ML19309B989
| ML19309B989 | |
| Person / Time | |
|---|---|
| Site: | Zion File:ZionSolutions icon.png |
| Issue date: | 03/31/1980 |
| From: | Office of Nuclear Reactor Regulation |
| To: | |
| Shared Package | |
| ML19309B983 | List: |
| References | |
| NUDOCS 8004080011 | |
| Download: ML19309B989 (32) | |
Text
5 e
O SAFETY EVALUATION OF THE REQUEST FOR REMOVAL OF CONSERVATISM IN THE WESTINGHOUSE PAD COMPUTER CODE 4
)
l tgnREGu
- ey, z%
g aiS u
.s.az
.'g C
k'..,
+...e$
j MARCH 1980 CORE PERFORMANCE BRANCH UNITED STATES NUCLEAR REGULATORY COMMISSION nAsHINGTON. D. C. 30868 g
Q A'
s004080
TABLE OF C0r:TEf1TS
- PAGE
' 1.
Introduction.........................
1-2.
Request for~ Reduction in Conservatism.............
2 3.
Conservatism. in -the Westinghouse PAD Computer Code......
3 3.1;- Design vs. As-Fabricated Conditions...........
5 3.2. Fuel Model Conservatism...,. '............
6 3.3 Audit of the Revised Analysis..............
10
- 4. -
Staff Criterion for Margin of Conservatism..........
15
. 4.1 Basis for Margin of Conservatism.............
15 4.2 The Staff Criterion...................
18 5.
Application of the Criterion.................
19 5.1 The Westinghouse PAD Computer Code 22' 5.2 The NRC Fuel Performance Codes..............
23 5.2.1 FRAP-S3......................
23 5.2.2 FRAPCON-1.....................
24 5.2.3 GAPCON......................
25 5.3 The State-of-the-Art...................
26 6.
Conclusions..........................
27 7.
References..........................
28 9
9 h
s e
!"E7v rYF.'?T!0?! 0F THE REQ"EST FOR REMOVAL OF CONSERVATISM IN THE WESTINGHOUSE PAD COMPUTER CODE 1.-
Introduction The thennal conditions within the fuel of a light water reactor during its normal lifetime must be described in the safety analysis for each reactor. The fuel temperatures are used as initial conditions in describing the response of the reactor to a number of hypothetical transients and accidents, such as the loss-of-coolant accident (LOCA).
Coninonwealth Edison Company (Com-Ed), the owner of Zion Station 1 and 2, has requested a license amendment for these two plants to increase the allowable LOCA peaking factor limit from'l.86 to 1.93.
This change is based on reanalysis of the loss-of-coolant accident wherein reduced initial fuel temperatures are assumed. The reduced fuel temperature are a result of the removal of some of the margin of conservatism in the fuel performance code used in the analysis. Both Coninonwealth Edison and th,e code developer, Westinghouse Electric Corporation, believe that t'he remaining conservative feature's of the code are adequate for the safety analysis.
A review of the proposed revisions to the fuel code, PAD-3.3, and our c,
-evaluation of these changes, are presented in the following sections.
The discussion will consist of a technical review of the submittal, com-parision of the Westinghouse code with a traditional staff audit code, and the development and application of a new criterion for margin of conservatism in codes of this type. All of these ;nethods lead us to conclude that the remaining conservative features of the code are adequate for safety analysis.
2-2.
kequest tor Keduction in conservatism On March 22, 1979, Commonwealth Edison Company requested (Ref.1) a license amendment to increase the allowable LOCA peaking factor limit.
This request was based on revised analysis of the emergency-core-cooling-system (ECCS) to meet 10 CFR 50.46 (Ref. 2) requirements. The LOCA peaking factor, also known as the limiting heat flux hot channel factor or the limiting F, is defined as the maximum local heat flux on the g
surface of a fuel rod divided by the core average fuel rod heat flux.
The naximum allowable local heat flux is calculated in the plant safety analysis, usually for the loss-of-coolant accident.
For most reactors, the LOCA peaking factor limits the operational flexibility or power maneuvering capability, but not the total power generating capability, of the plant.
In the case of Zion Station Unit 1, there is some evidence (Refs. 3-6) that the LOCA peaking factor may also limit the power production capability of the plant.
The potential peaking factor limitation caused Com-Ed to reanalyze the loss-of-coolant accident for the Zion facility. An increase in the allowa,ble LOCA peaking factor from 1.86, to 1.93 was projected based on reduced fuel temperatures. The reduced temperatures were calculated with a modified version of Westinghouse PAD-3.3 code (Ref. 7).
We reviewed the Com-Ed submittal and requested (Refs. 8-9) additional information with regard to the proposed changes. Com-Ed responded to these requests with additional information (Refs.10-11), which has also been reviewed. The details of the Com-Ed proposal are discussed in the next section.
g e.
w.
, 3.
Conservatisrn.in the Ws.stinghouse FAD Computer Code The Westinghouse PAD computer code iteratively calculates the interrelated effects of temperature, pressure, cladding elastic and plastic deformation, fission gas release, and fuel densification and swelling as a function of time and power density. The most recent version of the code, PAD-3.3, was described by Westinghouse in a Licensing Topical Report (Ref.
7). This report was previously reviewed and approved by the NRC staff
- (Ref. 12 ).
As part of the emergency-core-cooling-system evaluation requirements (Ref.13) for plant safety analysis, the PAD code utilizes a number of conservatisms-in the prediction of fuel temperatures. These include conservative inputs to the code, conservatisms within the code itself, and conservative margin applied to the code output. Some of these con-servatisms, such as the 102% of maximum allowable power that is input to the code, are specifically required by the regulations. Other conservatisms, such as the conservative margin applied to the code output, are not specifically required by law. These additional conservatisms were submitted by Westinghouse as part of earlier safety analysis reports or were required by the NRC staff during the review process.
