ML20137D375

Draft Reg Guide,Task Rs 501-4, Best Estimate Calculations of ECCS Performance

ML20137D375
Person / Time
Issue date:	07/12/1985
From:	NRC
To:
Shared Package
ML20137D362	List: IA-85-550, Draft Reg Guide,Task Rs 501-4, Best Estimate Calculations of ECCS Performance ML20137D375
References
FOIA-85-550, TASK-OS, TASK-RS-501-4 REGGD-01.XXX, REGGD-1.XXX, NUDOCS 8508220465
Download: ML20137D375 (34)

Text

i

?

HAFT REGULATORY GUIDE (Task No: RS 501-4)

BEST ESTIMATE CALCULATIONS OF EMERGENCY CORE COOLING SYSTEMS PERFORMANCE A. INTRODUCTION Paragraph 50.46(a)(1) of 10 CFR Part 50 " Acceptance Criteria for Emergency Core Cooling Systems for Light Water Nuclear Power Reactors" requires that light water nuclear reactors be provided with emergency core cooling systems (ECCS) that are designed such that their calculated core cooling performance conforms to certain acceptance criteria specified in Paragraph (b) of that part  ;

and that such performance be calculated by an acceptable evaluation methodology using models with certain required and acceptable features, as set forth in Appendix K to 10 CFR Part 50, "ECCS Evaluation Models." Paragraph (b) (1) of '

9 50.46 requires that the maximum calculated fuel element cladding temperature not be greater than 2200 F. In addition, Paragraphs (b) (2) through (b) (5) of 6 50.46 contain required limits for calculated maximum cladding oxidation and l maximum hydrogen generation, and require that calculated changes in core  !

geometry remain amenable to cooling and that long term decay heat removal be provided. Appendix K, "ECCS Evaluation Models," specifies many required and acceptable features of ECCS evaluation models but leaves unspecified many features of the evaluation methodology which the licensee provides subject to NRC staff approval.

The Nuclear Regulatory Commission staff is proposing to amend the require-ments of 6 50.46 and Appendix K so that the regulations reflect the improved understanding of ECCS performance during reactor transients obtained through the extensive research perfonned since the promulgation of the requirements in S January of 1974. The proposed revision of paragraph 9 50.46 (a) would, if pro-h mulgated, delete the requirement to use Appendix K features and require instead g that evaluation models, which would be subject to acceptance by NRC staff, 4/ include sufficient supporting justification to demonstrate that the analytic u.h technioues provide a reasonable best estimate of the behavior of the reactor M 0,,1=e DRAFT 1 me"Om-~e Sv5 TEM

I UKMri

~

system during a pastulat d isss-of-csolant accid:;nt. The propas:d rule changes would require that the uncertainty in the evaluation model be quantified and considered when compared with the applicable limits in 5 50.46 (b).

This regulatory guide describes the models and methods that are acceptable '

for meeting the requirements for a generic best estimate calculation of ECCS performance during a loss-of-coolant accident and for estimation of the uncer-tainty in that calculation. However, other models and methods will be consid-ered, if supported by appropriate experimental data and technical justification.

In any event, the models and methods which can be accepted in a generic sense would still have to be justified to the NRC staff for particular plant applications.

Paragraph (a) of the proposed revisions to S 50.45 would also permit licensees to continue to utilize the features of Appendix K in their evaluation models with the exception of some minor revisions to Appendix K which are not expected to affect any existing approved evaluation models.

This regulatory guide contains no new information collection requirements and, therefore, is not subject to the Paperwork Reduction Act of 1980 (44 U.S.C. 3501 et seq.). .

B. DISCUSSION The criteria set forth in 10 CFR 50.46 " Acceptance Criteria for Emergency Core Cooling Systems in Light Water Nuclear Power Reactors" and the calcu-lational methods specified in Appendix K were formally issued in January 1974 after extensive rulemaking hearings, and are based on the understanding of ECCS performance available at that time. In the decade following the promulgation of those rules, the Nuclear Regulatory Commission, the nuclear industry, and several foreign institutions have conducted an extensive program of research which has greatly improved the understanding of ECCS performance during a postu-lated loss-of-coolant accident. The methods specified in Appendix K are now known to be highly conservative; that is, the actual fuel cladding temperatures during a loss-of-coolant accident would be much less than the temperatures cal-culated using Appendix K methods. In addition to showing that Appendix K is conservative, the ECCS research allows for quantification of that conservatism.

The results of experiments, computer code development, and code assessment now allow more accurate calculations of ECCS perfoman:e, along with reasonable 07/12/85 2 RG EMER CORE COOLING SYSTEM

4 estimates of uncertainty, during a postulated loss-of-coolant accident, than is possible using the current preuriptive Appendix K procedures.

It is also known that some piarits are being restricted in operating flext-bility by limits resulting from conservative Appendix K requirements. These restrictions are preventing optimal operation of some plants. Based on the research performed, it is now known that these restrictions can be relaxed through the use of more accurate calculations without adversely affecting safety. The Appendix K requirements divert both NRC and industry resources from matters more relevant f,or reactor safety to analyses with known unphysical assumptions.

In recognition of the known conservatisms in Appendix K, the NRC adopted an interim approach described in SECY-83-4728, to accommodate requests for improved evaluation models, submitted by reactor vendors, for the purpose of reducing reactor operating restrictions and providing a realistic evaluation of plant performance. This interim approval is a step in the direction of basing licensing decision on calculations of actual plant behavior. Although the -

approach permits many "best estimate" methods and models to be used for licen-see submitals, it retains those features of Appendix K which are legal requirements. The proposed amendments to 10 CFR 50.46 will, if promulgated, permit licensee ECCS evaluation models to be fully "best estimate" and will i remove the arcitrary corservatisms contained in the required features of Appen-dix K for those licensees wishing to avail themselves of these improved methods.

Throughout this regulatory guide, the term "best estimate" will be used to in- l dicate that the techniques attempt to predict the.true reactor system thermal-hydraulic response and do not include artificial assumptions to provide a conservative bios. The NRC staff has been consistently tending toward the rationale that safety is best served when decisions concerning the limits within which nuclear reactors are permitted to operate are based upon the most accurate

/

calculations possible. This approach is currently being used in the resolution of almost all reactor safety issues (e.g., anticipated transients without scram, pr ssurized thermal shock, and operator guidelines) and is now being proposed for the last remaining issue which is still treated presciptively, the loss-of-coolant accident.

Because of the greatly improved knowledge of ECCS performance during loss-of-coolant accidents, best estimate calculations with reasonable and quantifi- '

able uncertainties can now be performed. The NRC staff is proposing to amend 07/12/85 3 RG EMER CORE COOLING SYSTEM

a

(

Section 50.46 of 10 CFR Part 50 to allow best estimate calculations to be used for the ECCS performance calculations in place of the current evaluation models which use artificial Appendix K features. This rule change will also require analyses of the uncertainty of the best estimate calculation and require that this uncertainty be considered when comparing the results of the calculations to the limits of 9 50.46 (b). In this manner, more realistic calculations will be available for regulatory decisions, yet appropriate conservatism will be maintained consistent with the accuracy of the calculation.

Ma'ny of the methods and models needed for a best estimate calculation are the same as those used previously for evaluation model analyses and, although licensees and applicants are well acquainted with them, explicit guidance on acceptable methods and models (based on NRC experience with its own best esti-mate aavanced codes such as TRAC-PWR, TRAC-BWR, RELAP5 and COBRA) is both pos-sible and appropriate. Further, acceptable methods for the uncertainty analysis have only been documented in SECY-83-4721. Therefore, the NRC staff decided that guidance in the form of a regulatory guide would be useful in order to document procedures acceptable to the reg 11atory staff.

