ NINTER = 1 l 20.0 - l 10 4 00 100.0 2004 300.0 400.0 5000 800.0 700 0 8000 900.0 1000.0 1100.0 1200 0 TIME - (MINUTE) Figure 3.15 Effect of *%RCH time step control during core / concrete interactions. (Note the time scales are different for the two pressure histories). 3-27
a. F15510N PRODUCT RELEASE A:tD TRANSPORT in this section we review the assessment in the PRA of fission product release - from the damaged reactor core and the transport of these fission products to the containment boundary. This aspect of our review is therefore specifically related to the following sections of the PRA; Section 5: Magnitude of Radioactive Release Appendix F.3: Fission Product Release and Transport in addition, the information provided by GE at the Third [1] ane Fourtn[2] GE-NRC Technical Update Meetings is relevant to this section and supplements the descriptions in the PRA. Also a number of Q1s were generated related to fission product behavior. GE responses to questions 720.33 and 720.80 provide further clarification of assumptions made by GE regarding primary system hold-up ano aerosol retention in cracks. The Q1s also contained many questions re-garding the potential for scrubbing of aerosols by the suppression pool, which strongly influences our calculations in this section. The determination of fission product (FP) release and transport in the GESSAR- !! PRA is a significant departure from other recently published " industry" PRAs (refer to references 3 through 5). In the past, PRAs have generally usec the prescriptions in the RSS[63 to detemine the release of fps from the fuel and the movement of these fps throughout the primary system and contain-ment building. However, there has been significant research activity in this area since the publication of the RSS in 1975. A basis for estimating FP be-havior was published [73 in 1981 by RES/NRC. In addition, upcated fission product source term methods are currently being developeo[6) and are re-ceiving extensive peer review. PRAs have generally recognized the potential influence of the new source tem methods but the impact has been expressed only in the fom of uncertainty (e.g., the level 2 risk curves in references 3 and 4). The point estimate risk curves for the PRAs in references 3 through 5 are based on the RSS pre-scription with regard to FP release from the fuel and transport in contain. ment. The GESSAR-II PRA moves away from this practice and incorporates new methods currently under development to determine the source tems. These 4-1
changes relative to the RSS are significant anc some of the more impo tant are listed below: Fission product release rate constants (based on NUREG-0772[73) e as a function of core temperature. Hold-up (permanent) of fission products in the primary system.[1] e e High suppression pool decontamination f actors based on GE experiments and modeling.[2] At this stage, BNL is unable to confirm the validity of the above changes rel-ative to the RSS approach. We are closely following the activities of RES/ NRC and their contractors and will factor into cur review of the GESSAR-Il PRA any developments made regarding new source term information at the appropriate time. However, prior to the release of this new source term information, it nas been cecided to follow two approaches in our review of tne GESSAR-Il PRA. Botn ap-proaches will use up-to-date information regarcing containment event trees, core meltdown phenomena and the cetermination of potential containment builc-ing failure modes. It is important that this aspect of our review reflects best estimate calculations and removes any " conservative' assumptions that may have been made by GE in this area. The first.approacn (or base case calculations) will then use fission proouct release from the fuel and transport in containment as prescribed in the RSS. Specifically, the core release fractions in Table 4.' will be used (note that we also included the core release fractions used in tne GESSAR-Il PRA in Taole 4.1). In addition, suppression pool decontamination factors of 100 and I will be used for subcooled and saturated pools, respectively. No primary system hold-up will be allowed and the CORRAL [93 code will be used to mooel tne transport of fps in containment. The second approaca (or advanced technology case) will reflect the new methods currently under development [83 by RES/NRC and their contractors. This will necessitate BNL staff following closely the activities of RES/NRC. In addi-tion, the various computer codes and models under development [83 will also have to be transferred to BNL to allow for the successful completion of this approach to our review of the GESSAR-II PRA. 4-2
This letter repo-% therefore initially concentrates on reviewing anc cleeMy understanding the methods used by GE to detemine the source tems used in the PRA. This is done in Section 4.1. We then compare, in Sections 4.2 and 4.3, the basic mathematical model (the CORRAL [93 code) used by BNL and GE to cal-culate fission product transport. In Section 4.4 we calculate how the source terms detemined by GE change if the base case approach is used. Finally in Section 4.5, we provide sensitivity studies to show how some of the new meth-ods (advanced technology approach) under development might influence the cal-culated source tems. However, it must be emphasized that the inclusion of Section 4.4 in no way endorses the new methods used by GE in the PRA. The sensitivity studies are being used as part of our review only to develop areas of concern that we con-sider might be potentially important in tems of health consequences. Some of these concerns have already been expressed [10] as 02s anc await GE respon-ses. Others will require the results of work underway by RES/NRC contractors before they can be resolved and factored into our review. 4.1 R[eleaseCategoriesUsedintheGESSAR-IlPRA In the GESSAR-II PRA, fifteen sets of release categories were calculated to represent the failure modes identified for the various accident sequences. We have reproduced these release categories in Table 4.1. In Section 1, we noted that this letter report concentrates on Class IT sequences so that only re-lease categories 1 through 6 in Table 4.1 are under review in this section. The methods used to calculate the release categories are described in Appendix F.3; however, the failure modes and sequence of events leading up to the re-leases are described in Appendices C.16 and C.17 of the PRA. It is rather an involved process tracking through the various sections and appendices of tne PRA to identify a particular set of release fractions in Taole 4.1 with the appropriate accident sequence, event tree and failure mode. Hence, we include below a brief description of the six release categories of interest. Release Category Number 1 (I-T-L3) Most of the probability (~70%) of this release category comes from the CTI-Pa containment event tree (refer to Section 2). Note that the CTI-Pa tree repre-sents Class 1 transients initiated by LOOP in which power is restored during 4-3
~ / the 60 minutes following the start of core damage. GE considers that for these sequences 30% of the time there will be local H2 combustion in the wetwell. GE further concluded that 60% of the time the containment would re-main in' tact during this local H2 combustion event. Hence, the containment is pre'dicted to fail approximately 12 hours after scram due to the build-up of noncondensible gases during core / concrete interactions (refer to Section 3.2.1). This release category, therefore, represents the gradual overpressur-ization failure mode described in Section 2. There is no bypass of the sup-pression pool assumed for this failure mode so that all of the fission pro-ducts are assumed scrubbed by the pool. Sensitivity studies have shown that, even if the high pool decontamination factors assumed by GE are reduced, aero-sol deposition in the wetwell (during the 8 to 9 hours after RPV f ailure that it takes to fail containment) significantly reduces the fission products released. Release Ca'tecory Number. 2 (I-T-E2) Virtually all of this release category probability comes from the CTI-Pa con-tainment event tree. GE concluded that for these sequences, 10% of the time there woulc be a global detonation that would result in containment failure j and cracks in the drywell head and in the drywell ceiling / roof slab. These. cracks would allow the upper pool to drain into the drywell within about one hour. GE assumes that the water in the drywell would prevent the vaporization release from occurring. For this sequence, GE assumed that all of the fission products released during the in-vessel release phases would pass through the suppression pool. However, GE also assumed that fission products released f rom the primary system as the vessel fails could potentially bypass the sup-pression pool through the cracks generated by the global detonation.
