ML20236Q292
| ML20236Q292 | |
| Person / Time | |
|---|---|
| Issue date: | 07/13/1998 |
| From: | NRC (Affiliation Not Assigned) |
| To: | |
| Shared Package | |
| ML20236Q285 | List: |
| References | |
| NUDOCS 9807200152 | |
| Download: ML20236Q292 (12) | |
Text
l M *ou t
UNITED STATES
- s
]
NUCLEAR REGULATORY COMMISSION J
WASHINGTON, D.C. 300H001 ENCLOSURE 1 SAFETY EVALUATION BY THE OFFICE OF NUCLEAR REACTOR REGULATION RELATING TO SIEMENS POWER CORPORATION LICENSING TOPICAL REPORT ANF-1125(P). SUPPLEMENT 1. APPENDIX D.
"ANFB CRITICAL POWER CORRELATION UNCERTAINTY FOR LIMITED DATA SETS" 1 INTRODUCTION in a letter from H. Donald Curet (SPC) to U.S. NRC dated April 18,1997, Siemens Power Corporation (SPC) requested NRC review of SPC Topical Report (Reference 1), ANF-1125(P),
Supplement 1 Appendix D, "ANFB Critical Power Correlation Uncer'.ainty for Limited Data Sets."
This topical report describes a statistical method to estimate the variance of a hypothetical larger data set using the observed variance of the existing limited data set for the ATRIUM-9 fuel. This method was developed in response to a NRC potential nonconformance, developed during NRC vendor performance inspection review 9990081/97-01, regarding the sufficiency of critical heat flux (CHF) data for SPC's ATRIUM-9 product.
The NRC staff was assisted in this review by its consultant, Pacific Northwest National Laboratory (PNNL). The consultant's evaluation and findings are described in the attached PNNL technical evaluation report (TER, Enclosure 2). The staff reviewed the SPC submittal (Reference 1) and the responses (References 2, 3,4, and 5) by SPC to the staff's letter requesting additionalinformation on ANF-1125(P), Supplement 1, Appendix D.
l 2 EVALUATION When ANF-1125(P), Supplement 1, Appendix D, was submitted, it was intended for use on the ATRIUM-9B and ATRIUM 9X fuel designs. After the acquisition, review, and analysis of additional CHF data on the ATRIUM 9B fuel design, the staff determined that the methodology described in this topicslis not applicable to the ATRIUM-9B design. Thus the review and approval of this methodology is limited to the ATRIUM-9X fuel design. The consultant's TER which includes summary, background, evaluation, and technical findings, is given in. The staff reviewed the consultant's evaluatior; as documented in the TER and in general concurs with tne findings. Differences are addressed below.
Recommendation 1 was reworded slightly to simplify the statement. There is no change in the meaning.
Recommendations 2,3, and 4 do not apply since the acceptance of the Appendix D methodology is for the ATRIUM-9X fuel design only.
9007200152 990713 3
PDR TOPRP ENVEXXN Lj C
PDR L;
2 3 CONCLUSION Based on the staffs review in conjunction with our consultant's evaluation (Enclosure 2), the NRC concludes that the proposed methodology is acceptable for ATRIUM-9X fuel only, subject to the following restriction to which SPC has agreed (Reference 5):
The Additive Constant Uncertainty for ATRIUM-9X fuelis 0.0201 for all rods with local peaking of 1.22 or less and 0.0235 for all rods with local peaking of greater than 1.22.
l 1
i
3 4 REFERENCES
- 1. *ANFB Critical Power Correlation Uncertainty for Limited Data Sets," ANF-1125(P),
Supplement 1, Appendix D, April 1997.
- 2. Letter (HDC:97:082) from H. Donald Curet to U.S. NRC, Responses to Request for Additional Information for Siemens Topical Report, ANF-1125(P), Supplement 1, Appendix D, "ANFB Critical Power Correlation Uncertainty for Limited Data Sets," August 7, 1997.
- 3. Letter (HDC:97:147) from H. Donald Curet tS U.S. NRC, Respons,es to Request for Additional Information for ANFB Critical Power Correlation Uncertainty for Limited Data Sets, ANF-1125(P), Supplement 1, Appendix D, (TAC No. 98478), December 22,1997.