C. REGULATORY POSITION Q

The features presented in Appendix A to this regulatory guide

" Description of Best Estimate Calculations of Emergency Core Cooling Systems /

Performance and Estimate of the Calculational Uncertainty" are acceptable to the NRC-staff for demonstrating reactor vendor's compliance with the proposed revision to Paragraph (a) (1) of 5 50.46.

D. IMPLEMENTATION The purpose of this section is to provide information to applicants and licensees regarding the NRC staff's plans for using this regulatory guide.

This draft guide has been released to encourage public participation in its development. Except in those cases in which an applicant or licensee t proposes an acceptable alternative. method for complying with specified portions of the Comission's regulations, the methods to be described in the final guide reflecting public comments will be used by the NPC staff in evaluating best estimate calculational methods submitted for NRC approval.

DRAFT .

07/12/85 4 RG EMER CORE COOLING SYSTEM

V a

E. VALUE/ IMPACT ANALYSIS A value impact analysis has not been prepared in support of this regulatory !

guide. Such a study was performed as part of the regulatory analysis which supports, the rulemaking effort and is available through the NRC's Public Document Room.

References:

1

1. Information Report from William J. Dircks to the Commissioners, dated November 17, 1983, " Emergency Core Cooling System Analysis Methods,"

SECY-83-472. I i

DRAF 07/12/85 5 RG EMER CORE COOLING SYSTEM

' plW11 I

. 's APPENDIX A Description of Best Estimate Calculations of Emergency Core Cooling Systems Performance and Estimation of the Calculational Uncertainty A. Best Estimate Calculations

1. General A best estimate calculation employs modeling that accurately describes the physical processes of heat transfer and f1'uid dynamics in a nuclear reactor.

There is no unique approach to the extremely complex modeling of the processes occuring during a loss-of-ccolant accident. The NRC has developed and performed assessment upon several best estimate advanced thermal-hydraulic transient codes.

These include TRAC-PWR, TRAC-BWR, RELAP5 and the COBRA series of codes (References A.1, A.2, A.3, and A.4, respectively). The NRC staff considers these codes and the models and methods employed in them acceptable in a generic sense.

This means that the codes predict the major phenomena observed for a broad range of thermal-hydraulic tests acceptably well. Licensees and/or applicants may use these codes and/or specific models within them for the purpose of performing '

their best estimate calculation provided that they demonstrate that the code and/or models are applicable to the specific facility and over the intended operating range.

Features expected in a best estimate calculation are described in this section and specific examples of features that are considered acceptable '

best estimate models will be given in paragraph A.2.

A best estimate model should be designed to provide a realistic calcula- '

tion of a particular phenomenon to the degree required or practical with the currently available data and knowledge of the phenomenon. The model should be compared with applicable experimental data and should predict some sean of the data, rather than providing a bound to the data. The effects of all important ,

variables should be considered. If it is not possible or practical to consider a particular phenomenon, the neglect of this phenomenon should not normally be treated by including a bias in the analysis directly, but should be included as part of the model uncertainty. In this manner, the importance of neglecting this phenomenon can be determined when the overall calculational uncertainty is tvaluated.

07/12/85 RG EMER CORE COOLING SYSTEM

UNAri Careful consideratten should.b2 given to the range of applicability of the mode)

. used in a best estimate code. When comparing the model to data, judgments of the applicability of the data to the situation that would actually occur in a reactor should be made. Correlations generally 1.hould not be extra-polated beyond the range over which they were developed.

If the model is to be extrapolated beyond the conditions where valid data comparisons have been made, judgments should be made as to the effect of this extrapolation and considered in the uncertainty evaluation.

The use of fundamental laws of physics, well established data bases (e.g. , steam tables), and sensitivity studies should be used to assist in estimation of uncertainty which results from extrapolation.

A best estimate code contains all the models necessary to predict the important phenomena that might occur during a loss-of-coolant accident. Best estimate code calculations should be compared with applicable experimental data (e.g. , integral simulations of loss-of-coolant accidents) to determine the over-all uncertainty of the calculation and biases, if any. In addition to provid-ing input to the uncertainty evaluation, integral simulation data comparisons should be used to ensure that all important phenomena that are expected to occur during.a loss-of-coolant accident are adequately calculated. The following par-agraphs list the primary features that are included in best estimate thermal-hy-draulic transient codes. In general these features will have uncertainties asso-I ciated with their use for predicting reactor system response. These uncertainties '

should be considered in the overall uncertainty analysis described in Section 8 of this Appendix.

The above discussion is an idealized definition of best estimate. In  ;

practice, best estimate codes may contain certain models which are simplified and/or contain conservatism to some degree. This conservatism may be introduced '

for the following reasons:

1.

The model simplification or conservatism has little effect on the result and, therefore, the development of a best estimate model is not justified, 2.

The uncertainty of a particular model is difficult to determine and '

only an upper bound can be determined, or 3.

The particular application does not require a totally best estimate calculation and a bias in the calculation is acceptable.

The introduction of conservative bias or simplification in otherwise best estimate codes should not, however, result in calculations which are unphysical, 07/12/85 7 RG EMER CORE COOLING SYSTEM

'. 's unrealistic, do not include all important phenomena, or contain bias and ,

uncertainty that cannot be bounded. Therefore, any calculational procedure determined to be a best estimate code in the context of this guide or for use under the revised 10 CFR 50.46 should be compared with applicable experimental data to ensure that realistic calculations of all important phenomena result.

2. FEATURES OF BEST ESTIMATE CODES The features that are acceptable for use in best estimate codes are based on NRC's experience with its own best estimate thermal-hydraulic transient codes.

Licensee best estimate codes which are benchmarked against the applicable NRC code will be censidered acceptable provided that the geometry and the scenarios assumed are comparable. In some cases, acceptable models are specifically iden-tified, but other models may be used with the basic requirement that a specific model is acceptable if it has been compared with applicable experimental data.

These models and others used in NRC best estimate codes are acceptable in the generic sense in that the codes themselves have been assessed against a wide variety of experimental data. These models have provided the best reoresenta-tion of the integral thermal-hydraulic test data to date. However, the use of these models in any particular plant analysis would have to be justified for that specific application. That is, the scalability of the data and analysis, the phenomena encountered and the importance of modeling must all be consid-ered. Reference A.5 provides a summary of the large experimental data base available. While inclusion in Reference A.5 does not necessarily guarantee that the data or model will be found acceptable, the report represents a large bndy of data generally applicable to best estimate models. In fact, it consti-tutes the data base used by the NRC to justify the assessment of its own best estimate codes.

The following paragraphs list the primary features that are included in best estimate thermal-hydraulic transient codes. NRC's review of 1icensee sub-mitted best estimate codes will be primarily focused upon the codes themselves and not normally upon individual features within them. While individual fea-tures that are discussed in this regulatory guide have been demonstrated to be acceptable, their use in conjunction with other features may or may not produce acceptable results. Therefore, sufficient justification must be provided to 07/12/85 8 RG EMER CORE COOLING SYSTEM

'. 'e '

substantiate that th) best estimate ccde performs ad;quately for th2 classss of transients to which it it applied. In general these features have uncertainties associated with their use for predicting reactor system response. These uncer-tainties should be considered as part of the overall uncertainty analysis de-scribed in Section B of this Appendix.

a. Basic Structure of Code (i) Numerical Methods A best estimate code consists of a numerical scheme for solving the equa-tions used to represent the various models. The numerical scheme is in itself a complex phenome:,on and can play an important role in the ove, ell calculation.