- However, NUREG-0772[73)
GE calculates (based on tnat most of the fission products will be released from the fuel prior to vessel failure anc will therefore be subjected to pool scrubbing. Tne GE core release fractions based on the methods in Reference 7 are included in Table 4.1 and are compared with the core release fractions used in the RSS.[6] The timing of fission product release relative-to the time of loss-of-drywell-integrity is of importance and is explorec in greater detail in Section 4.2. 4-4
o ~ Release Category Neber 3 (?-i-E3' Again, all' of this release category probability comes from the CTI-Pa tree. This category is composed of sequences that result in global H2 combustion or -local H2. detonation. This category represents about 65% of -the total probability of the CTI-Pa tree. GE assumes that the H2 phenomena result in early loss of containment integrity but that the drywell wall remains intact. Thus, all of the fission products are again subjected to suppression pool , scrubbing. In Section 3 it was noted that-our MARCH analysis indicates a much -greater potential for loial detonations to occur than assumed in the PRA.. _How-ever, it is not (at this stage) clear that local detonations will be _quite as benign as suggested by the GE analysis of this release category. Release Category Number 4 (I-T-12) All of. the probability of this rele'ase category comes from the CTI DD con-tainment event tree (refer to Section 2). Note that the CTI-Pb tree repre- ~ sents Class 1 transients initiated by LOOP in which power is restored after 60 minutes following the start of core damage. Fo'r these sequences, GE concluded that approximately 30". of the time a global detonation would occur resulting 'in containment failure and loss of drywell integrity. This category is simi-lar to Category 2 in that fission products released from the primary system at vessel failure could potentially bypass the pool. The release fractions are, however, also relatively minor because GE again believes that most of the fis-sion products are released prior to vessel failure and are hence subjected to pool scrubbing. Release Category Number 5 (I-T-I3) All of this release category comes from the CTI-Pb tree and is composed of H ' combustion (local H2 combustion contributes only sequences with global 2 about 10*.). These sequences fail containment early but leave the drywell wall intact. Thus, all of the fission products are subjected to suppression pool scrubbing. Release (6tegory Number 6 (I-T-L2) This release category is predicted by GE to be of much lower probability (ap-proximately two orders of magnitude) than the other five Class 1 categories. Containment failure occurs via gradual overpressurization ( Y containment 4-5
failure moce) curing extens1ve core / concrete Interactions. However, for these sequences, it is assumed that local H2 combustion.f ails penetrations between the drywell and the wetwell resulting in a potential bypass of the suppression 8 pool. The local H2 combustion was assumed not to fail containment. This category again has the potential for pool bypass but the 'GE calculated release fractions again reflect the assumption that most of the fission products will be released in-vessel. However, even with a different timing of release, the long time to fail containment for this sequence would allow significant aero-sol deposition. From the above discussion, one might conclude that release categories 2 and 4, which include the potential for fission product bypass of the suppression pool, might be rather more severe than the other four Class I categories. An inspection of Table 4.2 will indicate that although Category 2 is the most se-vere of the Class I categories, it is only marginally more severe. In fact, all of the Class I release categories, as calculated by GE, are remarkaoly similar inspite of the fact that they represent a very wide range of f ailure modes. The similarity is because GE assumes that most of the fission prooucts are released from the fuel while it is in the reactor vessel. If these in-vessel fission products are not retained in the prima ry system, they are subjected by GE to very high suppression pool decontamination factors. The rather modest amount of fission products released ex-vessel can, in some se-quences, bypass the pool but are subjected to aerosol ceposition and other removal mechanisms. The above release categories, as calculatec ey GE, rely on the CORRAL [9] computer code. A version of the CORRAL cooe is also in use at BNL and is usec as our basic mathematical model for calculating the quantity of fission pro-ducts released from containment. In the following section we describe modi-I (ficationsmadetoCORRALatBNL. 4.2 BNL CORRAL Modifications The CORRAL /KORALIN package has undergone some modifications at BNL. Pre-viously, the KORALIN code required two separate runs, the first of which listed the data output by MARCH onto the TAPE 7 file, and the second which wrote the file that inputs into CORRAL. The standard version requires careful hanc selection of a maximum of 20 data points for each parameter. These 20 4-6
points must characterize eacn parameter over very long time intervals and over a number of time steps ranging from a few hundred to several thousands. It was desired to totally eliminate the manual selection of data points from the TAPE 7 file and to increase the number of data points used in the CORRAL code in order.to more accurately define the input parameters. In addition, it was thought desirable to assure that for the most critical parameters, namely the material flow rates, the integrated total mass flow between compartments and to the outside should be transmitted accurately from the MARCH output to the CORRAL input. A scheme was, therefore, devised whereby the number of points transmitted through to CORRAL was increased from 20 to 500. For individual cell conditions such as temperature pressure, humidity, etc., the KORALIN code merely selects every Nth point on the MARCH output file such that the 500 se-lected points are evenly distributed through the original data sample. For the flow rates, however, an averaging routine is used so that all the original data points are factored into calculating the 500 input flow rates such that the area under the flow rate versus time curve is accurately preserved. The capacity of the KORALIN program was expanceo to take as many as 9600 original data points fran its previous 1300 points. Hence, except for those input parameters that are not obtained from MARCH, the interface between MARCH coce output and CORRAL code input is now automatic and more accurate at BNL.
- 4. 3 Comoarison Between GE and BNL Versions of CORRAL Code The CORRAL input deck for the I-T-L3 sequence was run, as received [Il3 from GE, on the BNL computing system and the results comDared with those outlined in 15.D.3 (Table.F.4.-2) of the PRA.
Table 4.3 shows the results of the com-parison. The noble gases and halogens agree exactly, wnile the aerosols ap-pear to differ by a f actor of.about 1.7, with the BNL CORRAL code predicting the higher releases. These differences may be due to differences in the ver-sions of CORRAL used by GE and BNL. 4.4 Impact of RSS Source Term Methodolooy (Base Case Acoroach) In this section the I-T-E2 accident sequence is reanalyzed using the RSS re-lease fractions and scrubbing factors. A two-volume CORRAL representations was used in this analysis'. The two volumes were representations of the 4-7
4 wetwell an-ine drywell. The gap and melt releases were multiplied by a f ac-tor of 10-2 to simulate a pool decontamination factor of 100 and then di-rected into the wetwell. This decontamination factor was applied to the iodine (!), cesium-rubidium (Cs-Tb), telluriun (Te), barium-strontium (Ba-Sr), ruthenium (Ru) and lanthanum (La) fission product groups only. The noble - gases (Xe-Kr) and organ'ic iodine (01) were not affected. With no oxidation release the undiminished vaporization releases were deposited in the drywell (over a 2-hour period) from which they were partially vented through the by- ~ pass flow path into the wetwell and subsequently out through the containment break. The fraction of fission products released from the containment for this case appears in Table 4.4 and is compared with the release fractions cal-culated by GE for this category. Note GE assumes for this sequence that the vaporization release will not occur (refer to Section 4.1) because the upper pool drains into the drywell. In our analysis,we assume a failure in the dry-well wall and no draining of the upper pool into the drywell, hence we allow for a full vaporization release. We are at present calculating a 1-T-E2 re-lease assuming that.the upper pool does drain into the drywell. This will be included in the final report. An inspection of Table 4.4 will indicate that in all cases the GE calculated release fractions are lower than those detennined by BNL. These differences are because the BNL calculations are based on RSS core release fractions, where as the GE results are based on the methods in Reference 7 regarding fis-sion product release. The effect that these different approaches have on the individual releases is discussed below. Noble Gases Both approaches handle the release of noble gases (Xe-Kr) in a similar manner and this is reflected in the results. lodine In the GE model, all the iodine (I) is released during the melt and gap re-lease periods and is thus subject to decontamination by suppression pool scrubbing. Thus, organic iodine (01) is seen to be the major contributor to the total iodine release from the reactor pressure vessel (RPV). In the case l of the BNL calculation using the RSS release fractions, approximately 9.9", of f I 48