It is the second category, those conservatisms which are not specifically required by law, which is the subject of this report.
The derivation and application of conservative margin applied to the PAD code output have been described previously (Ref.14). The margin is due to uncertainties in the fuel temperature predictions due to manufacturing variations. The parameters considered include:
a) cladding inside diameter.
b) pellet outside diameter c) pellet' density d) pellet sintering temperature
-s e
4_
Varictior.; in the first two parameters affect the calculated fuel-to '
cladding gap size. Variations in the last two parameters affect.the densification behavior of the fuel.
J For safety analyses, nominal design values of the above parameters are used as input to the PAD code and allowance for manufacturing variations are then added. The allowance for each of these four parameters is determined by using a bounding value for each quantity. The allowance is simply the difference between the nominal'and the bounded input code prediction. As an example, the PAD code will predict higher temperatures if an upper bound cladding inside diamete'r is used as input rather than the nominal design value of this parameter. The difference between the two predictions, in degrees Fahrenheit, is the allowance for manufacturing variatioon in cladding inside diameter. The bounding value for each input parameter is derived on a normally applied 95% probability basis at'a 95% confidence level. Each allowance is calculated at the time in life when fuel temperatures are maximum and at a power level of 15 kW/ft, the approximate LOCA limit.
The allowance calculated for each of the four input parameters is statistically combined with the others to form the total fabrication uncertainty. To the total fabrication uncertainty, a second, so-called model uncertainty, of 65 F.is added.
It is the Westinghouse position that this additional 65 F margin was added to ensure that the best-estimate model predictions would bound most of the experimentally measured fuel temperature data. The best estimate model is the same PAD-3.3 code using nominal input. values' and no explicit internal or external code conserva tisms. At the time the 65'F margin was accepted, it was the staff's opinion that this margin was used to account for uncertainties not explicitly considered elsewhere. Both Commonwealth Edison and Westinghouse believe that the 65 F margin is already considered else-
- where,in the analysis _and that the remaining conservative features of the code are adequate.
(
~
p q.
. In at" t'-
t:. ri.,a:st ts :rci. tt: 65'F model uncertainty, C:,m.sr.. - i t'.
Edison has requested that the use of as-fabricated, rather than as-desiqned, fuel conditions be allowed in the safety analysis. Generally, this would permit the use of nominal values of cladding inside diameter, pellet outside diameter, pellet density, pellet sintering temperature and their respective uncertainty allowances based only on the fuel supplied to each Zion Unit, rather than the entire Westinghouse product line.
In practice, the dimensional parameters (i.e., cladding I.D. and fuel 0 D.) and their uncertainties do not change significantly. As a result, the use of as-fabricated fuel conditions affects only the pellet density, pellet sintering temperature and their respective un-certainties.
A complicating feature of the request is a change in the current product line analysis by Westinghouse.
For the current Westinghouse fuel design:,
the fuel is sintered in such a manner that the statistical lower bound of the actual sintering temperature is always above the sintering temperiture used as input to the PAD code for safety analyses. This means the code input value is lower than virtually all of the sintering temperatures used ip manufacturing the fuel. As a result, the code predicts more densif'ication than is expected, but the' allowance for uncertainty in sintering temperature becomes zero.
The Com-Ed request for the use of as-fabricated values would cause the analysis to revert back to its original form. Namely, the use of a nominal sintering temperature and a non-zero allowance for uncertainty i
l in this tempe,ature. Because these values would be based only on fuel in each Zion Unit, the result is. a higher sintering temperature, an almost unchanged total fabrication uncertainty, and a reduction in average fuel temperature predictions of approximately 20*F.
3.1 Design vs. As-Fabricated Conditions i
i The request fsr the use of as-fabricated, rather than as-designed
(
fuel parameters is fundamentally sound. The LOCA' analysis should, I
g, w
e-g
- ( -
. ir. ;.rar:1, apply to a sp :ific plant rather tnan a generic design.
We therefore agree with the proposed use of plant-specific input conditions.
We have, however, examined the proposed change to~ determine whether the approach is indeed applicable to Zion and is statistically valid. Commonwealth Edison has stated that they will use generic fuel parameters (i.e., sintering temperature and pellet density) which are bounding to the actual values' determined for each Zion reload core but not for the entire Westinghouse product line.
These parameters will be selected on the basis of previous fuel reload data and anticipated future reload data. However, the fuel for each future reload will be measured to ensure that it meets the acceptance criteria for all fuel batches used in that reload. The statistics for the reload region are based en the complete set of data for all batches. At each Zion reload review, Commonwealth Edison will verify that the specific fuel parameters are bounded by the Zion generic fuel parameters.
Westinghouse, the fuel supplier for Zion Units 1 and 2, has also described (Ref.11) the statistical methods on which the change will be based. For Zion Unit 1, Region 7, the data include over 28,000 density measurements and the sintering temperature for each sintering boat used in manufacturing the fuel for the region (3.3 million pellets). The large number of observations used in the process is well in excess of the levels required for proper statis-tical analysis. We conclude that the request to use as-fabricated, rather than as-designed, fuel parameters is acceptable.
3.2 Fuel Model Conservatism As discussed previously, safetf analyses with the PAD code currently applies a conservative margin to the fuel temperature predictions of the code. *This margin is composed of a component due to
. fabrication uncertainties (cladding and peilet diameters, pellet density and sintering temperature) and the 65'F component termed the r.cdel uncertainty. Westinghouse has stated that the final component was added to ensure that the best estimate code predictions would bound most of the experimentally measured fuel temperature data. Because the evaiuation model or conservative version' of PAD, ratber than the best estimate version, is used in safety analyses, it is not necessary (from the regulatory standpoint) that the best estimate version bound any data.