Careful numerical modeling, sensitivity studies, and evaluations of numerical diffusion should be performed to ensure that the results of the calculations are representative of the models used in the code. Numerical simulations of complex problems such as those considered here treat the geometry of the reactor in an approximate manner, making use of discrete volumes or nodes to represent the system. Sensitivity studies and evaluations of the uncertainty introduced by noding should be considered to obtain optimal resultr.

(ii) Computational Models F A best estimate code typically contains equations for conservation of mass, energy, and momentum of the reactor coolant and noncondensible gases, if important (e.g. , air, nitrogen, etc). Energy equations are also used to calcu-

, late temperature distribution in reactor system structures and the fuel rods.

The required complexity of these equations will vary depending on the phenomena that are to be calculated and the required accuracy of the calculation. NRC staff experience with its own bes*

• stimate computer codes has found that sepa-rate flow fields for different fluid phases, or types, and calculation of non-equilibrium between phases may be required to calculate some important phenomena to an acceptable accuracy. NRC staff has also determined that certain phenomena require that the equations be solved in multiple dimensions. Therefore, this latter capability has been developed in TRAC and the COBRA series of codes.

07/12/85 9 RG EMER CORE COOLING SYSTEM

i How2ver, ene-dimensional apprcxications to three-dimensicnal ph2nomena will be considered acceptable, if those approximations are properly justified. Other basic code features include equations of state and other material properties.

Sensitivity studies and comparisons to data should be performed to determine the importance of the simplifications used.

b. Initial and Boundaey Conditions The heat generated by the fuel during a loss of-coolant accident depends on the power level of the reactor at the time of the loss-of-coolant accident and the history of prior operation. These initial conditions may be treated in one of two ways. Either the most likely initial conditions expected over the life of the plant may be used, and the expected variation in initial coaditions treated as part of the overall uncertainty evaluation, or the worst expected initial conditions (e.g., technical specification limits) used in the calcula-tion.

It is not necessary to assume initial conditions which are impossible to occur in combination. For example, beginning of-life peaking factor and end-of-life decay heat may be a unrealistic condition and, therefore, does not require consideration. Given the assumed initial conditions, relevent factors such as the actual total power, actual peaking factors, and actual fuel conditions should be calculated in a best estimate manner.

Calculations should be performed that are representative of the spectrum of possible break sizes from the full double ended break of the largest pipe to a size small enough that it can be shown that smaller breaks are of less conse-quence than those already considered. The analyses should also include the

! effects of longitudinal splits in the largest pipes, with the split area equal to twice the cross-sectional area of the pipe. The detail of break sizes con-sidered should be sufficient so that the system response as a function of break size is defined well enough to confidently interpolate between calculations, l without unexpected behavior between the break sizes considered. The break l should be assumed to occur instantaneously unless it can be shown through appro-priate analyses and comparisons with experimental data that break propagation can be realistically predicted.

l After accounting for a single failure and loss of offsite power, other boundary and initial conditions may be assumed to be that which are most probable over the life of the plant, with the variation in expected conditions or uncertainty in the conditions treated as part of the overall uncertainty 07/12/85 10 RG EMER CORE COOLING SYSTEM

~

evaluation. Tnese other conditions include, but may net be limited to, availability and performance of equipment, automatic controls, and operator actions. It should be noted that Appendix A to 10 CFR 50 requires that loss of offsite power and a single failure be considered when analyzing safety system performance.

As an alternative to an analysis in which the best estimate, or most likely, initial conditions are determined, the NRC staff will accept the following assumed initial conditions:

1. Loss of offsite power assumed,

2. Worst single equipment or control failure assumed, and

3. Technical specification limits on equipment availability or performance.

c. Sources of Heat During a Loss-of-Coolant Accident The sources'of heat discussed in paragraphs i through iv below, and the distribution of heat production should be accounted for.

(1) INITIAL STORED ENERGY OF THE FUEL - The steady state temperature dis-tribution and stored energy in the fuel before the postulated accident should be calculated in a best estimate manner for the assumed initial conditions, fuel conditions, and operating history. To accomplish this, the thermal con-ductivity of the UO 2should be evaluated as a function of burnup and tempera-ture, taking into consideration differences in ini.tial density. Thermal conductance of the gap between the UC 2 and the cladding should be evaluated as a function of the burnup, taking into con.,ideration fuel densification and expan-sion, the composition and pressure of the gases within the fuel rod, the initial cold gap dimension with its tolerances, and cladding creep. An acceptable method for calculating the initial stored energy is found in the FRAPCON-2 code and described in Reference A.6. Other best estimate methods will be considered acceptable provided their technical basis is demonstrated with appropriate data and analyses.

(ii) FISSION HEAT - Fission heat should be included in the calculation and should be calculated using best estimate reactivity and reactor kinetics calculations. Shutdown reactivities resulting from temoeratures and voids should also be calculated in a best estimate manner. The point-reactor kinetics formulation in the NRC best estimate codes is considered an acceptable best 07/12/85 11 RG EMER CORE COOLING SYSTEM

j esticato methed for determining fission heat in loss-of-coolant accident safety calculations during which scram occurs soon after the break is initiated. Other

_best estimate models will be considered acceptable provided their technical basis is demonstrated with appropriate data and analyses. Control rod insertion should be assumed if it is calculated to occur.

(iii) DECAY OF ACTINIDES - The heat from radioactive decay of actinides, including neptunium and plutonium generated during operation, as well as iso-topes of uranium, should be calculated in accordance with fuel cycle calcula-tions and know, radioactive properties. The actinide decay heat chosen should be that appropriate for the operating histcry.

(iv) FISSION PRODUCT DECAY HEAT - The heat generation rates from radioac-

{

tive decay of fission products should be. included in the calculation and should be calculated in a best estimate manner. The values in the ANs 1979 Decay Heat Standard (Reference A.7) using the assumed operating history, are acceptable for l use in calculating decay heat.

Other best estimate methods will be considered acceptable provided that their technical basis is demonstrated with appropriate data and analyses. '

(v) METAL WATER REACTION RATE - The rate of energy release, hydrogen gen- I eration, and cladding oxidation from the reaction of the Zircaloy cladding with steam should be included in the calculation in a best estimate manner. One I acceptable manner in which to perform this calculation is described by Cathcart I and Pawel (Reference A.8). Other best estimate models will be considered accept- !

able provided that their technical basis is demonstrated with appropriate data and analysis.

The reaction should not be assumed to be steam limited. For rods  ;

whose cladding is calculated to rupture during the loss-of-coolant accident, the  ;

inside of the cladding should be assumed to oxidize extending around the cladding inner circumference and axially no less than 1.5 inches in each

, direction from the location of the rupture, unless it can be shown through the combination of experiments and analyses that another assumption is justified.

(vi) REACTOR INTERNALS HEAT TRANSFER - Heat transfer from piping, vessel walls, and internal hardware should be included in the calculation and should be calculated in a best estimate manner. Acceptable modeling of these processes can be found in the NRC sponsored best estimate codes. Other models will be considered acceptable provided their technical basis is demonstrated with approx-inate data and analyses.

(vii) PRIMARY TO SECONDARY HEAT TRAN.iFER (not applicable to boiling water reactors) - Heat transferred between the primary and secondary systems through 07/12/85 12 RG EMER CORE COOLING SYSTEM

. . L)Mt i P tha ster.:a generators should b2 considered in th2 calculation and should b2 cal-culated in a best estimate manner. State-of-the art modelling of primary to secondary heat transfer is contained in the NRC developed best estimate thermal-hydraulic codes for pressurized water reactors.