- 4. Letter (HDC:98:012) from H. Donald Curet to U.S. NRC, Response to Request for Additional Information for ANFB Critical Power Correlation Uncertainty for Limited Data Sets, ANF-1125(P), Supplement 1, Appendix D, (TAC No. 98478), February 19,1998.
- 5. Letter (NRC:98:031) from J. F. Mallay to U.S. NRC, Siemens Power Corporation's j
Calculation of Additive Constant Uncertainties for the Appendix D Analysis, May 15,1998.
ENCLOSURE 2 TECHNICAL EVALUATION REPORT for ANFB CRITICAL POWER CORRELATION UNCERTAINTYFOR LIMITED DATA SETS, (ANF-1125(P), Supplement 1, Appendix D)
Judith M. Cuta June 1998 Prepared for Office of Nuclear Reactor Regulation U.S. Nuclear Regulatory Commission Washington, D.C.
20555 under Contract DE-ACO6-76RLO 1830 NRC FIN 12009 Pacific Nonhwest National Laboratory Richland, WA 99352 l
l
SUMht\\RY Approval of this methodology is recommended for application to SPC's ATRIUM-9X fuel only (i.e.,9x9 fuel with 3x3 water channel and bimetallic spacers). The methodology should not be granted a generic approval for application to limited data sets. Each specific application of the methodology to a given data set must be reviewed to determine (a) that the methodology is applicable to that data set and correlation, and (b) that the methodology is appropriately applied to the particular data set. Therefore, application of this methodology to the new data set being obtained for ATRIUM-9B will require review to determine that the additive constant uncertainty (ACU) for application to analysis of ATRIUM-9B fuel is obtained in an appropriate manner. Until such review is complete, the interim value of 0.029 for the ACU should be used for analysis of ATRIUM-9B fuel.
In the case of the ATRIUM-9X fuel, the final value of the additive constant uncertainty (ACU) is increased from the value of 0.0195 reported in Appendix D, (see Section 4.5, p. 4-6 of Ref.
- 1) to a value of 0.0235 for application to fuel bundles where the local radial peaking exceeds 1.22. For applications to fuel bundles where the local radial peaking is 1.22 or less, the ACU is increased from 0.0195 to 0.0201. The following table summarizes the recommendations for treatment of the ACU for 9x9 ATRIUM fuel, based on the evaluations of this review.
Table 1: Additive Constant Uncertainty Values for ATRIUM-9X and -9B Fuel ANFB-ratio to correlation convert additive form (identified E
ECPR constant by fuel type) upper uncertainty uncertainty overall ECPR uncertainty bound to ACU (ACU)
ATRIUM-9X with local radial peaking (0.00020 + 0.02822)as = 0.03155 0.0387 1.93 0.0201 s1.22 ATRIUM-9X with local radial peaking (0.00020 + 0.000377 + 0.02822)*5 0.0-4548 1.93 0.0235
>1.22
= 0.03704 ATRIUM-9B NA NA NA 0.0290 When the ACU values from this table are used in the analysis of ATRIUM-9X fuel in a specific core design, it will be necessary to show that the correct ACU is applied to each fuel l
bundle.
2-
BACKGROUND CHF and CPR correlations are used throughout the nuclear power industry to determine the thermal limits of a particular fuel design in a given reactor core. These correlations are ad hoc models that are statistically fit to data using least squares regression analysis over an appropriate data base. In most cases, the correlations are formulated in such a way that it is most efficient to derive a new set of empirical coefficients for each new fuel design. In some cases, additional components,are included in the model to account for new features in the fuel design that significantly affect thennal performance. The ANFB correlation, however, is constructed with the assumption that the basic functional relationships between boiling transition and the nominal bundle operating conditions of flow rate, pressure, and inlet subcooling are relatively unvarying between different fuel designs. Local effects due to such factors as flow channel geometry, radial power distribution, axial power distribution, and grid spacer design are correlated by means of a local energy balance term FEFF and a set of empirical additive constants on this term.