It is'our position, rather, that whichever version of PAD is used in plant safety analyses should conservatively predict fuel temperatures. This requires an explicit or implicit consideration of uncertainties, including uncertainties in the models used.
Both Comonwealth Edison and Westinghouse believe that the 65 F margin is already considered elsewhere in the analysis. They further state that the remaining conservative features of the code are adequate for ECCS analysis. The bases for this statement are:
13 The "best estimate" version of PAD bounds the majority of the experimental fuel temperature data considered.
2.
The evaluation model or safety analysis version of PAD always predicts fuel temperatures greater than or equal to those I
predicted by the 'best estimate" version of the code.
3.
The -limiting time in life for ECCS analyses is such that the conservative version of PAD always predicts fuel temperatures l
greater than those predicted by the best estimate version.
I l
4.
' The overall conservatism in the calculation of fuel temperature, i
the conservative applicatfon of thos,e temperatures in the LOCA analysis, and the conservatism associated with the overall l
LLOCA/ECOS evaluation warrant the elimination, of the 65*F fuel temperature model uncertainty.
k O -
. t c +ill discuss each' of these bases individually.
1.
Westinghouse.has submitted the results of a number of com-parisons between the best-estimate predictions of the PAD-3.3 code and experimentally determined fuel centerline temperature data. The data were taken from the Halden Heavy Boiling Water Reactor (AE-318, HPR-80 and IFA-226) and the Materials Testing Reactor (WAPD-228). Westinghouse selected these data because they represent helium-filled cylin'drical fuel rods near beginning-of-life with densities and gap sizes typical for the standard product line. We agree that these data are representative of the current Westinghouse product. However, it is clear that the "best-estimate" version of PAD-3.3 is not a best-estimate code at all, but a conservative one.
In other words, even with nominal input values and the removal of explicitly con-servative models within the code, PAD-3.3 tends to overpredict the experimental data.
2.
When the PAD code is used in safety analysis, certain evaluation model options are activated.. These include the fuel densifi-cation, gap conductance (gap closure) and cladding creep model s.
Because all of these processes are time-dependent, the difference between the conservative and best-estimate options is zero at time zero. At all non-zero exposures, the evaluation model options do result in higher fuel temperature predictions as stated by Westinghouse.
3.
Appendix K of 10 CFR 50 states that "the steady-state temperature distribution and stored energy in the fuel before the hypothetical accident shall be calculated for the _ burn-up that yields the
-highest calculated cla'dding temperature (cr optionally, the 4
I l
3..f.w. calcuiated stored energy.)"
(Ref.13)
In the hesting-house safety analysis, the limiting burnup occurs shortly ef ter beginning-of-life, at the point of maximum fuel densifi-cation.
If the best-estimate version of PAD is used, however, the burnup at which maximum burnup occurs is different than that cciculated with the evaluation model options. For'the best-estimate code, this burnup is at, rather than shortly af ter, beginning-of-life. We believe that the margin of conservatism between the best-estimate and evaluation model versions of PAD is misleading when measured at any specific burnup. -We conclude that the difference between the maximum temperatures predicted by each version of the code is a more appropriate basis of comparison. The margin calculated in this fashion is significant, but less than that assumed by Westinghouse.
4.
The overall conservatisn in the calculation of fuel temperatures and the overall conservatism in the LOCA analysis have not been rigorously demonstrated by Commonwealth Edison or Westinghouse.
Only individual details of the analysis, such as the impact of LOCA peaking factor uncertainties on fuel temperatures, have been described. The impact of the proposed modification on the overall analyses has not been addressed. As a result, we are unable to consider the overall conservatism in the LOCA analysis as a basis for the removal of the 65 F model uncertainty in PAD-3.3.. This conclusion is discussed further in Section 4.1 of this evaluation.
We conclude that Westinghouse has shown the evaluation model or safety analysis version of PAD-3.3 to be' conservative in calculating steady-state fuel temperatures for LOCA analysis..This alone is not sufficient to demonstrate that-the degree of conservatism, when used in LOCA analysis, is sufficient to warrant elimination of the 65 F model uncertainty. We will discuss this further in Section 4.
..- 4 3.3 ~ - t.udit cf the Revised Analysis The historical method of regulatory review of fuel codes may be divided into three areas: '(1) establishing the technical validity of th.e methods and supporting data described by the applicant, (2) verifying the bxistence of conservatism in the analyses, and (3) determining the degree of conservatism relative to traditionally accepted audit codes..The technical validity of the methods used in the Westinghouse PAD-3.3' code was established in Sections 3.1 and 3.2 of this report as well as earlier staff
~
evaluations (e.g., Ref.12). The fact that' the code is indeed conservative without the 65 F model uncertainty was also established in Section 3.2 of this report. The remaining item, a comparison of the PAD-3.3 predictions with a traditionally accepted audit code, is presented in this section of the report.
The GAPCON-THERHAL-2 code (Refs.15 and 16) is one of a series of fuel thermal performance codes developed by Battelle Pacific Northwest Laboratories for the Core Performance Branch of the Nuclear Regulatory Commission. Since 1975, it has been used by the staff to audit vendor fuel code submittals, including the Westinghouse PAD-3.3 code. GAPCON-2 predicts fuel temperatures, fuel densification and swelling, fission gap. release and other fuel conditions as a function of time and power in a fashion much like that of PAD-3.3.
GAPCON-2 also has a number of conservative model options similar to PAD.
r In order to audit the proposed modifications to the PAD-3.3 code, a current version of the GAPCON-2 code was used to calculate (Ref.17) volume average fuel temperatures as a function of burnup for the Westinghouse 15x15-fuel design used in Zion Unit 1.
The results of these calculations are shown in Figure 1.
The two lines represent the best-estimate and conservative predictions for GAPCON-2.