r ?. ' the iodine is released during the vaporization release period anc is t9us not subject to pool scrubbing in this sequence. Furthermore, the f raction of io-dine released as organic iodine in the RSS approach is approximately 23 tines as large as in the GE approach based on Reference 7 methods. Cesium-Rubidium Group In the RSS approach, approximately 19% of the cesium-rubidium (Cs-Rb) group is released during the ex-RPV vaporization period, which is not subject to pool scrubbing. The GE approach releases all the Cs-Rb group during the gap and melt release, subjecting it to pool scrubbing. Tellurium Group The large difference between the BNL ano GE release fractions for the tellu-rium (Te) group is primarily due to the fact that the largest fraction of the group (85%) is released during tne vaporization release period in the RSS method, while it is all released during the melt release in the GE approach. Thus, in the RSS approach 85% is not subject to pool scrubbing, wnile in the GE approach 100% is scrubbed. Barium-Strontium Group Most of the barium-strontium (Ba-Sr) group is released during the melt periods in both approaches, with even more of the core inventory being released in the GE study. Thus, differences between the two approaches are due to different assumptions regarding pool decontamination factors. Ruthenium Group In t.he case of the ruthenium group, the GE stuoy assumes that approximately half is released during the melt and in-RPV vaporization phase and the re-mainder during the ex-RPV vaporization (which is assumed not to occur for these sequences by GE), while in the RSS approach, approximately 60% is re-leased in the ex-RPV vaporization period. Furthermore, the total f raction of core inventory available for release is slightly higher in the RSS method compared to the GE method (.08 vs. 074). Thus, the difference in the fission product release are due to different decontamination factors applied to the melt release and no decontamination applied to the vaporization release in the BNL approach. 4-9
r c Lanthanum Grovo ' Finally, in the case of the lanthanum (La) group, the RSS nethod and the GE method release essentially all of the fission products during the vaporization period. In the RSS approach, approximately twice as much of the original core inventory as in the GE approach is released during the vaporization period. The difference in final release fraction is because there is no ex-vessel va-porization release in the GE calculation and due to the f act that in the BNL model there is no pool scrubbing. The pool decontamination f actors,' which were applied to iocine and aerosols only, were taken to be 100 in the BNL study and 10000 in the GE study. These are the values used for subcooled pools in the RSS and the GE study, respec-tively. No credit for decontamination during the vaporization release period was taken in the BNL case. 4.5 Sensitivity Studies In this section we assess the impact of the changes made by GE relative to the RSS methodology. In Section 4.5.1 the impact of calculating the release of fission products based on NUREG-0772 release rate constants is assessed. The impact of fission product hold-up and re-emission from the primary system is calculated in Section 4.5.2. Finally, in Section 4.5.3, we describe our in-put to the assessment of the suppression pool decontamination f actors used in the GESSAR-II PRA. 4.5.1 Fission Product Release Rate Constants Based on NUREG-0772 The release of fission products based on tne NUREG-0772 rate constants is con-sistent with the direction of RES/NRC activities [7,83 in tnis area. At BNL we nave been modifying MARCH to calculate directly f rom the predicted tempera-ture history of the core the release of fission products from the fuel. This worn is currently being pursued by W. S. Yu[123 at BNL. Preliminary calcu-lations ' indicate that the release of noble gas and volatiles fission products are controlled by temperature rather than time at temperature, while tne re-lease of low volatility fission products is controlled by time at temperature. In general, these results indicate that the BNL-calculated release fractions for melt and gap release are lower than those detennined by GE for transient events. For noole gases, iodine and cesium, this difference is approximately 4-10
r-3%, while for tellurto, barium and ruthenium, the difference is approximately 15%. The. difference for lanthanurn is larger (factor +2) and depends on which release rate coefficients are used. In the current calculations, coefficients corresponding to uranium are used. l l However, at this stage of our review we assess the impact of the NUREG-0772 rate constants by using the GE core fission product release fractions for L transient sequences (Class I-T). This data includes release fractions for the t i gap, melt and ex-RPV vaporization, and an additional in-RPV vaporization. In l this study the in-RPV vaporization will be added to the melt release, since there is no other way of handling this release in the BNL CORRAL code. The results of' this calculation are shown in Table 4.5 and compared with the re-l' lease fractions calculated for this category using the RSS prescription (refer i to Section 4.4 above). It is seen that the noble gases are essentially unaffected. Furthermore, the release fractions for cesium, ruthenium, and lanthanum groups cased on the RSS core release fractions are approximately two times larger than those based on t I the GE calculated fractions. This is partially due to the _ larger releases { during the melt release phase for cesium, and ruthenium in the GE based calcu-l lation, which implies more pool scrubbing in this case. In the lanthanum group, the difference is due to the fact that approximately.005 of the fis-sion - product inventory is available for release in the GE calculation and twice as much is available in the BNL calculation. Organic iodine differs by approximately a factor of 22 and this is due to the fact that, in the GE based calculation, the frac-ion of iodine available for release as organic iodine is reduced by approximately this amount. The larger difference in the case of the tellurium group is due to the fact that, in the RSS based calculation, 85% l is released in the vaporization period and thus not subject to pool scrubbing. j. Because of the large unscrubbed vaporization release in this case, this dif-ference could potentially be larger. However, since the release takes place in the drywell and then has to pass through to the wetwell before passing out into the environment, the difference is narrowed. In each volume it is subject to agglomeration and settling. In the GE based calculation,100% is released during gap and melt release periods and subject to pool scrubbing.
- Finally, j-for the barium-strontium group, a larger fraction of the core inventory is I
L l 4 11 i {