One significant advantage of this approach is that purely from the practical standpoint of data manipulation, it makes it easier to derive a new form of the ANFB correlation for a new fuel design. The values determined for the flow and pressure coefficients for the ' base' ANFB correlation and the formula for determining the FEFF term are not changed in the process of fitting the correlation to a new data set. All of the differences between the new data set and the j
original database are assumed to be captured by the process of fitting the correlation to the new j
data set to determine'the set of additive constants for that fuel design.
l This approach requires some extra care in the evaluation of the correlation's fit to the data base for a particular fuel design. It is necessary to check the validity of the underlying assumptions about the character of the different coefficients with each new data set that the correlation is l
applied to. The separate treatment of coefficients related to nominal operating conditions, l
wherein they are assumed to be constant and invariant with changes in geometry and power distributions, means that the correlation must be examined closely for biases with such parameters as flow, pressure, and inlet enthalpy over the full range of application each time the correlation is applied to a new data set representing a new fuel design. In addition, the method of determining the additive constants deliberately induces conservative biases into the fit of the correlation to the specific data set. This is not inherently bad, but it confounds the l
determination of the true variance of the fit to the data set, which is the significant parameter in determining the uncertainty of the correlation's predictions.
l As a result, the development of new sets of additive constants for the ANFB methodology requires data that span the full range of application of the correlation for the new fuel design.
This is necessary in order to assess the goodness of fit and to ensure that the overall variance of the fit to that specific data set correctly represents the uncertainty in the correlation's predictions when applied to the given fuel design. '
For SPC's 9x9 ATRIUM fuel designs' it has been determined 2 that the range of conditions tested is insufficient tojustify the additive constant uncertainty (ACU) values used in the Safety Limit determinations for cores with this fuel. In response to this, SPC has submitted ANF-1125(P), Supplement 1, Appendix D (Reference 1), which describe.s a statistical method to estimate the variance of a hypothetical larger data set, using the observed variance of the existing limited data set for the 9x9 ATRIUM fuel and supporting analyses with GE 7x7 ATLAS test data.
EVALUATIONS After a thorough technical review of Appendix D, the SPC responses to Requests for Additional Information (see References 2, 3,4, and 5), and the main statistical references used by SPC (see References 6 and 7), it has been determined that the proposed methodology is in general adequate for the purpose of obtaining an estimated additive constant uncertainty (ACU) for the ATRIUM-9X fuel design. However, there are a few modifications that are required in order to make the methodology fully acceptable for Safety Limit determmations.
The method used in the analysis to convert the ECPR uncertainty to the additive constant uncertainty (ACU) is not acceptable. The ECPR uncertainty is converted to the additive constant uncertainty (ACU) by dividing the overall ECPR uncertainty by the average of the ratios of the ECPR uncertainty and the FEFF uncertainty for the bundles in the data set. This method of computing the ratio ignores the variation in the ratio of ECPR uncertainty and FEFF i
uncertainty for each bundle, which represents a real source of uncertainty in the ACU. A limiting value such as a 2-0 limit or 95/95 tolerance limit on the distribution of the ratios i
should be used. For this application, the 2-0 limit of the ratio is approximately 1.93. This value should be used instead of 1.99.
l The variance due to high radial power peaking is inappropriately characterized as negligible l
and ignored in the analysis. The effect of high radial power peaking can be evaluated in the same manner as was presented for different non-uniform axial peaking (see p. 4-4,4-5 of Ref.1) using the 7x7 GE ATLAS data set. This analysis consists of a one-way ar.alysis of variance on local power peaking effects. The variance is given by the formula, (n-1)[f,.i,,2_1(f,.y,)2 )
g I
2 s,
n 1
l This design has a large square water channel that replaces a 3x3 rod array at the center of the bundle.
)
The specific designs considered here are ATRIUM-9X, which has bimetallic spacers, and ATRIUM.8B, which has ULTRA. FLOW mixing vane grids.
2 See NRC Inspection 999081/97-01. t
.~
~
where n = number of groups (in this case, n=2) l x, = mean ECPR of the ith group l
l For this analysis, the apprq riate grouping of the data is by high and low radial peaking. Low radial peaking is defined by radial peaking factors of 1.25 or lower, and high radial peaking is defined as 1.40 or greater. This grouping divides the ATLAS data base of 2544 points into two subse'.s, as shown in Table 2. The variance between these two groups, as computed using Eq. (1), is a measure of the effect oflocal radial peaking on the fit of the ANFB correlation to this data set.