Figure 2 of Ref.11 is similar, showing (1) the best estimate PAD-3.3 prediction, (2) the conservative PAD-3.3 prediction (3) the conservative
s9 e
WESTINGHOUSE 15X15 FUEL DESIGN m
i m
CC C
NRC EVALUATION MODEL 8< 8 cc mn. g 2
m P
i w 8-
?
O \\
x W
4 $_
NRC BEST-ESTIMATE MODEL t
ANb
~
"8 s
O 1000 2000 3970 l
ROD AVERAGE BURNUP, MWD /MTU Figure 1 i
i
o
-12 ~
ne-
..n _ g t,, n g.....,,
=.g 7 fe 7 fg ication uncertair.tir,
and (4) the conservative PAD-3.3 prediction plus the margin for fabricatior uncertainties and the 65'F model uncertainty.
The fourth Westinghouse curve (the highest):is representative of the.-
~ current LOCA ar,alysis. The third Westinghouse curve is the version requested by Comonwealth Edison. All calculations (both GAPCON and PAD) were made at a local: linear power of 15 kW/ft, approxi-mately the LOCA limit.
t A number of conclusions can be drawn in examining these two figures.
First,. the current (highest) Westinghouse curve.is higher than all
- of the other. curves, including the conservative GAPCON-2 prediction traditionally accepted for audit of LOCA analyses.
It is not until much higher burnups are reached that the. Westinghouse prediction is exceeded by the conservative version of GAPCON-2.
The change at higher exposures is due to the effect of cladding creepdown, which is considered in PAD-3.3, but not in GAPCON-2. The second ob-servation to be made from these figures is that all of the PAD predictions are higher than the best-estimate version of GAPCON-2.
This confirms our earlier conclusion that even the "best estimate" v'ersion of PAD-3.3 is not best-est'imate at all, but conservative.
There is additional evidence, not presented here, that even the best-estimate version of CAPCON-2 is conservative with respect to the data.
In order to obtain a more representative comparison between these two figures, a second set of results were generated in which cladding creepdown (which is already considered in PAD) was included in the-GAPCON-2 predictions. The creepdown values used by GAPCON-2 were i
generated with the Zircaloy creep model from a second code called BUC KLE' (Ref. 18). The results are shown in Figure 2.
The revised results are very similar to those shown previously except both GAPCON calculations exhibit a'significant decrease in fuel tem-peratures as 3 function of burnup. The conservative versions of L
both GAPCON-2'and PAD predict rising fuel tempera'tures from l
e.
e
.'+t-..w,
,--.er---t-
<-w,-.e
- w-v te--'
ev-
WESTINGHOUSE 15X15 FUEL DESIGN w
rdh-aD 8
q NRC EVALUATION MODEL WITH CREEPDOWN cr:8-Nci,iG N
2 N
E-m 8_
9 eM 4
cr:
N y l-4 J
NRC BEST-ESTIMATE MODEL WITH CREEPDOWN m
l3
~
m$
W i
i O
1000 2000 30r0 ROD AVERAGE BURNUP, MWD /MTU Figure 2 i
.- beninnbt-of-life to a low burnup peak then falling temperatures.
~
It should be noted that the conservative GAPCON-2 and PAD predicted rz,.iau:.a occur'at diffbrent burnups, but the maximum value calculated by the proposed version of PAD is not unreasonable compared to 'the conservative version of GAPCON It may also be noted the best-estimate versions of'both GAPCON-2 and PAD-3.3 predict high tem-peratures.at beginning-of-life and monotonically decreasing tem-peratures thereafter. For the burnup range considered, the best-estimate version of.GAPCON was continually and substantially over-predicted by the best-estimate version of PAD.
From our audit calculations of the PAD-3.3 code, we observe a similar, but not identical, behavior between this code and GAPCON-2. We also note that the proposed modifications to PAD result in peak volume average fuel temperatures in reasonable agreement with that tradi-tionally accepted for LOCA analyses with a conservative version of the NRC audit code, GAPCON-2.
8
+
t M
a 9
O s
m e
. 4.
St " W':-ir-f a "- -? c' Cc,:cr.ztism The Appendix K requirements for fuel thermal performance codes state that:
"The steady-state. temperature distribution and stored energy in the fuel before the hypothetical accident shall be cal-culated for the burn-up that yields the highest calculated cladding temperature (or, optionally, the highest calculated stored energy.) To accomplish this, the thermal conductivity of the U09 shall be evaluated as a fun'ction of burn-up and temperatuPe, taking into consideration differences in initial density, and the thennal conductance of the gap between the U0 a bud.n nd the cladding shall be evaluated as a function of the up, taking into consideration fuel densification and expansion, the composition and pressure of the gases within the fuel rod, the initial cold gap dimension with its toler-ances, and cladding creep." (Ref.13)
There is no explicit requirement within this section of the Code for conservatism in the fuel performanc,e codes.
4.1 Basis for Margin of Conservatism Although there is no explicit requirement for conservatism in the calculation of the initial stored energy of the fuel, the Commission has expressed an opinion on the subject.
"The assumption of 102% of maximum power, highest allowed peaking factor, and highest estimated thermal resistance between the 00 and the cladding provides a calculated 7
stored heat that is possible but unlikely to occur at the time of a hypothetical accident. While not necessarily a margin over the extreme condition, it represents at least an assumption that an accident happens at a time which is not typical." (Ref.19)
This opinion establishes the requirement for conservatism in the calculation of initial fuel temperatures for Appendix K calculations.
However, the degree of conservatism for this heat source was never l
established.
It is possible that even a best-estimate fuel code
.I. wuvio predict conservative fuel temperatures in the LOCA analyses because of conservatisms imbedded elsewhere in the calculation.