o released in the GE approach than in the SN; approaa.. Since this is released primarily during the melt period in both cases, it is subject to the same de-gree of scrubbing. Thus, approximately 2.3 times the RSS-based release frac-tion is computed in the GE based calculation. In the case of the iodine release, app'roximately 9.97, is released during the vaporization release period, which is subject to removal mechanisms such as agglomeration and settling and surface plateout, in the drywell and wetwell. The efficiency of these processes, in this particular sequence, is high since the flow out cf the building during this time period is low, allowing enough time for the processes to take place. Thus, the net leakage to the environ-ment of iodine is very similar for the two methods. 4.5.2 Primary System Hold-Up The effect of possible hold-up and the re-emission of fission products was also investigated. To carry out this analysis more mechanistically would re-quire a liARCH-MERGE-TRAP-MELT analysis.[83 This was not done here. Instead a fraction of the fission products was retained in the RPV. This fraction was obtained from an elaborate analysis [8] of a TQUV seouence presented at the Third Source Tem meeting. Thus, the fraction of the in-RPV release was mul-tiplied by a factor to represent the hold-up of fission products in the RPV due to plate-out. The remainder of the fission products was subjected ;o pool scrubbing and released to the wetwell. Fission products which are retained in the RPV represent a heat source on the inner surfaces of the RPV and thus start to heat these surfaces. The more volatile species would be expected to be re-evaporated and plate-out on a cooler surface. This reduces the heat source at the original site, and increases it at the new plate-out site. This moving heat source has the overall effect of moving the more volatile fission products to cooler region and possibly out of the RPV. The approach outlined above ignores surface chemistry effects and the other chemical compounds fomed by fission products. However, it serves as a basis for making esti-mates regarding re-emission of fission products. The following assu 1ptions were made regarding primary system hold-up and re-emission; The following retention factors will be used: e Noble gases, organic iodine, and elemental iodine will not be subject to retention. 4-12
o 1 o The cesium-rubidium group will be subject to 407, retention. The tellurium group will be subject to 65% retentipn. e All rema4ning fission product groups (barium-strontium, ruthenium, lantha- .o num) will be subject to 60% retention. ~ Re-emission of the fission products assumed retained in the primary system will be dealt with as.follows: e 10% and 100% re-emission of the Cs-Rb and the Te groups will be assumed. Nore-emissionoftheBa-Sr,RuorLa.groupsbillbeconsidered. e / The fraction of fission products re-emitted are added to the ex-RPV vaporiza-tion release, and are thus not subject to decontamination by pool scrubbing in this sequence. Results' for this analysis are shown -or. Taole 4.6. It is seen that the first three fission product groups (Xe-Kr, 01, and I) are unchanged from the case with no hold-up. The last three groups (Ba-Sr, Ru and La) are either reduced or uncha'ngeo. when compared to ine no hold-up case. The change is proportional to the in-RPV release fraction which is the largest for Ba-Sr and the smallest for La. The Cs-Rb ant Te groups, which were sucject to hold-up and re-emission of 10% and 100% both show increases over the case with no hold-up. Since the Cs-Rb, Te and Ba-Sr groups release most of their fission products during the in-RPV period (GE based release fractions), they would be expected to behave in a 'similar manner if no re-emission were allowed. Thus, it would be expected that with no re-emission of fission products the release fractions for Cs-Rb and Te groeps would be below those corresponding to the no hold-up case, as is the case for the Ba-Sr. Frcm Table 4.6 it can be seen that for even a modest amount of re,-emission (10%) for this particular sequence, the release fractions are already very close to those ca.ses with no hold-up. This sensitivity points out the need for an accurate re-emission determlination of fission products. The consequences of these increased releases will be dis- ~ ' ~ cussed in Section 5. '4.5.3 Suppression Pool Decontamination Factors A detailed review of the pool DFs claimed by GE is the responsibility of the AEB/DSI/NRC and their contractors. BNL will provide to AEB sequence dependent 4-13
a gas and vapor flow rates to the suppression pool during the various stages C core meltdown and attempt to define a range of appropriate aerosol particle sizes. AEB will then calculate appropriate pool DFs for application to the GESSAR-II PRA. These DFs will be included in the final report. 4.6 CORRAL Analysis Used as Input to CRAC For the purposes of comparing GE and BNL versions of the CRAC code, the re-lease category corresponding to sequence I-T-E2 (refer to Table 4.4) were used in Section 5.1.- All of the release categories calculated in Section 4.5 (re-fer to Tables 4.5 and 4.6) were used as input to the CRAC analyses in Section 5.2. The results in Section 5.2 give a measure of importance, in terms of health consequences, to the sensitivity studies in Section 4.5. 4.7 References for Section 4 1) BNL Letter Report to J. Meyer (RSB/DSI/NRC) from W. T. Pratt (BNL), dated August 20, 1982. 2) BNL Letter Report to J. M. eyer (RSB/DSI/NRC) from W. T. Pratt (BNL), dated October 27, 1982. 3) "Probabilistic Risk Assessment of the Limerick Generating Station," Philadelphia Electric Company, Septemoer 1982. 4) " Zion Probabilistic Safety Study," Commonwealth Edison Company, September 1981. 5) " Indian Point 2 and 3 Probabilistic Safety Study," Power Authority of :ne State of New York and Consolidated Edison Company of New York, Inc., March 1982. 6) Reactor Safety Study, "An Assessment of Accident Risks in U. S. Commer-cial Nuclear Power Plants," WASH-1400, HUREG/75-014, October 1975. 7) " Technical Bases for Estimating Fission Product Behavior During LUR Acci-dents," USNRC Report NUREG-0772, June 1981. 8) "Radionuclide Release Under Specific LWR Accident Conditions," Draft BMI-2104 Report, 1983. 9) R. J. Burian and P. Cybulskis, " CORRAL 2 User's Manual," Battelle Colun-bus Laboratories, Columbus, Ohio, January 1977. 4-14 t
c_ 10) BNL Letter Peport to J. Meyer (RSS/D51/NRC) fror.. ett (BNl,) Cated May 18, 1983. 11) D. A. Hankins, " Transmittal of MARCH and CORRAL Input Files to the NRC," August 25, 1982. 12) W. S. Yu, " Fission. Product Release Rates - Application.to GESSAR-!! PRA Review " BNL memorandum to W. T. Pratt, dated May 23, 1983. ) e h s \\ 4 4-15
Table 4.1 Fraction of core inventory released Nuclide Group Gap Nelt Yaporization WASH-1400 GE WASil-1400 G,0
- WASH-1400 GC Xe-Kr 3.0(-2)**
8.7(-1) 1.0(-1) 01 1.2(-4) 6. 1 11 ( - 3 ) 7.0(-4) I-ilr 1. 6118 ( - 2 )
- 8. 76fl(-1 )
9.93(-2) ? Cs-Rh 5.0(-2) 7.6(-1) 1.9(-1) Te 1.0(-4) 1.5(-1) ) 8.499(-1) Ba-Sr 1.0(-6) 1.0(-l) 1.0(-2) Ru 0.0 3.0(-2) 5.0(-2) La 0.0 3.0(-3) 1.0(-2) NOTE: GE calculations based on methods in Reference 7 s
- 3.0(-2) = 3.0x10-2
.o Table 4.2 GE fission product releases and release characteristi: 2 4 i acciotar etassiricarle= istC sectiew 4.4 or main acPonti ace rese
- acelocar Peasanititv. Evewtiva flat to Ptunt atttast. MR ith Oueaflom et ACLtast, at Oct Evacuaften waamike flPC. we tvf mtat OF *LLet. Cab /ste Mip
- Clow? DF PLupt attgast m Mt 98 e
- AA010150f 0Plc AcLEASE '# ACTION tv CMCMiCaL'OkowP WLPeta 4-17
Table 4.3 Comparison of release fractions as predicteil by GC anil llNl. versions of the COltitAl code for the I-i-l.3 category 4 Fission Proddct Groups ~ Cases Xe-Kr 01-I Cs-ith Te na-Sr Ru la P. 'o1 liased on 1.0 3.UE-4 1.8E-4
- 1. llE-6 1.8E-6 9.0E-6 1.5E-6 IINL version of C0llRAL
- \\
Table 4.4 Comparison between GE seituence 1-T-l'2 release;. and ItNt. release liaseil on 1t55 metlioitoloijy Fission I'ro<fuct Groups Cases Xe-Kr 01 1 Cs-lib Te lla-Sr Ru la ? G; BNL ilased on .9916 6.94(-3) 5.24(-3) 2.0S(-2) 6.39(-2) 1.56(-3) 3.93(-3) 7.62(-4) RSS Mettiods
Table 4.5 Effect of core release fractions based on the RSS compared with GL calculated fractions. Fission Product Groups Cases Xe-Kr 01 I Cs-Rb Te Ha-Sr Ru La BNL Based on .9916 6.94(-3) 5.24(-3) 2.0S(-2) 6.39(-2) 1.56(-3) 3.93(-3) 7.62(-4) RSS Release Model ,3 l 'As Above but .9980 3.00( 4) 5.83(-3) 8.27(-3) 8.28(-3) 3.60(-3) 2.58(-3) 3.69( 4) 4 with GE As-( sumptions N Regarding FP Release from llU / l
~ Table 4.6 Comparison of GE hased release fractions with and without" selective hold-up and re-emission ll Fission Product Groups Cases Xe-Kr 01 I Cs-Rh Te lla-S r Ru La l / Based on GE .9980 3.00(-4)* 5.83(-3) 8.27(-3) 8.23(-3) 3.60(-3)' 2.58(-3) 3.69(-4) ,1 Core Release 3. i Fractions but with No lloid-up in Primary System c. bl 1 ! (i Release . 99f50 3.00(-4) 5.83(-3) 7.91(-3) 7.69(-3) 1.66(-3) 2.36(-3) 3.69( 4) / Fractions with floid-up and then 10% Release ** a GE Release .9980 3.00(-4) 5.113(-3) 3.44(-2) 5.0l!(-2) 1.66(-3) 2.36(-3) 3.69( 4) fractions with lloid-up and then 100% Releast *
- i
- 2.57(-4) = 2.5/ x 10-4
- 10 and 100% release refers to the eventual re-emission of the Cs-Rb and Te fission product groups initially assumed held-up in the primary system.