Table 2: Esu.nated Variance using Different Groupings of the 7x7 ATLAS DATA 2
estimated variance (s ) due ATLAS data set data grouping to local power peaking complete 7x7 ATLAS data n=2; data grouped into low set,2544 data points local peaking and high local peaking sets;
- 1) low peaking (s1.25) 0.000377 2329 data points
- 2) high peaking (21.40) 515 data points In the absence of high radial peaking data in 9x9 ATRIUM fuel bundle geometries, it has been assumed that the variance exhibited by the ANFB correlation with the ATLAS data base is representative of the correlation's behavior with local radial peaking in other fuel geometries.
This variance should be included in the calculation of the overall ECPR uncertainty for ATRIUM-9X fuel bundles w..h radial peaking above 1.22. Following the formula described in Section 4.5 (see p. 4-5, Ref.1), this means that for ATRIUM-9X fuel bundles with high peaking (i.e., > 1.22), the overall ECPR uncertainty is given by overall ECPR uncertainty = (0.00020 + 0.000377 + 0.02822)os
= 0.03704 For ATRIUM 9X fuel bundles with low peaking (i.e., s1.22), the overall ECPR uncertainty is given by overall ECPR uncertainty = (0.00020 + 0.02822)os
= 0.03155
.S.
1 i
The appropriate value of the overall ECPR uncertainty, (which depends on the local radial power of the bundle) should be used in the Ch;-Square analysis (see p. 4-6, Ref.1), rather than the value of 0.0316 (see Section 4.5, p. 4-6, Ref.1).
The methodology described in Appendix D contains many assumptions, approximations, and specialized modifications to standard analysis of variance techniques. In particular, the method used to determine equivalent sample size for determining the "truebetween mean variance" with unequal sample sizes is not correct for the one-way analysis of variance used in this methodology. However, since most of the samples are indeed of equal size for this particular application, it is reasonable to suppose that the approximation is adequate. In addition, the values of the variance due to the effects of radial and axial power differences are estimated using information from the correlation's fit to the 7x7 ATLAS test bundles of GE fuel. This requires the assumption that the vanance in the ATLAS data for these factors is the same as (or perhaps greater than) that which would be obtained in tests with high radial peaking and different non-uniform axial power distributions for the ATRIUM fuel design.
This is a reasonable engineering assumption, based on the benavior generally observed in rod arrays with different axial power shapes. Also, the variance of any correlation for data not included in its data base is usually larger than the variance for data that is part of the correlation's data base. However, since SPC is in the process of obtaining additional data for the ATRIUM-9B fuel geometry, including tests with different non-uniform axial power distributions, it would be reasonable to check thir. assumption with respect to axial power distribution. Such a check can be performed by evaluating the variance of the correlation for ATRIUM-9B data with different non-uniform axial power distributions. If the value obtained i
is smaller than 0.00020 (which is the value obtained in the analysis of variance of the 7x7 l
ATLAS data), the assumption would be shown to be conservative. If the testing were to l
discover that the ANFB correlation is more sensitive to axial power shape for the ATRIUM-9B fuel geometry, then the actual value obtained in the analysis of variance should be used in place of the 0.0002 value to obtain the overall ECPR uncertainty.
Alternatively, with a more complete data set in hand, SPC might chose to re-derive the additive constants for the ANFB correlation when applied to ATRIUM-9B fuel. In that case, it would be necessary to re-evaluate the need for the analysis described in Appendix D. If it were determined that the expanded data set still does not provide full coverage of the data space, the applicability of the methodology described in Appendix D could be evaluated for the new data set. If the method is appropriate, this approach could be used to determine an estimate of the additive constant uncertainty (ACU) for the ANFB correlation when used with that fuel design.
It must be noted, however, that the methodology described in Appendix D is a specialized analysis tailored to the specific shortcomings of the 9x9 ATRIUM data sets. The methodology should not be given a generic approval for application to other fuel designs with limited data sets. Each such application must be reviewed on a case by case basis, to determine that all l.
relevant sources of variance are considered and treated appropriately in light of the deficiencies of the panicular limited data set.