For the number of vendor fuel codes (including the Westinghouse
. PAD-3.3 code) that have previously been approved by the staff, this is probably the case. They also exhibit various degrees of con-servatism by themselves, depending on the vendor and the type of calculation performed. Similar behavior is exhibited by the fuel performance codes utilized by the NRC staff. These are discussed in Sections 3.3 and 5.2 of this report.'
It may be noted, however, that a staff opinion has been developed for another heat source, the energy due to the decay of fission products.
"A best judgment evaluaticn of these factors leads to the conclusion that a suitable probability level is 95%.... A change to 997 ar 99.9% would increase these margins but not substantialif (i.e., not produce a fundamental change in the niture of the margins).
This level is viewed as the intent of the Appendix K rule development." (Ref. 20) and further that "An additional factor to be considered is the inter-action with criteria of other energy sources such as stored energy. Logically they too should be developed with the same uncertainty probability levels as used for decay heat." (Ref. 20)
There are other. examples of the application of a 95% probability level in the calculation of heat sources and other portions of the LOCA analysis. The choice of probability level appears to be more traditional than analytical.
O I
s
. 'here is no rigoron uplicit basis for a 957 level, however it appears to be a conservative level and is compatible with other acceptance levels.
The uncer-tainty and small variation analysis is inbedded in a large LOCA matrix where, for example, F nominal seldom n
if ever occurs and other elements of required LOCA energy sources are undoubtedly conservative by some large but as yet undecided upon amount. Studies of power distributions for Westinghouse reactors as a function of reactor operation modes have indicated that F limit levels are at least 15% (and usually over 20%) greater than steady state operation nominal peaking factors, and F extremes of allowable load n
follow transients are Peached, if at all, well less thsn 5% of-the time during a cycle for any presently envisioned operation." (Ref. 21)
These two staff opinions suggest the acceptability, without a rigorously derived basis, of the probability and confidence levels proposed. Such levels are usually submitted in safety analysis reports and judged acceptable by the staff. We are aware of no submittals in which a basis for the 95% probability level has been established by the industry.
It is recognized, however, that the establishment of such a basis should involve a statistical analysis of the entire LOCA problem. Such a study does not, as yet, exist.
We are also aware of a step in the direction of determining the overall Appendix K conservatism. Westinghouse has proposed (Ref.
- 22) a statistical combination of the uncertainties in the LOCA heat sources. The proposal was not accepted by the staff (Refs. 23 and 24). An appeal by Westinghouse (Ref. 25) resulted in a second rejection by the staff (Ref. 26) on the basis that modification to explicitly-required conservatisms (such as is the case for decay heat) in Appendix K analyses should be implemented by a change in the regulation itself. We conclude that a reduction in the margin of conservatism in Appendix K stored energy analysis cannot, at this time, be based on conservatisms in other portions of the LOCA analysis. Because there is no explicit requirement for margin of conservatism in the stored energy analysis itself., we also conclude
. that the conservative margin of this heat source nay be established through the review, rather than rulemaking, process.
Indeed, this has teen the practice within the NRC in the past.
4.2 The Staff Criterion In order to develop a more uniform review of stored energy codes, we intend to use the following criterion for these models:
Assuming best-estinate input conditions, an acceptable fuel performance model shall yield a required out-put parameter such that the predicted value bounds
. a large proportion of the experimental values for this parameter.
This means that an acceptable fuel performance code, given-best-estimate input values, will, at a high probability level, correctly predict the peak fuel temperatures, fuel stored energy, fuel-cladding ' gap conductance or other parameter required as input to subsequent LOCA analysis codes. We believe an appropriately high p,robability level is 0.95 or 0.95/,0.95 where a confidence level is required.
e s
L e
-~ '
~19-i 5.
Annlication of'the Criterion L
The_ principal required output parameter from the Westinghouse PAD-3.3 code lis volume-average fuel temperature. Higher volume-average' fuel temperatures are conservative for LOCA analysis. Therefore, to; meet the proposed acceptance criterion, the PAD-3.3 code should show thel ability to overpredict volume-average fuel temperatures 95% of the time at a 95%
confidence level based on experimental data.
This criterion should b; established with experimental data prototypic of the Westinghouse product line and, where possible, taken near LOCA conditions.
1 i
As discussed in Section 3.2 of this report, Westinghouse believes the experimental fuel temperature data shown in Ref.11 to be representative i:
of their standard product line. These-data were taken at linear power levels of up to 15 kW/ft, which is approximately the LOCA limit. A limitation of these data, however, is that they are based on fuel centerline thermocouple measurements. Therefore, the data are an indication of fuel centerline temperature rather than volume average fuel temperature.
We ard not aware of any experimental data which directly measure in-reactor' volume average fuel temperatures.
It is possible, however, to relate fuel centerline and volume average temperatures analytically.
This is shown in Figure 3, where best-est'imate BOL fuel'centerl'ine and volume average fuel temperature predictions from GAPCON-2 are $hown as a function of linear power. This figure indicates that the fuel l centerline i
temperature rises much more rapidly than volume average temperature as a function of power. This is an expected result because fuel sufface temperatures remain relatively close to the coolant temperatury whereas the fuel centerline temperature rises. Thevolumeaveragetemherature may be approximated by the average of the fuel surface and centerline temperatures. Figure 4 shows the same prediction replotted with fuel volume
' average temperatures expressed as a function of fuel centerline temperature.
We.will use this f,igure to relate uncertainty in centerline temperature to uncertainty in volume' average temperature.
i i
e I
l l
i n
L l.
' WESTINGHOUSE 15X15 FUEL DESIGN 4
f
% a-4 i
id D
CENTERLINE I
~
i W
~
Qy O
a Mb
?
VOLUME AVERAGE g
W D
4 N
.A~~
O A
o i
r l
0 5
10 15.