Other best estimate models will be considered acceptable provided that their technical base is demonstrated with .

appropriate data and analyses.

d.

Swelling and Rupture of the Cladding and Fuel Rod Thermal Parameters A calculation of the swelling and rupture of the cladding resulting from ,

the temperature distribution in the cladding and from the pressure difference between the inside and outside of the cladding, both as a function of time, should be included in the analysis and should be performed in a best estimate manner. The degree of swelling and rupture should be taken into account in calculations of gap conductance, cladding oxidation and embrittlement, hydrogen generation, and in calculating heat transfer and fluid flow outside of the cladding. The calculations of fuel and cladding temperatures as a function of time should use values of gap conductance and other thermal parameters as functions of temperature and time. Acceptable best estimate methods for calculating the swelling and rupture of the fuel cladding and fuel rod thermal parameters are contained in the FRAPCON-2 code and described in Reference A.6.

Other best estimate methods will be considered acceptable provided that their technical basis is demonstrated with appropriate data and analysis.

e. Blowdown Phenomena (i) BREAK CHARACTERISTICS AND FLOW - In analyses of hypothetical loss of-coolant accidents, a spectrum of.possible break sizes should be considered as indicated above. This spectrum should include double-ended breaks ranging in cross-sectional area up to and including the largest pipe in the primary coolant i system.

The analysis should include longitudual split in the largest pipes, with the split area equal to the cross-sectional area of the largest pipe.

Discharge flow rate should be calculated with a state-of-the-art critical flow rate model which considers the fluid conditions at the break location, upstream and downstream pressures, and break geometry. The critical flow model should be justified'by comparison to applicable experimental data over a range of con-ditions for which the model is applied. The model should be a best estimate 07/12/85 13 RG EMER CORE COOLING SYSTEM i

UtWI calculation, prsdicting th2 mean of applicable experimental data. Such state-of-the art models are included in the NRC developed best estimate thermal-hydralic transient codes.

Other best estimate models will be considered acceptable provided that their technical basis is demonstrated with appropriate data and analysis.

The NRC staff will not normally consider the adequacy of the break flow model separately, but will evaluate it in conjunction with the other models in the code and for the specific scenarios it is applied to and the ranges of applicability in question.

(ii) ECCS BYPASS - The best estimate code should contain a calculation

{

of the amount of injected cooling water which bypasses the vessel during the blowdown phase of the loss of-coolant accident.

This water calculated to be j expelled from the vessel prior to the "end-of-bypass" should be accounted for in the primary coolant mass balance. The calculation of end of-bypass should i

be a best estimate calculation using analyses and comparisons with applicable  !

experimental data.

NRC Research Information Letter 128 (Reference A.9) presents 1 the results of NRC's Emergency Core Coolant Bypass Program. The appendix to this letter presents best estimate models of the ECC bypass phenomenon. These best estimate models are considered acceptable to the NRC staff for predicting reactor coolant system inventory distribution during ECC-bypass conditions. The NRC expects improved information concerning the ECC-bypass phenomenon for large scale facilities to be available from tests conducted in the Upper Plenum Test Facility in Federal Rupublic of Germany. These data are expected to reduce the scaling uncertainty in the models presented in RIL 128.

Although it is clear that the dominant processes governing ECC bypass are multidimensional, single dimensional approximations justified through suffi-cient analysis and data may be acceptable. Also other best-estimate methods will be considered accpetable provided that their technical basis is demonstrat-ed with appropriate data and analysis. Cooling water injected prior to the end-of-bypass that is not expelled, but remains in piping or is stored in parts of the vessel, is not required to be subtracted from the vessel inventorv if justification for retaining this liquid can be made using best estina'e analyses based on applicable experimental data.

f. Modino near the Break and ECCS Injection Point The break location and ECCS injection point are areas of large thermal non equilibrium and contain phenomena that are often difficult to calculate.

07/12/85 14 RG EMER CORE COOLING SYSTEM

~

UlW) f The results of these calculations are often highly d: pendent on the noding.

Sufficient sensitivity studies should be performed on the noding and other important parameters to ensure that the calculations provide realistic results and are not influenced by numerical phenomena.

g. Frictional Pressure Drops The frictional losses in pipes and other components should be calculated using models that include variation of friction factor with Reynolds number and two phase friction multipliers. Best estimate models that have been adequately compared with experimental data should be used. Standard two phase friction multipliers have been acceptably utilized in the NRC sponsored best estimate codes are considered suitable for use in best estimate thermal hydraulics codes.  ;

0ther best estimate methods will be considered accpetable provided their techni-cal basis is demonstrated with appropriate data and analysis.

h. Momentum Equation

-The following effects should be taken into account in the two phase conser-vation of momentum equation: (1) temporal change in momentum, (2) momentum convection, (3) area change momentum flux, (4) momentum change due to compressi-bility, (5) pressure loss resulting from wall friction, (6) pressure loss re-sulting from area change, and (7) gravitational acceleration. Acceptable I methods for handling momentum during a loss-of coolant accident are incorporated in NRC developed best estimate themal-hydraulic codes. Other best estimate methods will be considered acceptable provided their technical basis is demon-strated with appropriate data and analysis.

1. Critical Heat Flux Best estimate models developed from appropriate steady-state or transient experimental data should be used in calculating critica. heat flux (CHF) during loss-of-coolant accidents. The codes in which these models are used should contain suitable checks to assure that the range of conditions over which these correlations are used are within those intended. Acceptable models for predict- i i

ing critical heat flux include the Biasi correlation used in TRAC-PWR anc TRAC- ,

BWR (References A.10 and A.11). TRAC-BWR also uses the CISE-GE correlation  ;

(Reference A.11).

07/12/85 15 RG EMER CORE COOLING SYSTEM I

f k

, o .,

DRAFT Other models will be acceptable provided that their technical basis is demon- i strated with appropriate data and analysis. After CHF is predicted at an axial-fuel rod location, the calculation should use nucleate boiling heat transfer correlations if the calculated local fluid and surface conditions justify the re establishment of nucleate boiling. The heat transfer model used to cal-culated return to nucleate boiling should be justified by comparison to appli-cable experimental data and the uncertainty of the model considered in the un-certainty evaluation.

j. Post-CHF Heat Transfer Models of heat transfer from the fuel to the surrounding fluid in the post-CHF regimes of transition and film boiling should be best estimate models j

based on comparison to applicable steady-state or transient data. State of-the-art best estimate correlations for the minimum stable film boiling temperature should be included in the calculation. At the time of the writing of this regulatory guide, there is no single minimum stable film boiling temperature correlation which provides acceptable results for all loss-of-coolant accident scenarios.

Any correlations should be evaluated to demonstrate that they provide accept-able results over the applicable ranges. Other correlations will be considered acceptable provided their technical basis is demonstrated with appropriate data and analysis. The NRC has an effort in place to develop improved correlations for post-CHF heat transfer. The results will be incorporated in improvad versions of NRC sponsored codes.

k. Pump Modeling The characteristic of rotating primary system pumps should be derived from a best estimate dynamic model that includes momentum transfer between the fluid and the rotating member, with variable pump speed as a function of time. The pump model resistance and other empirical terms should be justified through comparisons with applicable data. The pump model for the two phase region should be verified by comparison to applicable two phase performance data.

Pump coastdown following pump trip should be treated in a best estimate manner.

A locked rotor following a large break loss of-coolant accident should not be assumed unless it is calculated to occur.