~ 5. RISK ANALfSIS An in-depth review of the consequence (site) model used in the GESSAR-!! PRA is the responsibility of the Accident Evaluation Branch (AEB) DSI/NRC. The i determination of. risk in this section is for the purpose of assigning a mea-sure of importance (in tems' of risk to the public) to the various accident sequences and containment failure modes. These calculations are based on the GE site model and do not yet reflect the AEB. review. However, these calcula-tions provide a useful way of ranking the relative importance of our concerns regarding the containment event trees (Section 2) core meltdown modeling (Sec-tion 3) and fission product releases (Section 4). The consequence model used by GE-is described in the following sections of the PRA; Section 5: Magnitude of Radioactive Release Appendix F.4: Consequence Analysis In addition, further infomation on the risk analysis was provided at the Third [13 GE-NRC Technical Update Meeting. The GE response to question 720.73 also provided further justification for the use of site 6 as an average site appropriate for use in the consecuence analysis in the PRA. A final res-olution of the acceptability of this site depends on the AEB review. In this section the risks of the various sensitivity analyses described in the preceeding subsections will be discussed. The determination of the risks was carried out using the CRAC code operational on the BNL computer system. A CRAC input deck for the GESSAR-II 238 Nuclear Island was based on information supplied by GE (in the GESSAR-II PRA) and was obtained from data available on standard NRC files. The GE supplied data covered the population distribution, core inventory, and plant dimension data. The health effects data and the me-teorological data were obtained from standard NRC files. The latter data re-ferred to standard site number 6. Because the evacuation model used in the version of CRAC operational at BNL is different from the evacuation nodel used -in the version of CRAC in use at GE, two models were used to estimate the sen- .sitivity of the consequences to this input. Finally, the leakage data used in 4 these analyses varied from run to run, and was characteristic of tne scenario considered. 5-1
anc evacuation model corresponcing to Scheme 3 (as outlined above in Section 5.1) were used in all the calculations presented in this section. Fission product release fractions and timing of the release are shown in Table 5.4 It is seen that the time of release is the same for all sequences since up to that point the sequences are identical. This time corresponds to the time of containment fail'ure. The duration is chosen to proceed up to the end of the vaporization release. The warning time is defined as the difference between containment failure and the start of core melting. Since the duration of release. is over 4.5 hours, we have assumed that the energy of release will be very low. The height of release was assumed to be 10 m in all cases. It can be seen from these results that using the release fractions based on RSS[33 methodology leads to the most severe risks, for this particular se-quence, for three of the five risk indices. In the case where the GE basec release fractions (which are based on new methods) are used, much more benign risks result. Including hold-up and subsequent re-emission of selected fis-sion product groups results in risks which increase in severity with in-creasing re-emission. The risk resulting from' the case with 100% re-emission is approximately as severe as the case using release fractions based on tne RSS methodology, while the risk resulting from the case with 10% re-emission is approximately as severe as the case using release fractions, based on GE methodol ogy.
- 5. 3 References for Section 5
- 1) BNL Letter Report to J. Meyer (RSB/DSI/NRC) from W. T. Pratt (ENL), dated August 20, 1982.
2)
- 1. A. Papazoglou, et al.,
"A Review of the Limerick Generating Station Probabilistic Risk Assessment," NUREG/CR-3028, BNL-NUREG-51600, February 1983.
- 3) Reactor Safety Study, "An Assessment of Accident Risks in U. 5. Commercial Nuclear Power Plants," WASH-1400, NUREG/75-014, October 1975.
5-3 n
-1 Table 5.1 Comparison between GE anc Schene 1 evacuation mooels Scheme 1 GE-Model Maximum Distance of Evacuation (m) 1.61(+4) 1.604(+4) Evacuation velocity (m/s) 5.66(-1) 5.66(-1) Time Lag Before Evacuation (hrs) 0.0 0.0 Radius of Circular Area Near Reactor 8.0(+3)
- 8. 0(+ 3)
Angle of Evacu'ation -- 45.0 45.0 Evacuation Cost-(S/ Evacuee / day) 95.0 95.0 -Criteria:of Duration of Release 3.0 3.0 for Evacuation Distance Movec by Evacuees '(m) 2.42(+4) Sheltering Racius (in) 1.61(-4) Evacuation Scheme 1.0 Cloud Shielding Stationary People .75 .75 Cloud Shielding Moving Evacuees 1.0 1.0 Cloud Shielcing Sheltering .75 Cloud Shielding tio Emergency Action .75 _ Ground Shielding Stationa ry Evacuees .33 .33 Ground Shielding Moving Evacuees .50 .50 Ground Shielding No Emergency Action .33 Breatning Rate Stationary Evacuees 2.66(-4) -2.66(-4) Breathing Rate Moving Evacuees 2.66(-4) Breathing Rate Sheltering Region One 2.66(-4) Breathing Rate Sheltering Region Two 2.66(-4) 5-4
Table 5.2 Comparison between Scheme 1 and Scheme 3 evacuation models' Scheme 1 Scheme 3 Maximum Distance - of Evacuation (m) 1.61(+4) 1.61(4) Evacuation Yelocity (m/s)- 5.66(-1) 6.70(-1) Timk lag Before Evacuation (hrs) 0.0 2.0 . Radius of Circular Area Near Reactor 8.0(+3) 8.0(+3) Angle of Evacuation 45.0-90.0 . Evacuation Cost (S/ Evacuee / Day) 95.0 245.0 Criteria of. Duration of Release 3.0 3.0 for Evacuation Distance Moved by Evacuees (m) 2.42(+4) 2.42(+4) 4 ~ Sheltering Radius (in) 1.61(+4) 1.61(+4) c Evacuation. Scheme 1.0 3.0 Cloud Shielding Stationary Peopie .75 .75 Cloud Shielding Moving Evacuees 1.0 1.0 Clouo Shielding Sheltering .75 .75 Cloud ' Shielding No Emergency Action .75 .75 Ground Shielding Stationary Evacuees .33 .33 Ground Shielding Moving Evacuees .50 .5 Ground Shielding No Emergency Action .33 .33 Breathing Rate Stationary Evacuees 2.66(-4) 2.66(-4) Breathing Rate Moving Evacuees 2.66(-4) Breathing Rate Sheltering Region One 2.66(-4) Breathing Rate Sheltering Region Two 2.66(-4) 5-5
r. s y Table 5.3 Comparison of GE and BNL CRAC code prec1ctions for sequence 1-T-E2 i BNL CRAC BNL-CRAC Consequence GE-CRAC (Scheme 1) (Scheme 3) Total Thyroid NA* 1.52(-6)** 1.53(-6) Latent Cancer 9.64(-7) 6.38(-6) 6.43(-6) (excluding thyroid) Early Fatalities Early Injuries NA 5.15(-8) Total Person Rem 1.61(-2) 1.16(-1) 1.17(-1)
- Not Available
- 1.52(-6) = 1.52x10-6 Note these calculations are preliminary and may be changed as a result of the AEE/DSI/NRC review.
5-6
Table 5.4 Release fractions developed in sensitivity studies for the I-T-E2 category GE Release frac-GE Release. Frac-WASH-1400 Release GE Release tions with floid-tions with Hold. Fractions No Fractions up and 10% up and 1001 lloid-up No lloid-up Re-emission Re-emission Xe-Kr .lil60 / .8707 .8707 .8707 /' 01 5.71(-3)* 2.57(-4) 2.57(-4) 2.57(-4) 7 i 6.01(-3) 3.80(-3) 3.80(-3) 3.80(-3) Cs-Rb 2.29(-2) 4.65(-3) 6.83(-3) 4.32(-2) Te 11. 6 7 ( - 2 ) 4.65(-3) 8.20(-3) 6.74(-2) Ba-Sr 1.48(-3) 2.32(-3) 1.23(-3) 1.23(-3) ui is Hu 5.20(-3) 3.24(-3) 3.12(-3) 3.12(-3) La 1.03(-3) 5.06(-4) 5.06(-4) 5.06(-4) Time of 1.146 1.146 1.146 1.146 Release 4 Duration 4.5 4.5 4.5 4.5 Warning .27 .27 .27 .27 Time q. s. Energy of 0.0 0.0 0.0 0.0 Release lleight 10.0 10.0 10.0 10.0
- 5.71(-3) = 5.71x10-3
Table 5.5 Impact of sensitivity stuelles for accitlent sequence'l-T-E2 Ihyroit! Latent I:arly Early Sequence 1)escription Cancers Cancers Fatalities injuries Person-Rem WASil-1400 Release Fractions 3.70(-5)* 2.111(-4) 2.94(-8) 1.69(-5) 4.01(0) No lloid-up ~ ! GE Release Fractions from 1.53(-6) 6.43(-6) 5.15(-8) 1.17(-1) PHA l '(' ,/ GE Release Fractions No. 1.48(-5) 1.01(-4) 5.25(-9) 7.62(-3) 1.88(0) Ilolti-up GE Release fractions lloid-1.41(-5) 9.53(-5) 4'. 39( -9) 7.35(-6) 1.76(0) up Plus 10% Re-emission l \\ GE Release Fractions lloid-3.55(-5) 2. 11 6 ( - 4 ) 2.12(-8) 1.39(-5) 5.36(0) l up Plus 100% Re-emission j -= 5
- 3.70(-5) = 3.70xlG Note these calculations are preliminary an<l may he changeil as a result of the AED/OSI/NRC review.