20 LOCAL LINEAR POWER, KW/FT i
Figure 3
4 WESTINGHOUSE 15X15 FUEL DESIGN w h-i 1
MARGIN N
Qh-N Nd E-m (b
m*
M i
4 o
i
' a 8~
M D
i 4
i j
O j
0 500 1000 1500 2000 2500 3000 3500 4000 4500 i
FUEL CENTERLINE TEMPERATURE, F i
i Figure 4 f
It sh:;dif c'1be n'oted that the PAD-3.3 predictions of the experimental data.uti' ia w...ir.al input values. This is in keeping with the criterion.
While we recognize (Ref. 27) the conservatism which may exist in the input values to PAD-3.3 when used in LOCA analyses, such uncertainties are difficult to quantify from experimental data. As discussed in Section 4 of this report, we also believe it to be inappropriate to base
.the reduction of conservatism in one segment of the LOCA analysis on
-i possibly excessive conservatism in another. Such a change is more appropriate to the rulemaking process in which the conservative margin in the overall problem may be examined.
For'the purpose of this review,-
we shall assume that the PAD-3.3 code is provided nominal input conditions as part of the LOCA analysis.
5.1 The Westinghouse PAD Computer Code A statistical analysis was performed on the measured and predicted fuel centerline temperature data shown in Figure 2 of Reference 11.
We determined the mean and standard deviation of the difference (e.g. measured temperature minus predicted temperature) assuming the distribution to be normal. As the proposed criterion requires the PAD-3.3 code to conservatively predict volume average fuel temperature with a 95% p'robability at a 95% confidence level, the
~marain to be added to the best-estimate prediction should be 1.64 times the standard deviation plus the best-estimate code bias (data mean), if any.
- To relate this uncertainty in fuel centerline temperature to volume average fuel temperature, we refer to Figure 3.
At a LOCA limit of
'15 kW/ft, we find best-estimate fuel volume average and fuel center-line temperature of 2200*F and 3600*F respectively. Adding the margin due to uncertainty in fuel centerline temperature to the best-estimate centerline temperature yields a conservative prediction of centerline temperature. From Figure 4, we find the' corresponding best-estimate, conservative and equivalent margin values for volume average temperature. Using this process, we conclude that the PAD-3.3 code meets the proposed criterion based on the experimental data comparison supplied by Westinghouse.
e 5.2 The ':??..Fucl Performance Codes
-In order to check the validity of our conclusions regarding the-overall uncertainty in the PAD-3.3 code, we reviewed a number of other fuel performance. codes and Ltheir predictive uncertainties.
We have attempted to show that the predictive uncertainty in the PAD-3.3 not only meets the proposed criterion, but is also repre-
. sentative.of similar fuel performance codes.
5.2.1 FRAP-S3 The FRAP-S3 code (Ref. 28) was <feveloped by Idaho National Engineering Laboratory for the thermal and mechanical analysis' of light water reactor fuel rods. The code considers the effects of fuel and cladding deformation, temperature dis-tribution, internal gas pressure, and material properties like PAD and GAPCON.
FRAP-S3 was developed by Idaho National Engineering Laboratory for NRC's Office of Reactor Safety Research.
It is a representative example of a state-of-the-art fuel perfonnance code.
The FRAP-S3 verification repo'rt (Ref. 29), presents predicted versus measured fuel centerline temperatures based on thermo-couple measurements from approximately 100 rods, representing over 800. data points. All fuel rods used by Westinghouse, except.-WAPD-228 rods 22-3 and 22-4 were included in this study.
The standard error between measured and predicted fuel center-line temperature was stated to be 356 F and 457 F for un-pressurized and pressurized rods respectively. Assuming the
. standard deviation is independent of fuel centerline temperature (as was assumed in the study), this would result in a maximum uncertainty in fuel centerline temperature of 1.64 x 457 F = 750 F.
This is a.95/.95 statistical tolerance interval.
l To relata this uncertainty in fuel centerline temperature to volume average fuel temperature, we again re'fer to Figure 3.
L Adding 750*F margin to the best estimate centerline temperature
. yields a conservative centerline temperature prediction of 4350'F..From Figure 4, we find that this 750 F margin on centerline temperature-is approximately equal to a margin of 273'F on volume average fuel temperature. We conclude that FRAP-S3 volume average fuel temperature predictions have a maximum uncertainty of 273 F at a.95/.95 tolerance level.
There is evidence (Ref. 30) that these values would drop considerably if the data base were restricted to helium pres-surized rods near beginning-of-life with typical densities and gap sizes as proposed'by Westinghe'use. We have not considered that possibility in this evaluation.
5.2.2 FRAPCON-1 The FRAPCON-1 code (Ref. 31) is a more recent version of the FRAP-S3 code discussed previously. This computer program is the most recent of the fuel performance codes developed for the NRC. As is the case for its predecessor, FRAPCON-1 is intended to calculate the effects of power and burnup on fuel behavior under normal operating conditions.
The FRAPCON-1 verification report (Ref. 32) presents the results of predicted versus measured fuel centerline temperatures for approximately the same number (93 rods /740 data points) of fuel centerline thermocouple measurements as FRAP-S3, The standard oeviation between measured and predicted values is 306 F for unpressurized rods and 529 F for the pressurized rod data. The latter value is larger than that calculated for FRAP-S3 but no explanation for the regression in predictive ability is presented in the report.
The FRAPCON-1 assessment report again assumes that the standard deviation is constant for the range of centerline temperatures considered. Using the same process described for FRAP-53, we calculated maximum FRAPCON-1 predictive uncertainties of 868'F for fuel centerline temperature and 319'F for volume average fuel temperature at a 95/95 tolerance limit.
6
. E.'