RECOMMENDATIONS 1.
The proposed methodology described in Appendix D (with additional explanatory material provided in Ref. 2, 3, 4, and 5) can be approved for Safety Limit calculations for ATRIUM-9X, with the following modifications; a) A limiting value such as a 2-0 limit or 95/95 tolerance limit on the distribution of the ratios of the ECPR uncertainty and the FEFF uncenainty for the bc,xiles in the dsta set should be used to conven the ECPR uncenainty to the additive constant uncertainty (ACU). For this data set, the 2-0 limit of the ratio is approximately 1.93. This value should be used instead of 1.99.
b) The variance due to high radial power peaking, as estimated from the ATLAS da'a, cannot be neglected. Ut.ing the complete ATLAS data set, it can be shown that this variance is on the order of 0.000377. This value thould be included in the overall ECPR uncenainty calculation for fuel bundles with local radial peaking greater than 1.22.
Therefore, the ACU for ATRIUM-9X fuel can be calculated using the foliawing formula; l
- l ##""
ACU=
1.93 i
X2 bound of [(0.00020 + F,(0.000377) + 0.0282 )"]
2 1.93 1
where F, = 1.0 for bundles with local radial peaking > 1.22 F, = 0.0 for bundles with local radial peaking s 1.22 2.
For ATRIUM-9B fuel, the interim value of 0.0290 should be used for the Additive Constant Uncenainty (ACU) while additional new data is being obtained for this fuel design. A new topical repon describing the approach to be used to determine the ACU i
for the new constants for ATRIUM-9B fuel must be submitted, reviewed, and approved l
in order to obtain an ACU value based on the fit of the new constants to the new data base.
3.
The Appendix D methodology should not be given a generic approval for application to
. fuel designs with limited data sets. Each such application must be reviewed on a case by case basis, to determine that all relevant sources of variance are considered and treated appropriately in light of the deficiencies of the particular limited data set.
4.
In this methodology, it is assumed that the variance due to the effects of radial and axial power differences in 7x7 ATLAS test bundles of GE fuel is the same as (or perhaps greater than) that which would be obtained in similar tests for 9x9 ATRIUM test bundles. This assumption should be evaluated by one-way analysis of variance of the additional data currenity being obtained by SPC in ATRIUM-9B fuel geometry test bundles with different non-uniform axial power distributions. If the actual variance with axia! power shape is greater than the value of 0.00020 estimated from the 7x7 ATLAS data, SPC must evaluate the effect this would have on the overall ECPR uncertainty for ATRIUM-9X and ATRIUM-9B fuel in applications of the Appendix D methodology.
-S-
l l
REFERENCES 1.
- ANFB Critical Power Correlation Uncertainty for Limited Data Sets," ANF-1125-(P:
)
Supplement 1, Appendix D, April 1997.
j l
2.
August 7,1997, HDC:97:082, Response to Request for Additional Information for ANFB Critical Power Correlation Uncertainty for Limited Data Sets," ANF-1125-(P),
Supplement 1, Appendix D, (TAC No. 98478).
3.
December 22,1997, HDC:97:147, Response to Request for Additional Information for J
ANFB Critical Power Correlation Uncertainty for Limited Data Sets," ANF-1125-(P),
Supplement 1, Appendix D,(TAC No. 98478).
4.
February 19,1998, HDC:98:012, Response to Request for Additional Information for ANFB Critical Power Correlation Uncertainty for Limited Data Sets," ANF-1125-(P),
Supplement 1, Appendix D,(TAC No. 98478).
5.
May 15,1998; NRC:98:031, "Siemens Power Corporation's Calculation of Additive Constant Uncertainties for the Appendix D Analysis."
6.
NUREG/CR-4604, PNL-5849; " Statistical Methods for Nuclear Material Management",
W.M. Bowen and C.A. Bennett, December 1988.
7.
Mary G. Natrella, Experirantal Statistics, NBS Handbook 91, reprint of the Experimental Statistics Portion of the AMC Handbook, issued August 1,1963 and reprinted October 1966, US Government Printing Office, Washin%, D.C.
l l
i 1
1 i
--