1 i bsidier.4
a 6.
SUMMARY
AND FURTHER WORK In this section we summarize the results of our review and indicate the direc-tion of further work. It was noted in Section 1 that this letter report con-centrates on our review of Class 1 transients, initiated by LOOP. Other lower probability sequences are currently being reviewed and will be discussed in the final report. In Section 2 we discuss our review of the containment event trees used by GE in the PRA. In particular, we have recalculated, based on the MARCH analyses discussed in Section 3, the conditional probabilities associated with the var-ious H2 phenomena identified in the PRA. We concluded that there is a much higher probability of local detonations occurring than assumed in the PRA. We .were unable to identify conditions under which a global detonation would be expected. However, these conclusiens are preliminary and based on simple cal-culations using MARCH. More detailed calculations are being made to provide an accurate representation of H2 concentrations in containment. In adci-tion, several problems were encountered when attempting to understand some of the branch point probabilities in the containment events. These concerns have been sent to GE in the form of Q2s and depending on the responses, will be ap-propriately addressed in the final report. In Section 3, we compared the predictions of the GE version of the MARCH code against the BNL version. Differences between the two codes were due to the use by GE of a more up-to-date decay heat correlation than used in the origi-nal version of MARCH 1.1. The new decay heat correlation includes actinide decay. The t1 ARCH 1.1 version in use at BNL has been updated to include acti-nide decay. The primary concern that emerges from the sensitivity studies carried out in Section 3 is the importance of hydrogen-related processes on both the acci-dent sequences and on the probabilities applied to the event tree analyses discussed in Section 2. In particular, a multi-volume MARCH model nas pro-duced evidence that hydrogen detonations localized in the wetwell adjacent to
- he drywell wall may occur with a significantly higher probability than as-sumed in the PRA as noted above.
In the PRA, it was determined that a global detonation was required before drywell integrity would be lost. A local deto-nation would apparently leave the wall intact. Our MARCH analysis would 6-1
suggest redistr;: sting the conditional probabilities (ref er to Section 2) in the PRA from global events (both detonations and combustion) to local deto-nations. If the structural analysis in Appendix C of the PRA is correct such a change would result in a lower probability of loss-of-drywell integrity than assumed i,n the PRA. Loss-of-drywell. integrity is important because it may supply a rechanism by 'which a bypass flow path could be established which circumvents the scrubbing capacity of the suppression pool. The impact of local detonations on drywell wall integrity is under review at NRC.
- Finally, our analyses in Section 3 also indicated that there is uncertainty in the metal-water reactioa modeling in the MARCH code.
It remains for us to assess the impact of the uncertainty reg,arding H2 generation. In Section 4 we reviewed in detail the methods used by GE to determine the fission product release fractions used in th'e GESSAR-II PRA. This review is restricted to those release categories associated with Cl as s 1 transients. The I-T-E2 and I-T-12 categories were identified as a having a relatively high probabilitly (~200 of tne total core nelt frequency) and potentially an impor-tant contribution to health consequences. A CORRAL input deck was obtained from GE and input to the BNL version of the CORRAL code. The predictions of the two codes were found, as expected, to be very close. The predictions agree exactly for the release of noble gases and halogens but differ (by less than a factor of two) for the release of aero-sols. The calculations used by GE co determine the release categories in the GESSAR-II PRA are based on methods that are a significant departure from those used in the RSS. At this stage of our review we are unable to confirm the validity of these new methods. Two approaches have therefore been used; firs.tly, we recalculate the potentially most important release categories using RSS core release fractions and suppression pool DFs (refer to Table 4.4 in Section 4) and secondly, we assess how the new methods influence the release of fission products. Table 4.4 indicates that the new methods, as used by GE, have a significant potential for reducing the release fractions. However, there are aspects of these new methods, not yet addressed in the GESSAR-II PRA, that could affect the reduction in release fractions indicated in Table 4.4. The sensitivity studies presented in Section 4.5 are included not as an endorsement of these new methods but rather to indicate that they do 6-2
not always predict a reduction in tne release f ractions rela'.ive to the RSS approach. An important result of the sensitivity study relates to hold-up of fission products-in the primary system. If this hold-up is permanent then the release fractions will be lower than predicted by the RSS methods, which ig-nored this affect. However, there is the possibility of re-emission of some of these fission products (namely, the Cs-Rb and Te groups) after f ailure of the reactor vessel. For sequences such as the I-T-E2 category, this is im-portant because at later times drywell integrity may be lost and Fps can po-tential.ly bypass the suppression pool scrubbing. The impact of this is clearly shown in Table 4.6 and indicates how important it is to accurately detennine the potential for re-emission of fission products from the primary system. In Section 5 we compared the predictions of the BNL version of the CRAC code with the GE version; this was done by using similar input to the two versions of the CRAC code and comparing the results. Tne BNL version of CRAC predicts 6.5 times higher mean latent fatalities than the GE version. Note that the review of the site consequence model in the GESSAR-l! pRA is the responsibil-ity of AES/DSI/NRC. The calculations in Section 5 do not reflect this review and are consequently preliminary. They are included only to give a prelimi-nary indication of how some of our concerns in Sections 2, 3, and 4 might in-fluence risk. The impact on risk of the sensitivity studies for the I-T-E2 category are clearly shown in Table 5.5 and it is again clear that the poten-tial for re-emission of fission products from the primary system after vessel failure is potentially an important issue. 6-3