N C0" The GAPCON series of computer codes, which are also utilized by the NRC staff, have not been subjected to the same verifi-cation process used for FRAP-S3 and FRAPCON-1. All of the GAPCON series codes have been verified with experimental data but the measured and predicted values have not been statis-tically analyzed. However, the developers of GAPCON, Battelle Pacific Northwest Laboratories, have attempted to establish the predictive uncertainty in these codes from first principles (Refs. 33 and 34).
s A recent investigation by Cunningham et al. (Ref 35) deter-mined the effect of input and'model uncertainties on fuel temperature and stored energy calculations. The study identified analytical models necessary for calculating stored energy and then utilized both the method of linear propagation and Monte Carlo technique to determine prediction uncertainties.
Re-sults were generated for a typical BWR fuel rod, but the study is also applicable to PWR fuel designs. The authors estimate the maximum uncertainties for fuel centerline temperature at a linear power of 500 W/cm (15.2 kW/ft) to be 15.5% for the Monte Carlo technique and 18.2% for the linear propagation method.
These figures are given at a 3a (99.9%) confidence level, but will be assumed.95/.95 tolerance intervals in this report.
Using these figures, we concluded that the expected uncertainty in predicting the fuel centerline temperature of a PWR rod operating at 15 kW/ft to be 558 F and 655 F by two first-principles methods. These values correspond to 206 F and 239'F uncertainties on the volume average fuel temperatures.
s O
!~
i
.. 5.3 T!
St:te of the Art We have exa,ined a sample of state-of-the-art fuel' performance n
codes.-for an indication of the expected uncertainty.in predicting volume average fuel temperatures at approximately 15 kW/ft. This-sample included two similar data-prediction studies (Refs. 29 and
- 32) and two similar first principles methods (Ref. 35). The maximum uncertainties assumed in this sample are summarized below:
4
' Maximum uncertainty in volume average Study fuel temperature at 15 kW/ft Data-prediction (Ref. 29) 273 F Data-prediction (Ref. 32) 319'F Linear propagation (Ref. 35) 206 F Monte Carlo (Ref. 35) 239'F Average 259 F To determine the state-of-the-art uncertainty in volume average fuel temperature, we have taken the average of these values. The value obtained is 259 F above the data mean. The Westinghouse margin for vplume average fuel temperature is also the difference between the evaluation model prediction and the data mean. This value is the maximum volume average temperature predicted by the evaluation model version of the code minus the maximum volume average tem-perature predicted by the best estimate version plus the bias in the best estimate code, if any. We have determined this value and conclude that the PAD-3.3 meets the proposed criterion without the use of the 65*F model uncertainty and that the remaining margin of conservatism is similar to the expected uncertainty in other state-of-the-art fuel codes.
I r
9-
. 6.
- :. A ss :...s
'lle have examined the proposed revisions:to the Westinghouse fuel code,
. PAD-3.3, as described by the code developer and Commonwealth Edison Company.
These changes consist of the-use of as-fabricated, rather than as-designed, values of fuel. density, fuel sintering temperature a'nd their associated tolerances, and the deletion of the 65'F model uncertainty from the Westinghouse fuel thermal performance analysis. Based upon our
-technical review of the submittal, comparison of the Westinghouse code predictions with a traditional staff audit code, and the development and application.of a new criterion for. margin of conservatism in codes of this-type, we conclude that these changes are acceptable. This acceptance is limited to the current version of the Westinghouse PAD-3.3 code as approved by the staff (Ref.14) for application in LOCA analyses, but acceptance is' not limited to Zion.
l l
s e,
M;
.: 7.
References 1.
' C. Reed (Com-Ed)-letter to H. R. Denton (NRC) on " Zion Station Units 1 and'2 Proposed Change to Facility Operating License Nos.
DPR-39 and DPR-48," dated March 22, 1979.
2.
Title 10, Code of Federal Regulations,'Part 50, Section 46,.
." Acceptance Criteria for Emergency Core Cooling Systems for Light Water Nuclear Power Reactors," January.1979.
3.
C. Reed (Com-Ed) letter to H. R. Denton (NRC) on " Zion Station Unit 1 Base Load Operation Under Revised ECCS Evaluation" dated
. January 24, 1979.
4.
C. Reed (Com-Ed) letter to H. R. Denton (NRC) on " Zion Station Unit 1 Additional Information for Base Load Operation Under
. Revised ECCS Evaluation" dated Janaury 25, 1979.
5.
C. Reed (Com-Ed) letter to H. R. Denton (NRC) on " Zion Station Units 1 and 2 Revised ECCS Evaluation" dated January 29, 1979.
6.
C. Reed (Com-Ed) letter to D. G. Eisenhut (NRC) on " Zion Station Units 1 and 2 LOCA Reanalysis Review" dated July 16, 1979.
7.
J. V. Miller, Ed., " Improved Analytical Models Used in Westinghouse Fuel Rod Design Computations," Westinghouse Electric Corporation Report WCAP-8720, October 1976 (proprietary) and WCAP-8785, October 1,976 (non-proprietary).
8.
.P. S. Check (NRC) memorandum to A. Schwencer (NRC) on " Questions
.Concerning Removal of Conservatism from the Westinghouse PAD Code" dated April 18, 1979.
9.
P. S. Check (NRC) memorandum to A. Schwencer (NRC) on " Zion Request for Technical Specification Change to F Based on Conservatism in 9
the PAD Computer Code" dated August 30, 1979. Enclosure reprinted as Appendix A to Ref 11.
10.
W. F. Naughton (Com-Ed) letter to H. R. Denton (NRC) on " Zion Station Units 1 and 2 Additional Information for Increase in F Peaking Factor" dated May 3,1979.
q 11.
W. F. Naughton (Com-Ed) letter to H. R. Denton (NRC) on " Zion Station Units 1 and 2 Additional Infonnation for Increase in F Peaking Factor, NRC-Docket Nos. 50-295 and 50-304" dated n
January 25,'1980 (proprietary):'
12.