07/12/85 16 RG EMER CORE COOLING SYSTEM

. o ,

1. Core Flow Distributicn During Blowdown .

The core flow through the hottest region of the core during the blowdown

]l should be calculated as a function of time. For purposes of these calculations, il the hottest region of the core should not be greater than the size of one fuel assembly. Calculations of the flow in the hot region should take into account any cross flow between regions and any flow blockage calculated to occur during the blowdown as a result of cladding swelling or rupture. The numerical scheme should ensure that unrealistic oscillations of the calculated flow do not result.

m. Post-Blowdown Phenomena

, (i) CONTAINMENT PRESSURE - The containment pressure used for evaluating cooling effectiveness during reflood and spray cooling should be calculated in r

I a best estimate manner including the effects of containment heat sinks. The calculation should include the effects of operation of all pressure reducing equipment assumed to be available as discussed in Paragraph A.2.6 of this Appendix.

(ii) CALCULATION OF REFLOOD RATE FOR PRESSURIZED WATER REACTORS - The re-filling of the reactor vessel and the time and rate of reflooding of .the core should be calculated by a best estimate model that takes into consideration the thermal and hydraulic characteristics of the core and of the reactor system.

The primary coolant pumps should be assumed to be operating in the expected manner based on the assumptions of Paragraph A.2.b when calculating the resis-1 tance offered by the pumps to fluid flow. Acceptable correlations for calculating j core refill and reflood rate were developed as part of the FLECHT-SEASET experi-

] mental program (References A.12 and A.13) and are documented and may be used in a best estimate code. Mechanistic models which treat these conditions are included in the NRC sponsored best estimate codes and are also considered accept-able by the staff. Other correlations will be considered acceptable provided that their technical basis is demonstrated through comparison with appropiate j data and analysis. The total fluid flow leaving the core exit (carryover) j should be. calculated esing a best estimate model which includes the effect of cross-flow on carryover and core fluid distribution. The effects on reflooding rate of the compressed gas in the accumulator which is discharged following accumulator water discharge should be included in the calculation. Prediction of the reflood rate is one of the most difficult aspects of loss-of-coolant 07/12/85 17 RG EMER CORE COOLING SYSTEM l'

i

o. ,

1 accident analysis, and any model or code used for 'this calculatio e assessed against available experimental data.

(iii)

STEAM INTERACTION WITH EMERGENCY CO WATER REACTORS - The thermal-hydraulic interaction between th e steam and the emergency core cooling water should be taken into account e in ca

core reflooding rate and the steam flow through the reactor s dur- coolan ing the time that the accumulators are discharging water t (iv) .

REFILL AND REFLOOD HEAT TRANSFER FOR -

i During the refill and reflood portions of a postulated loss accident of-coolan calculation, the heat transfer should be based on a bestnestimat!

of the fluid flow through the core and should include thew effects blockage calculated to occur as a result of cladding swelling upture. Heat or r transfer calculations which account for two phase conditions in th  !

e core during refill should be justified through comparisons with experimental . Accept-data able correlations for calculating core refill and reflood heat were transfer

developed as.part of the FLECHT-SEASET experimental program a d i_n References A.12 and A.13. n are documented Mechanistic models which address refill and re flood heat transfer are incorporated in the NRC developed best esti and are considered acceptable by the staff. mate codes Other models will be considered acceptable provided that their technical basis is demonstrated pari-through(

son with appropriate data and analysis.

n.

Convective Heat Transfer coefficients for Boiling Water Reactor l Rods Under Spray Cooling I

NRC developed best estimate thermal-hydraulic transient ng codes f water reactors contain acceptable methods for treatment of convective transfer during spray cooling. i Other methods will be considered provided their technical basis can be justified with appropriate data and .

analysis These methods should contain the following:  !

(i)

, Following the blowdown period, convective heat transfer coeffic should be determined based on the calculatede fluid co and the calculated rod temperatures.

Heat transfer models should be best esti-mate models that have been compared to appropriate experimental data .

l 1

07/12/85 18 RG EMER CORE COOLING SYSTEM l

' * l l

(ii) During the period follping flashing, but prior to ECCS initiation, l I

heat transfer models should include cooling by steam flow or by a two phase mixture, if calculated to occur.

(iii) Following initiation of ECCS flow, but prior to reflooding, heat transfer may be based on the actual calculated bundle fluid conditions and best estimate heat transfer models which take into account rod-to rod variations in heat transfer. As an alternative, convective heat transfer coefficients of 3.0, 3.5, 1.5, and 1.5 Btu-br~1-ft'# ~lF may be applied to fuel rods in the outer corners, outer row, next outer row and to the remaining rods in the inte-rior, respectively, of the assembly once full rated core spray is reached.

(iv) After the two phase reflood level reaches the level under considera-tion, either a best estimate heat transfer model may be used or, as an alterna- i tive, a heat transfer coefficient of 25 Btu-br~1 ft -2*p-1 may be applied to all fuel rods.

(v) Thermal-hydraulic models which do not calculate multiple channel ef-fects and do not make use of the simplified alternatives of (3) and (4) above should be compared with applicable experimental data or more detailed calcula-tions to ensure that important phenomena are not missed by not explicitly cal-culating multiple channels.

o.

The Boiling Water Reactor Channel Box Under Spray Cooling Following the blowdown period, heat transfer from the channel box and wet- '

ting of the channel box should be based on the calculated fluid conditions on both sides of the channel box and should make use of best estimate heat transfer and rewetting models that have been compared with applicable experimental data.

The following heat transfer coefficients and wetting time correlations are acceptable:

(i) During the period after lower plenum flashing, but prior to the core l

spray reaching rated flow, a convective coefficient of zero should be applied to the fuel channel box.

(ii) During the period after core spray reaches rated flow, but prior to wetting of the channel, a convective heat transfer coefficient of 5 Btu-hr'A-ft' # -1 F should be applied to the sides of the channel box.

07/12/85 19 RG EMER CORE COOLING SYSTEM

• , o.,

(iii) Wet, ting of the channel box should be assumed to occur 60 seconds after the time determined using the correlation based on the Yamanouchi analysis (see Reference A.14) '

Acceptable be'st estimate models are also contained in the NRC sponsored best estimate thermal-hydraulic codes. Other methods will be considered accept-able provided that their technical basis is defronstrated with appropriate data and analysis.

p.

Other Features of Best Estimate Codes No list of best estimate code features could be all-inclusive because the important features of a best estimate code may vary depending on the transient to be calculated and the required accuracy of the calculation. Because of this, no attempt has been made to construct an exhaustive list of best estimate code featt.res. Rather, features which were identified as important for inclu-sion in Appendix K were used as a basis for the above list. These features are  ;

not necessarily any more or less important than other code features, but were highlighted because it is necessary to give specific examples of how current best estina'te models may vary from methods used traditionally in evaluation model codes using the various Appendix K conservatisms. The NRC staff believes that the best examples of best estimate thermal.-hydraulic transient codes are those developed by the NRC: TRAC-PWR, TRAC-BWR, RELAPS, and COBRA. Although these codes are subject to further improvement, based on their ongoing use and '

assessment, they currently provide a reasonable and acceptable best estimate calculation of loss-of-coolant accident in a full-scale light water reactor.

This is substantiated through the code development and assessment literature '

generated by the NRC and its contractors over the past several years.

It is possible, however, to generally describe how other features of best estimate codes should be constructed. Two basic criteria should be applied:

(1) COMPLETENESS - Best estimate code:: should contain all models in sufficient detail to predict all phenomena that are important to the desired result of the calculation (e.g.. , peak cladding temperature). Simplifications are acceptable as long as code uncertainties or bias do not become so large so as to cast doubt on the actual behavior that would occur and the true effect of assumed initial and boundary conditions (e.g., equipment sizing, safety system 07/12/85 20 RG EMER CORE COOLING SYSTEM 4

UKAtI sotpoints, etc.).