APPENDIX A BWR Mark-Ill Suppression Pool By. Pass Model e e
BROOKHAVEN N ATION AL LABORATORY MEMORANDUM DATE: March 1, 1983 TO: SNL MARCH Users FROM: R. D. Casser M susJECT: BWR MARK III Suppression Pool Vent Modeling One of the MARCH code limitations is evident when attempting to model the dryvell-to-vetvell vents in a General Electric BWR with a MARK III contain-ment. For the ex-vessel phase of an accident sequence (or for dryvell LOCAs) steam, non-condensibles and fission products pass from the dryvell to the vet-well through 3 rows of circular vents which are accessed through a weir ar-range =ent. The pressure differential between the adjoining volu=es dictates the water level in the weir and thus the cross sectional flow area of the vents. The MARCH code, however, is limited to a single constant vent area (AVBRK). To mitigate this modeling limitation, a set of UPDATES to MARCH were developed which allow for the calculation of the vent flow area as a function of the static pressure differential. Figure I shows a rough view of the MARK III vent arrangement together with the appropriate dimensions. There are three sets of vents spaced at 8* increments circumverentially around the dryvell vall (45 vents per row for a total of 135 vents). Each vent has a diameter of 27.5 inches. For each row of vents the area which is exposed for intercompartment transfer is calculated by: z/7,72+R sin sR-2 Avgn7 - 45 + ~ where R is the radius of the vent and Z=Z -E n Here, H, is the level of water in the weir below the suppression pool water level, and Z is the mid-plane level of the nth row of vents. n The IVENT=-12 option in the MARCH code is utilized to implement this mod-eling. The input break area, AVERK, is assumed to be a break which bypasses the pool (although MACE continues to put it through the pool). The CONVENT subroutine then calculates a new total flow area which consists of the bypass A-1
Mero to 3h'L MARCH Usets ~ March 1, 1983 Fage 2 flow area plus the vent flow area (AVENT) based on op. If the flow is from the vetwell to the dryvell, AVEh'T is set to zero and flow occurs only through the bypass. Attached is a list of the updates for the vent modeling. They have been incorporated in MARCHBk'R which is a version of MARCH 1.1 that contains previous updates specific to Bk'R analysis. RC:tr Attachments G l l l l i t { l I I l i r A-2
1 = on x 9 =. u .,9 N N l -~ 5z //// / / / / /////// ,/ / ~ [ =, / } i l / / / //l l l/ / l ' l/ /l l'l /,/ l J I p = la' L/ / s /n, / / / / / / //
3 MARCH UPDATES FOR CE B*a7. MK III l 00 MARCH.2995 3 1 Av8R KTCVBRKWVBRRTVABRK7tySRS1 o0 MARCH.3095 3,. (B R *T CVS R KT*VS RMTvd3 R A,1 v 3 n S, 00 MARCH.5553 . sven00TINC C0NympTTF2T5TMTASVRTVO( i AVBRKTCVBRK,WVBRKTVABMK,1) o! MARCH 5554 7 Dai A R.C.Pi il.1 45c33,v.*313553,3.A*1u927/ DATA 21,22,Z3,Z4,25 /6.35'167,7.5,8.645833,10.85'167,12.0/ U-i-2o,47,4 3,41 4Aa.i*563a,15.a5-4ei.lL.5,Ai.6~4583 7< )- FU.N ( 8, B, X 1 = 4 5 0 * ( A * ( B *
- 2 ) /2. 0 + ( B + + 2 )
- A S I N ( X/B ) + X * ( B + + 2-x *
- 2 ) + + 0. 5 )
TD NANCn.536u IF(I.EO.1) GO TO 50 )
- VETT=v.v GO TO 700 av i-DFi;
) IF(Y.GT.Z1) GO TO 100 KVENTro.o GO TO 700 ) Auu ifLT.G.433 GC is cuu Z:y-Z2 AvEhT=FvM'LPi,R,4i GO To 700 ) 70 Ar(T.61.4+1 60 10 J0u AVENT=185.61156 ) Gv TG 100 300 IF(Y.GT.26) GO TO 400 Z s Y - Z3 AVENT= FUN (PI,R,2)+185.61156 Gv iu evu 400 IF(Y.GT 27) GO TO 500 ) avEnT=37A.22313 GO TO 700 -- 500 Ifti.0T.Zv; ww.- ca. ) Z=i-Z8 AVENT=r0N't P i
- R,4 i
- 3 i r. c 431 s GO TO 700 3 Muu A V EttT= 5 S6T834 7 0 700 WBAK=(AVBRK+AVENT)* AMIN 1(G1,G2)
) ) ) u
( APPENDIX B MARCH OUTPUT DATA 9
-4 w e MBE 5 35.55. 5 FWI N Jihe, BW1 m lWIgg, _ __ _ _ 1e pigyug g g,3 ^ l GESSAR TQUV (ITI) CASE GSR00 4 4=a 1 n 40000-M D: D 3600 0-i 1 E< i D: ga) 3000 0-D. My 25000-m m L D: 2000 0-o O W 15000-O h 10000-i i S000-i 00-j 00 10 0 20 0 30 0 400 60 0 00 0 70 0 80 0 90 0 1000 110 0 TIME - (MINUTE) i FIGURE 8-1 l
.... - -. - m. i GESSAR TQUV (ITI) CASE GSR00 1400 0 l 1200 0-D h 10000- \\ m D. y 800 0-W b m', M a000-O' 4 4003-E D. 200 3-I 00-o' 0 NO NO NO 100 0 110 0 00 10 0 NO N0 400 60 0 0 TIME - (MINUTE) FIGURE 8-2 I i
t .a. 4 g I l 1 O .e. O o .g oO o e -E CD 0 e -8 kJ CO< U G i- .o m F -E z 2 n, v v m i w ~E g ra E D 2 E O' es s E--- e U -8 to c CD E l-e o .g C i o _i -s i a eo e o 4 4 4 S E 8 2 a s a g r s = 13A313HD1XIM WY3JS - E31YR m 5 i B-3
,a y d' RBI 85 IG.M. 95 FWI N Jila., IgH JW-et ME st, p. _ -Ma pil1Put up g,3 J GESSAR TQUV (ITI) CASE GSR00 ~. 026 ^ o 020-tmJ b p E f O is - Q F 5 oc U 7 0 10 - g 5 ~ M ~ [a. 006- /
- f 0 00 00 10 0 20 0 30 0 400 60 0 60 0 70 0 80 0 90 0 100 0 130 0 TIME - (MINUTE)
FIGURE B-4 l l
's ~a i O 3 O O O -g CD 0 -8 F4 0 -R n o s 2 -E i F x = w w k -8 [ a ca C 2 SY -I " s a li l "E T (O O N "O i 0 C g -2 O = = { O 4 e w N O
== 0 O O O O E i 03rI3M 3803 NOIJDYH3 e 5 B-5
en e e R5E I 15.W.E FWIN M IW3 MINI 45, - - -e WI53rul MR
- 5. 3 GESSAR TQUV (ITI) GASE GSR00 8GD
~ W 700-b- 000 D. E-. b ~ sl )i m a a< 400-D. 2O U 300-Ahg a,0 10 0 - 00 200 0 4000 000 0 000 0 1000 0 1200 0 I400 0 1600 0 TIME - (MINUTE) ~ rIGilRE n-6 VOi,UME NO.1 j
e 9 1 4 EEI IO E= E N E N IM ME M G5, _ _ _ != we,M 6 5,3 GESSAR TQUV (ITI) GASE GSR00 i 7000 hp k 4' La e000-tea 'M E-. 1 4: h 5000-f a. ta E-4 4000-l cp Ea e Z . M (y 3 2 i k 3000-
- C O.