J. F. Stolz (NRC) letter to T. M. Anderson (W) on " Safety Evalu-ation of WCAPt8720" dated February 9,1979. '-
l l
e
a
- 13. Title 10, Code o_f Federal Regulations, Part 50, Appendix K, "ECCS Evaluation Models" January 1979.
14.
R.- Salvatori.(W) letter NS-RS-133 to D. B. Vassallo (AEC) with Supplemental Information on Fuel Densification dated February 5, 1974.
Also Appendix B.1 of " Fuel Densification Experimental Results and Model for Reactor Application," J. M. Hellman, Ed.,
Westinghouse Electric Corporation Report WCAP-8218-P-A, March 1975 (proprietary) and WCAP-8219-A, liarch 1975 (non-proprietary).
15.
C. E. Beyer, C. R. Hann, D. D. Lanning, F. E..Panisko,.and L. J.
Parchen, "GAPCON-THERitAL-2: A Computer' Program for Calculating the Thermal Behavior of an Oxide Fuel Rod," Battelle Pacific Northwest Laboratory Report BNWL-1898, November 1975.
16.
C. E. Beyer, C. R. Hann, D. D. Lanning, F. E. Panisko, and L. J.
Parchen, " User's Guide for GAPCON-THERMAL-2: A Computer Program for Calculating the Thennal Behavior of an Oxide Fuel Rod,"
Battelle Pacific Northwest Laboratories Report BNWL-1897, November 1975.
17.
K. Kniel (NRC) memorandum to S. Fabic (NRC) on "GAPCON Input Parameters for the WRAP Program" dated September-27,1979.
18.
P. J. Pankaskie, "An Analytical Computer Code for Calculating Creep Buckling of an Initially Oval Tube," Battelle Pacific Northwest Laboratories Report BNWL-1784, May 1974.
- 19. Opinion of the Comission, "In the Matter of Rulemaking Hearing, Acceptance Criteria for Emergency Core Cooling Systems for Light Water-Cooled-Nuclear Power Reactors," USAEC. Docket No. RM-50-1, December.28, 1973.
20.
H. J. Richings (NRC) memorandum to P.- S. Check (NRC) on " Potential Changes in CPB Position on Decay Heat for LOCA" dated July 14, 1976.
21.
H. Richings (NRC) memorandum to P. S. Check (NRC) on " Policy Questions Arising from W Maxi Convolution" dated April 6, 1978 (Proprietary).
~~
22.
C. C. Little, S. D. Kopelic and H. Chelemer, " Consideration of Uncertainties in the Specification of Core Hot Channel Factor
' Limits," Westinghouse Electric Corporation Report WCAP-9180, September 1977.
.23.
D. F. Ross, Jr. (NRC) memorandum to R. J. Mattson (NRC) on "A Proposed Change to LOCA Peaking Factor Uncertainty Require-
-ments" dated July 24, 1978.
e
.24.
R. J. Mattson (NRC) letter to T. Anderson (W) dated August 9, 1978.
- 25..T.! M. Anderson (W) letter NS-TMA-1929 to R. J. Mattson (NRC) dated September II,1978.
26.
R. J. Mattson (NRC) letter to T. Anderson:(W) dated October 26, 1978.
27.
H. Richings-(NRC) memorandum to P. S. Check (NRC) on "Some Notes on PWR (W) Power Distribution Probabilities for LOCA Probabilistic Analyses
- dated July 5,1977.
28.
J. A.
Dearfen,
.G. A. Eerna, M. P. Bohn, J. D. Kerrigan and D. R.
Coleman, "FRAP-S3: A Computer Code-for the Steady-State Analysis of 0xide Fuel Rods, Volume 1, FRAP-S3 Analytical Models and Input Manual," EG&G Idaho, Inc. Report TFBP-TR-164, March 1978.
29.
D. R. Coleman, E. T. Laats and N. R. Scofield, "FRAP-S3: A Computer Code for the Steady-State Analysis of Fuel Rods, Volume 2, Model Verification Report, EG&G Idaho, Inc. Report TFBP-TR-228, I
August,1977.
30.
J. D. Kerrigan and D. R. Coleman, " Standard Design Analysis - A Statistical Detemination of Corewide Initial Accident Conditions -
Fuel Stored Energy Results," Table X, EG&G Idaho, Inc. Report CAAP-TR-034, December 1978.
1 l
31.
G'. A. Berna, M. P. Bohn, D. R. Coleman and D. D. Lanning, "FRAPCON-1:
A Computer Code for the Steady-State Analysis of 0xide Fuel Rods,"
EG&G Idaho, Inc. Report CDAP-TR-78-032, August 1978.
32.
E. T. Laats, G. B. Peeler and N. S. Scofield, " Independent Assessment of the Steady State Fuel Rod Analysis Code FRAPCON-1," EG&G Idaho, j
Inc. Report CAAP-TR-050, May 1979.
j 33.
D. D. Lanning, C. R. Hann and E. S. Gilbert, " Statistical Analysis of Gap Conductance Data for Reactor Fuel Rods Containing 00 Pellets,"
7 Battelle Pacific Northwest Laboratories Report BNWL-1832, August 1974.
34.
C. R. Hann, D. D. Lanning, R. K. Marshall, A. R. Olsen and R. E.
Williford, "A Method for Detemining the Uncertainty of Gap Con-ductance Deduced from Measured Fuel Centerline Temperatures,"
Battelle Pacific Northwest Laboratories Report BNWL-2091, February 1977.
35.
M. F. Cunningham, D. D. Lanning, A. R. Olsen, R. E. Williford and C. R. Hann, " Stored Energy Calculation: The State of the Art,"
Battelle Pacific Northwest Laboratories Report PNL-2581, May 1978.
4
,-,.n
-