Comparisons of the ovarall calculations to integral experi-ments should be performed to ensure that all important phenomena can be predicted and to help in making judgments on the effect of code simplifica-tions.

Consideration should also be given to the uncertainty and validity of the experiment to ensure that meaningful comparisons are being made.

(11)

DATA COMPARISONS - Individual best estimate models should be compare to applicable experimental data to ensure that realistic results are predicted and that all relevent experimental variables are included. Uncertainty analys-es are required to ensure that a major bias does not exist in the models and that the model uncertainty is small enough so as to provide a realistic esti-mate of the effect of important experimental variables. Uncertainty analyses should also consider experimental uncertainty to ensure that meaningful compar-isons are being made.

B.

Estimation of Overall Code Uncertainty

1. General The term " uncertainty," when applied to best estimate thermal-hydraulic transient codes, is used at two levels.

At the lower or more detailed level, it refers to the degree to which an individual model, correlation, or method used within the code represents the physical phenomenon which it addresses.

These individual uncertain' ties, when taken together, contribute to the overall uncertainty associated with the code. This higher level uncertainty is an estimate of the confidence which can be placed upon a code's prediction of reactor system thermal-hydraulic response during loss-of-coolant accidents.

For example, if a best estimate thereal-hydraulic transient code predicts that the maximum cladding temperature would be 1000*F during a certain large break loss-of-coolant accident and the overall uncertainty evaluation indicates that the standard deviation assoc uted with that prediction is 50*F, then it can be concluded that there is 95% confidence that the actual maximum temperature will not exceed 1082*F (one-sided hypothesis test with z = 1.64a). The preceding example assumes that the observed temperature is drawn from a population which is normally distributed with zero bias. This assumption must be justified and l

the distributional differences and biases accounted for in the uncertainty analysis. The following section will describe those features which should be 07/12/85 21 RG EMER CORE COOLING SYSTEM

e d

• included in the calculation of the estimate of the overall uncertainty associated with the use of a best estimate code.

2. Features of the Overall Uncertainty Calculation Uncertainties associated with individual models within the best estimate codes do not account for all of the uncertainty associated with its use. Vari-ous other sources of uncertainty are introduced when attempting to use best estimate codes to predict full-scale plant thermal hydraulic response. These include uncertainty associated with the experimental data used, the input bound-ary and initial conditions, the fuel behavior, and the code user. Additional sources of uncertainty stem from the use of simplifying assumptions and approxi-mations. A careful statement of these assumptions and approximations should be made and the uncertainty associated with them snould be accounted for. A 95%

probability level is considered acceptable to the ~NRC staff for comparison of best estimate predictions to the applicable limits of S 50.46 (b).

The basis for selecting the 95% probability level is primarily for consis-tency with what has been standard engineering practice in regulatory matters involving thermal-hydraulics. Many parameters, most notably the departure from nucleate boiling ratio (DNBR), have been found acceptable at the 95% probabil-ity level by the staff in the past. It is useful, however, to examine this criterion on a risk basis. To illustrate by way of an example, assume that the probability of a large pipe rupture is approximately 1 x 10 ' per reactor year.

Assume also that a hypothetical large break loss-of-coolant accident calculation results in a best estimate peak clad temperature 2000 F with a standard devia-tion of 122 F at the 95% probability level. Then the probability of exceeding the 2200 F limit in 10 CFR 50.46(b) would be .05. Thus, the probability of a large pipe rupture resulting in a peak cladding temperature exceeding 2200 F would be 5 x 10 8 Only the hottest rods would exceed the 2200 F limit under these assumptions and the 2200 F limit is itself a conservative requirement chosen to prevent significant embrittlement of the cladding to guarantee that a coolable geometry is maintained. The accident at TMI-2 resul*.ed in a signifi-cant percentage of the core exceeding the 2200 F limit for long periods of time. Despite the significant damage, the core still remaired coolable.

Therefore, if it is further assumed that gross core melt oc:urs in only one in ten cases where the 2200 F limit is exceeded, the overall probability of core 07/12/85 22 RG EMER CORE COOLING SYSTEM

[HWI me,t would be on the order of 5 x 10 7 This probability level is well below many other significant contributors to core melt risk and, therefore, the use of a 95% probability level should ensure that large break loss-of-coolant  ;

accidents are not.important contributors to risk.

This 95% probability level would also be applied to small loss-of-coolant accidents which have a higher probability than large breaks. The dominant factors influencing risk from small loss of coolant accidents include equipment availability and operator actions. Calculational uncertainties are much less important than factors such as operator recognition of the event, the availabi-lity of equipment, and the correct use of this equipment. The use of a best estimate calculation with reasonable and quantifiable uncertainty is expected

~

to provide a reduction in the overall risk from small break LOCA by providing acre realistic calculations with which to evaluate operator guidelines and to determine the true effect of equipment availability.

This section provides a description of the features that should be includ-ed in the overall uncertainty evaluation associated with best estimate thermal hydraulic transient codes.

The uncertainty evaluation should make use of sta- ,

tistically based methods to determine the code uncertainty and the bias (if any).

For a calculation of this complexity, a completely rigorous mathematical treatment of the statistics is neither practical nor required. In many cases, approximations and assumptions may be made to make the overall uncertainty eval-untion possible. A careful statement of these assumptions and approximations enables a judgement of their affect to be made. ,

The purpose of the uncertainty i, evaluation is to demonstrate that for postulated loss-of-coolant accidents a I' given plant will not, with a probability of 95% or more, exceed the applicable limits specified in 10 CFR 50.46 (b).

d i

2. Code Uncertainty i I

t t

l This regulatory guide makes a distinction between the terms " code uncer-tainty" and "overall code uncertainty." The latter term is defined in Para-graph B.1 of this Appendix and includes the contributions to the uncertainty described in Paragraphs B.2 and 8.3.

The features of the code uncertainty, the contribution to the overall uncertainty due to the models and numerical methods used, will be described in Sections B.2.a through B.2.g. The code uncertainty should be evaluated through direct data comparisons with relevant integral l

07/12/85 23 RG EMER CORE COOLING SYSTEM l

th2rm31-hydraulic Gxperiments and should address the important parameters for a broad range of transients,

a. Break Spectrum i The uncertainty analysis should be applied to a range of break sizes in order to demonstrate that transients which were not analyzed would have less severe consequences than the ones analyzed.

Events which should be included are large intermediate and small break loss of-coolant accidents.

Sensitivity studies aimed at determining ifmiting case transients may be performed in order to reduce the number of analyses required. Such studies should be accompanied with sufficient supporting information to corroborate the inferences drawn.

b. Key Parameters

-l Accuracy statements should be based on a nell chosen, limited number, of parameters, which are referred to as key parameters. The key parameters selected for use in the uncertainty analysis should include both single-valued parameters and continuous-valued parameters.

A set of continuous valued key parameters, not all of which apply to a given problem, but which include all potential major '

phenomena are discussed in the following paragraphs. These are supplemented a number of single-valued key parameters, generally measuring either maximum or minimum values of the continuous-valued parameters, or event timings.

Because of its direct relationship to fuel damage, cladding temperature is the primary key parameter and its uncertainty should always be calculated.

Depending on the transient other parameters will be required so that the uncer-tainty in clad temperature is better understood. To avoid large numbers of temperature comparisons, the core hot spot may be used as the location of most interest.

Primary system pressure is also a key parameter. Many accident and trans-ient signatures are identified by their unique pressure behavior, and many subordinate plant actions are controlled by the primary system pressure (e.g. ,

high or low pressure trip setpoints, ECC injection rates, break and/or valve flow rates). Pressure maximums and/or minimums and the transient times 07/12/85 24 RG EMER CORE COOLING SYSTEM

they are reached become corresponding single valued key parameters.