2 f O 2000-J ) j 100 0 00 200 0 4000 000 0 000 0 8000 0 1200 0 I4000 1000 0 } TlME - (MINUTE) FIGURE B-7 VOI.UME NO. t O 4
O o O N .8 2 = w OO 3 3 O O Q O E2 M e .$_n w O t n Z c:' r-os E .g - W v i 5 w D 2 O E cr - e-a g 1, 1 -8 VJ e W E2 l C O -a E ( a n E O a a a 4 i O O O O O O O O 5 l MY315 NOIDYHJ S:i1On B-8
9 e 4 MF IW.M.B NI N A, 4W1 Apetzcg, gemseev0s Oglyut vCa c.) i GESSAR TQUV (ITI) GASE GSR00 035 030-ZW 025-O> O20-Z S b b 015-in. W 0 10 - J O M 0 06-000-00 200 0 4000 0000 800 0 1000 0 12000 14000 1000 0 TIME - (MINUTE) l I FIGilRE B-9 VOLUME NO. I
Ptst a se.as.n rev a m. surs mesures. : - sisyun we s.: GESSAR TQUV (ITI).G ASE GSR00 020 1 Z [4 015-O M O i Z O 0 10 - b i toa o M i I** t Ed d J O 0 05-2 1 o i i 0 00 00 2000 4000 0000 8000 1000 0 3200 0 1400 0 1000 0 TIME - (MINUTE) FIGURE B-10 VOLUME NO.1 4
O N e O _8 Z o w O 3 e 5 y) e o 0 e CD o <O .$ w _ e O E E-eE E -g - m i E d 3 y y D e3 6 CV -l & a E-
- i e
l l _g Cn w E CD E W [ C e -R E_ di = a E e A 6 4 e e e 8 R S S S R R 2 9 .i 3HOSS3Hd 1N3K18YdMOD 1Y101 4 j B-11
o N d e 8 Z O 1 w O 2 D d M e 0 -l Ed CD <O -! G- ^ D ~ ~ z E F o 2 -w w v c:: 5 y W C i D 2 .l-CR' -g i: E >= .I i = -s Cn E CO 5_ [4 O 4 E i X 5 e a o e o a 6 4 4 6 3 R R R R E a = i 3801Yg3dM31 IN3Migydgo3 ~ s t l' ^ B-12
0 006 3 2 0 O 0 N .0 4 0 E 1 0 M R U. I S 0 O 00 V 'G 2 3 ES 0 A 0 .0) C 0E 1 T l ) i 3 I N 1 T I 00M n I 0( ( 8 tn u V r. E i U M r 0 Q 00T I 0 T R = A 0 S 0 ,0 4 SE = G = 0 ,002 n g 0 0 8 8 4 2 0 8 8 4 1 1 8 1 s 0 0 0 M 0 0 0 0 0 0 0 0 O 2<W$ zO lu. ]8M
s e N d o .8 Z o ca o 2 3 cn o o 0 e CO o <U .g_' -s 2 O z b
- s a
-g-i E W 5 E p 2 e l CY ~$ H s F 6 \\. =.s Cn ) ~ t in 8 e g { C / o -R E m ( m .e t g g g e e g g o o o o o e o Ji N30080AH NOIDYHJ 310K B-14
s i O N 6 o .8 w oo E e a C C y O C.3 (.D a <o .g g c E e E-oE a -g w i 5 w 2 E O e3 CY -$ F a E-- li _s CD ~ l D E EG [ C e -R E 8 R' R 2 2 8 8 e o o o o F .i N30AXO NOIDYUd 370N B-15
.@^ s I 6 I O E O h 6-En a 0 -8 M r En i -E E i U F i e c s a t-a,- w l 3 = Q2 C E O CY .g F i F 0 e -g in CD i M 4 0 = 1 -e I e C b b b b d b b L f E, I l l l 8 e 3H31YH3dW313E00 3DYH3AY E t B-16
O E O R B-W W O -g CC Cn o <O 'E E F= m z ~ p ex o-e v w E S E D i a _ g g l M l a (O .g CD -i Cd 5 0 e -e i i a a c C O O O O O e g E l I I I E 5 3HnSS38d N31SAS ABYMlHd E t B-17
e 6. s e 6 O E a O
- k b-CD o
0 -E ta CD n Q Z t O E 5 E-a ~ a U w I = S w i D 2 h - 8 I" E F ? = CD -2 CD -i [d i O t e-s I_ E 4 4 4 4 4 5 s a a e = s a 9 i 13A37 38n1XIX MY31S - 831YA Ee B-18
o t E o R g e V) - = O -E C:] CO< -E g U Fo m Z E F
- 2 a
, ~ - v w 1 s G w I D ob h 7 - g I-F s o CD -R in - E Cz2 i 0 O t -e E i O N 5 s g E O o o o O a k l g i 03DY38 GY'13 Notrygg E E t c 0 l l l B-19 i--
e._,r a 4 ..m.4 .a e -m 0 O E O h B- -W CO E O 00 C gE o C e~ o Z N i -b
- 2 gv w
-v cr O L3 c.: i D 2 i Cr -R F E F I -8 tn in .i W i f C t -9 i. E' .a 9 O 8 I e o o I 8 03r13W 3803 N0113YH,i E r l .B-20
~. sitsu e es.us.se ses s are, eEi1 m-ea mres, ese==*to sa m e c e.i CESSAR TQUV (ITI) CASE GSRr7 i 35 0 300-a 25 0-D' n. j PZ 20 0-i M A / 1 3 j IS O - I J A M O O 90 0 - .3 4 P O 80-00 i i i e i i 1 1 = 00 1000 200 0 300 0 400 0 6000 000 0 MO 8000 900 0 10000 1l000 12n00 + ] TIME - (MINUTE) rir.IIRE R-21 i 1 + -
2' 7 A 1 R S a G n E o A S x A ) C oET AU ) z N z IT .I -n 4 (M I( tn u o r. s i U su r ~ Q T i a T o A = R A o = = S A S E G g ~ 0 0 0 o o 0 0 0 0 n 0 0 0 0 5 0 s o 2 2 i l ~ mMo4:aw @,we EZIjl<D! a =',~
E_ g w D 2 e 3 y) e o 0 -E C=3 = CD -l O
- if s
-$ 2 m O z ~ i -$*5 l E-- c = C g i = 3 o fr D -{ g 2 i O P l F 8 F i en R en C:2 o
- 4 0
t i e i ( 4 t \\ o* e E 8 i I R R. 10 2 o o e a NY31S NOIDY8.i S310M l s t 9 B-23
0 002 x I 0 I 00 .O I 1 N 0 E 0 7 .0 M 1 0 R U. 1 I S 0 O 0 V G 09 E 0 0 S 00 A ) C 0E 0T 0 U ) 7 4 N I 2 T I 00M R I ( 0( 0 E R I I V 'p, 0 r. 0 E i U 0M F 6 Q T I T 0 0 0 4 R r A 0 0 S 0 3 S E 0 G 00 2 0 = 00 1 0 0 5 0 5 0 5 ~ 2 2 1 1 0 0 0 0 0 0 0 = m. $oOE>n Zo b:wE a. lfJ a e (t e r so n
s l I l o!
- E 6Z b
ca -l 2 D CD 'o e C .g Cx3 Cn "I <o .c -J $ m ~ z e b n . g 3. = v m E d i i D '$ w C i 3 i Cr i F j.E ' ~g e l O 1 y c' CO I C:] a a a It I i = -8 I I g 4 6 o E B E 8 B. O O i N3DAXO NOIDY83 3Io}s E t i s B-25 i
o N E ci z k -h N B CQ o O O -l Cz] CO -g <O e ta s ~$ D n e 7 ~ g-o 2 = -[ - v u i s G a i D -] $ C i e p t_. e 8 [i t o m .g m. i k o l C -E i-O j! -E i 1 ) C O e i g C C C O O O O O E R E E ? 8 2 E =i 380SS3Hd 1N3M18YdNO31Y101 E C B-26
3 O 9 9 0 002 3 0 2 00 .O I 1 N 0 E 0 7 '0 M 0 1 U 1 R L 0 O S 0 V 0 G '9 E 0 0 S '0 8 A ) C 0E 0T 0 '7 U 7 ) N I 2 I 0 T 0M I
- 0(
( 6 E R a I t V E G 0 e 0 l ed U =0M f 6 I a Q T h T 0 0 s =0 4 = R A L. 0 0 0 S 3 S a, E 0 s u s 0 e. G 02 s A 0 3 0 b 0 1 U* d 0 5 0 m 0 0 0 0 0 0 0 0 0 0 0 0 0 0 s 0 5 0 6 s. 2 1 1 v s. i eZtX$4 OU a M g< mar s w t i ru n a.
- U J
,i 4!.
4 9 e 0 00 2 3 2 ~ 0 e 0 ,0 O. 9 1 N 0 E 0 7 ,0 M 1 0 R W 3 S 0 0 Y G 09 E 0 0 S 08 A ) C 0E 0T 0 U P 8 ) N 7 I I T 00M n I 0( ( 0 E R u 3 V 0 r. g E i e a 0 g U 0M r 6 E I M Q T T 0 I 0 04 R A 0 u 0 S 03 SE 0 0 G 0 5 2 A 0 ) 0 S 0 SI 3 M 0 1 0 1 M O n 6 4 2 0 o 0 0 0 8 I R. 5 3 i 3 <bb Z2 l<Mh BaO I I U U
T-9 e-> 9 9 O E o N 6 = e b -f E e CO g o o f e CQ .I <O eG og H m 2 O ~ O' 'I E m 2 = D .$e E 3 s Cy i b s .h II l m ~$ h M i e o C 4 o
- h 8
E I 8 j j j' f N3Do80AH NOIDVEd alon B t B-29
p s.o x e'
O O
N Cs z
b
-k 2w e
S
(/)
O O
O
'h Cz2 O
CD
-l
}}