Break flow or, more generally, the mass flow rate out of the primary system (thus including flow rate out a power operated relief valve, for example) may be considered a key parameter.

The integrated outflow (which is normally.a smoother function of time) may be used as a key parameter, but temporal break flow data is more readily available and the break flow prediction is generally of interest because it gives an indication as to the adequacy of the best estimate critical flow model.

Just as the break (and/or valve) flow measures the mass lost from the primary system, defining the total emergency core cooling injection flow rate as a key parameter allows evaluation of the mass flow into the system, to complete the overall primary system hydraulic behavior description.

The accuracy in predicting important aspects of the local distribution of i

the primary sy; tem inventory should be evaluated using a comb 6i.1on of single-valued key parameters (i.e., times for loop seal clearing, or for pressurizer filling and/or emptying) and the next key assessment parameters, the vessel or core collapsed liquid level for primary system loss of-coolant accidents and/or the pressurizer collapsed liquid level for secondary side loss-of-coolant acci-dents.

Level peaks and/or minimums and the transient times at which they occur are additional single-valued key parameters.

The secondary side pressure should be used as a key parameter. The onset and maintenance of steam generator reverse heat transfer have a significant <

impact on important phenomena such as the formation and clearance of Coop seals.

Pressure peaks and/or minimums and the transient times at which such trip set-points are reached should be used as corresponding single valued key parameters.

Finally, the natural circulation flow rate should be considered a key para-meter, if the coolant pumps are tripped or fully degraded. Natural circulation i

flow is chosen because of its strong influence on core coolability. Equivalent single-valued key parameters are the magnitudes of the subcooled natural circu-4 lation flow and the peak two phase natural circulation flow and the fractional system inventories corresponding to transitions between different natural circu-lation modes.

07/12/85 25 RG EMER CORE COOLING SYSTEM

The key parameters for accuracy quantification for large break loss-of-coolant accidents would be the temperture at the hot spot, primary system pressure, the break flow, and the ECC injection. Pressure in the intact loop cold leg is important because.this location is far enough from the break planes to avoid the shnp pressure gradients present very near the breaks, the pressurizer pressure can be misleading because it can be affected by test-dependent surge line effects, and if the pressure nearest the ECC injection point is not well predicted, then the ECC injection will be mispredicted as a direct consequence. The break flow in a larga break loss-of coolant accident consists of two components, the l

vessel-side and pump-side break flows, which can be evaluated separately, but also require a combined accuracy statement, j

For small break loss-of-coolant accidents, the key parameters for accuracy quantification are the cladding temperture at the hot spot, primary system pressure, the break flow, the ECC injection, the vessel or core collapsed liquid

)

level, the secondary side pressure, the secondary side valve flow (if any) and the natural circulation flow rate (if the pumps are tripped or degraded). The primary pressure in the intact loop cold leg should be evaluated for the same reasons given in the large break loss-of-coolant accident discussion. The se-condary side pressures shouid be evalueted in the steam generator steam domes.

c. Partitioning A quantitative accuracy statement may be derived over the entire time range covered by both measurement and calculation. Alternatively, individual accuracy statements can be derived for major time intervals chosen to isolate different governing phenomena during different portions of the accident or transient being I studied.

A given key parameter may be significant only during some part of the entire test and/or calculation. Furthermore, the particular behavior of interest (e.g.,

core heat-up) may occur in the test but not in the calculation, or vice versa.

9 Major intervals in which the code accuracy is quantified could therefore be limited to regions where phenomena of interest are expected and where the code predicts the same phenomena as occurred in the experiment. The code accuracy quantification should be carried out in all the major intervals defined, even 07/12/85 26 RG EMER CORE COOLING SYSTEM

, .,e, UM{j if tiu results are exp cted to b2 mora meaningful in some time regions than in others.

The main difficulty with limiting quantification only to those major time intervals where approximately " correct" behavior is being calculated is that the resulting statistics will be skewed to those cases and the resulting code error estimates will be biased on the low side. If in some analyses core heat up is observed, but not calculated (or contrariwise), this should appear as a large error in the overall accuracy quantification, not as a zero-input. Always providing the accuracy quantification information, divided into well chosen major time regions, will allow later processing in various combinations, including all regions or only sae.

The calculation and test should cover the same period of time. Thus, the end boundary of the last major time interval is defined. Another obvious time boundary is transient start. If a steady state period before time zero is present in both the data and the calculation (which is useful for plotting purposes), it forms an obvious major interval.

The internal time boundaries betwee'n major intervals for most types of transients can be tied to single-valued key parameters which correspond to the timings of various events.

These include time of accumulator initiation, time of loop seal clearing, time of MSIV main steara isolation valve closure, time of power operated relief valve cycling onset, etc., which often mark the beginning or end of various transient phenomena of interest.

All such possible key timings to define major transient intervals need not be used. The total number of major intervals for a given transient should be limited, and should be chosen to allow the maximum possible overlap and/or cross-comparison between different break sizes. Two to four major time intervals (not counting steady state), based on significant changes expected in governing physics, should be sufficient for most break sizes.

The physically-based major intervals for one key parameter may not be iden-tical to those for a different key parameter. An effort should be made to

+

07/12/85 27 RG EMER CORE COOLING SYSTEM

b keip as many common b3undaries as p3ssible to cinimize the number of overlapping major intervals for different key parameters. The objective is to facilitate cross comparisons of individual code accuracy statements.

The major regions for accuracy quantification for all key parameters for a large break loss-of-coolant accidents should be the " blowdown" (from transient start to accumlator initiation) and " refill /reflood" periods (from accumulator initiation to end of transient). The steady state before transient start can i

ba considered a third region. The boundary between the two major large break loss-of-coolant accident transient time intervals is taken as the time when accumulator injection begins because the accumulator flow is usually the dominant component of the total ECC injection, compared to either HPI or LPI, and usually significantly changes the governing physics of the transient. If '

desired, the period or " refill /reflood" could be further divided into an ECC bypass period and the actual refill /reflood period when the vessel level rises.

Since the times at which the bouncaries between the two majur internals will usually differ for the experiment and test, some measure of care should be taken in choosing either the test time or measurement time or some time in between as the transition point for accuracy quantification purposes. The rationale for selecting the transition times used should be clearly set forth and should be consistent. .Within each region a sufficient number of intervals should be used for comparison of code prediction and measured value and for a determination of the code uncertainty for that region.

l Major regions for accuracy quantification for most small break loss of-coolant i accident key parameters are also steady state, " blowdown" (from transient start  !

to accumulator initiation) and " refill /reflood" (from accumulator initiation to end of transient). Potential small break loss-of-coolant accident scenarios ,

vary sufficiently that a single definition of the major regions for all such '

transient types is not possible. For liquid levels, the blowdown period should usually be subdivided into an initial period when flow effects significantly perturb differential pressure (i.e., level) instrumentation and a later period when the flow has stagnated:

the boundary should be the time that the primary cooiant pumps coast down and/or degrade, t

07/12/85 28 RG EMER CORE COOLING SYSTEM

.... . UMri P Anoth2r time b3undary potentially affecting most key parameters is time to loop seal clearance. Secondary side pressure accuracy quantification should be divided into separate time intervals based on the onset of reverse heat transfer. Natural circulation flow should be quantified after the time that the primary coolant pumps coast down and/or degrade, a time boundary also used i for the liquid levels.

For all transients, within each major interval a number of equally-spaced subintervals should be defined. These are used to reduce but still accurately represent the quantity of (code measurement) difference values carried to later stages of the accuracy quantification process. Ten to fifteen such minor inter-vals in each major interval is acceptable. However, a large number of code /

data difference points should De retained so that the interval sizes and/or the particular statistical evaluation can be modified in the future,

d. Data Smoothing Key parameters, both in integral thermal hydraulic test data and best es-timate predictions of them, frequently contain large oscillations over relatively

'short periods of transient time (i.e. , " noisy" information). Also, discrete measurement times and computer time steps frequently do not coincide. Thus, it '

may be necessary to smooth one or both of the sets of data to facilitate data comparisons. The uncertainty evaluation should include any uncertainty which can be ascribed to the smoothing of the curves. If code predictions are smoothed prior to comparison with data it should be shown that oscillations which occur are a result of attempting to simulate actual physical phenomena and not due to limitations of the numerical techniques used.

e. Truncated Calculations Frequently, when best estimate computer codes are used, analysts will trun-cate the calculation prior to the termination of the transient (e.g., end of reflood) because the code end measured results have equilibrated in a fashion which demonstrates that they will not again converge. While this does not necessarily adversely affect the accuracy quantification for an individual transient in the common time period, the overall accuracy estimate being used 07/12/85 29 RG EMER CORE COOLING SYSTEM

. . UMri in combining th2 r3sults from a number of analys2s will again be skewed toward those cases where the " correct" behivior is being calculated. For quantification .

purposes, calculation should cover the entire test period, whether or not the predicted results exhibit the observed behavior,

f. Qualitative Results The determination of code uncertainty is largely a numerical exercise but it can, in some cases, mask important information about the code's predictive capabilities.

For example, if the code predicts system response well but has the event timing off significantly, the code uncertainty will be large particu-larly in regichs where the value of the key parameter is changing rapidly with respect to time.

Also, when measured and calculated data are behaving in a man-ner which causes them to change relative positions as a function of time (i.e.,

curve crossing), the uncertainty analysis may mask the fact that important pheno-menological conditions are not predicted correctly. Such qualitative findings should be clearly set forth.

g. Simplifying Assumptions and Approximations As mentioned in Paragraph A.1 of this Appendix, simplifying assumptions or approximations may be used under certain conditions in order to facilitate the best estimate calculation. Justification should be provided to demonstrate that any biases or errors introduced are conservative and the effect of such errors should be quantified. .In addition, it should be demonstrated that the l i

total contribution that these assumptions and approximations have upon the code uncertainty is sufficiently small that the best estimate calculation will real-istically predict all the important phenomena occurring during the transient.

h. Documentation The documentation provided for NRC review of the uncertainty quantifica-tion should clearly explain the methods used to arrive at code uncertainties for the various key parameters selected for the individual transie:1t types from I the various experimental comparisons used. All of the tests used should be listed and specific comparisons made should be provided. Any best estimate 07/12/85 30 RG EMER CORE COOLING SYSTEM 1

l

calculaticns which were mad 2 but were not used in the determination of the {

uncertainty, should be identified and a justification for not utilizing the a data provided.

i

3. Other sources of uncertainty When a best estimate methodology is used to predict reactor transients, l sources of uncertainty other than the limitations in the individual models and -

- numerical methods used are introduced. These contributors to the overall cod uncertainty should be considered in the uncertainty analysis. '

t

a. Experimental Uncertainty >

I i

The thermal-hydraulic behavior of a test facility is determined by the experimentalist by taking various measurements at key locations in the facility.

Limitations of the individual measurement instruments will introduce a random error in the measurement and errors associated with the calibration of the in-strument produces a systematic error.

The errors for individual instruments should be combined and an experimental uncertainty component should be included in the overall code uncertainty.

I

b. Initial and Boundary Conditions f

When a plant input model is prepared certain relationships describing the plant's behavior are defined.

These include factors such as pump performance, valve actuation, and control system functioning. Uncertainties associated with the characterization of the perfomance of such components should be accounted '

for in the uncertainty evaluation,

c. Scaling The experimental information used to determine code uncertainty will usu-ally be obtained from facilities that are much smaller than nuclear power reac-tors.

Thus uncertainties will result when the results of best estimate calcula-tions are extrapolated to the larger sca'a. These scaling uncertainties should be quantified and considered in the overall code uncertainty analysis.

07/12/85 31 RG EMER CORE COOLING SYSTEM

u ..',

d. 79,3 Variability of the results of plant transient calculatior.s can result from uncertainties associated with fuel behavior. This uncertainty includes many t effects sucts as fuel conductivity, gap width, gap conductivity, peaking factors and decay heat.

These uncertainties should be quantified and used in the deter-mination of overall code uncertainty.

07/12/85 32 RG EMER CORE COOLING SYSTEM

, . e' a -

References A.1 Los Alamos National Laboratory " TRAC-PF1/ MODI:

An Advanced Best Estimate

^

Computer Prcgram for Pressurized Water Reactor Thermal-Hydraulic Analysis," to be published.

A.2 Idaho National Engineering Laboratory " TRAC-BD1/M001:

An Advanced Best Estimate Computer Program for Boiling Water Reactor Transient Analysis,"

NUREG/CR-3633, Aprii 1984.

A.3 Idaho National Engineering Laboratory, "RELAPS/M002 Code Manual, Vol.1&2 EGG-SAAM-6377, April 1984.

A.4 Pacific Northwest Laboratory, " COBRA / TRAC-A Thermal-Hydraulics Code for Transient Analysis of Nuclear Reactor Vessels and Primary Coolant Systems," NUREG/CR-3046, Vol.'l-5, March 1983.

A.5 W. D. Beckner and N. Zuber, " Technical Basis for Revisions of ECCS Rule,"

to be published.

s A.6 k G. A. Berna et al., "FRAPCON-2: - A Computer Code for the Calculation of j

Steady State Thermal-Mechanical Behavior of Oxide Fuel Rods," Pacific l

Northwest Laboratory, NUREG/CR-1845, January 1981.

E A.7 1 American Nuclear Society Standard, ANSI /AN55.1-1979, "American National Standard for Decay Heat Power in Light Water Reactors," August 1979.

A.8 J. V. Cathcart and R. E. Pawel, et al., " Zirconium Metal-Water Oxidation Kinetics: IV Reaction Rate Studies," ORNL/NUREG-17 August 1977.

A.9 Memorandum from Rober B. Minogue to Harold R. Denton, "Research Information Letter No.128, "PWR Lower Plenum Refill Research Results,"

December 8, 1981.

i A.10 L. Biasi et al., " Studies on Burnout: Part 3," Energia Nucleare 14, 1967.

07/12/85 33 RG EMER CORE COOLING SYSTEM

DWt\

A.11 R. T. Lahey and F. Moody, "The Thermal Hydraulics of a Boiling Water Nuclear Reactor, ANS, 1977.

A.12 H. C. Yeh, C; E. Dodge, and L. E. Hochreiter, "Reflood Heat Transfer

, Correlation," Nuclear Technology, Vol. 46, 473, 1979.

A.13 G. P. Lilly, H. C. Yeh, C. E. Dodge, and S. Wong, "PWR FLECHT skewed 1 profile Low Flooding Rate Test Series Evaluation Report," WCAP-9183, i November 1977.

A.14 A. Yamanouchi, "Effect of Core Spray Cooling in Transient State After loss-of-Coolant Accident," Journal of Nuclear Science Technology, Vol. 5, No. 11, pp. 547-558, November 1968.

i I

f s

07/12/85 34 RG EMER CORE COOLING SYSTEM

/

/

- _ _ _ _ _