ML17303A340
| ML17303A340 | |
| Person / Time | |
|---|---|
| Site: | Palo Verde  |
| Issue date: | 03/06/1987 |
| From: | Haynes J ARIZONA PUBLIC SERVICE CO. (FORMERLY ARIZONA NUCLEAR |
| To: | Knighton G Office of Nuclear Reactor Regulation |
| References | |
| 161-00043-JGH-L, 161-43-JGH-L, TAC-56662, NUDOCS 8703110052 | |
| Download: ML17303A340 (22) | |
Text
I REGULATOR NFORNATION DISTRIBUTION:
TEN tRIDS)
~h ACCESSION NBR: 8703110052 DOC. DATE: 87/03/06 NOTARIZED:
NO DOCKET 4 FACIL:STN-50-528 Palo Uerde Nuclear Station>
Unit li Arizona Publi 05000528 AUTH. NANE AUTHOR AFFILIATION HAYNES> J. Q.
Av izona Nuclear Pouev Pv o Ject (formerly Av i zona Public Serv RECIP. NANE RECIPIENT AFFILIATION KNIQHTON. Q. W.
PWR Prospect Div ectov ate 7
SUBJECT:
Forwards pv oposed pv ogv'am for fuel assembly= guide tube fretting wear insp pv ogv am fov v evicts Zc appv oval bg 870015.
DIBTRIBVTION CODE:
A047D COPIEB RECEIVED: LTR 1
ENCL g BITE:
TITLE:
OR Submittal:
Inservice Inspection/Testing NOTES: Standav'di zed plant. N. Davis> NRR: 1Cg.
05000528 RECIPIENT ID CODE/NANE PWR-B ADTS PWR-B PD7 LA LICITRA>E CO< IES LTTR ENCL 1
1 0
1 1
RECIPIENT ID CODE/NANE PWR-B EB PWR-B PD7 PD 04-
'OPIES LTTR ENCL 1
5 5
INTERNAL: ACRS AEOD/PTB 01 16 10 10 1
1 1
1 1
- DN/LFMB NRR/DSRO/EIB OQC/HDS1 1
0 1
1 1
0 EXTERNAL LPDR NSIC NOTES:
03 05 1
1 1
1 NRC PDR 02.
1 1
TOTAL NUNBER OF COPIES REQUIRED:
LTTR 29 ENCL
'26
gl J q Vh, hh hh hf 'h
-Y k.
h h
hf 'i h
h h hh!!
\\ )(~h>
hg
Arizona Nuclear Power Project P.O. BOX 52034 o
PHOENIX, ARIZONA85072-2034 March 6, 1987 161-00043-JGH/LJM Director of Nuclear Reactor Regulation Attention:
Mr. G.
W. Knighton, Project Director PWR Project, Directorate ¹7 Division of Pressurized Water Reactor Licensing B
U.S. Nuclear Regulatory Commission Washington, D.C. 20555
Subject:
Palo Verde Nuclear Generating Station (PVNGS)
Unit 1
Docket No.
STN 50-528 (License NPF-41)
CEA Guide Tube Wear Program for Unit 1
File: 87-B-056-026
Dear Mr. Knighton:
Per FSAR Section 4.2.4 and SSER 2 Section 4.2.5 ANPP is required to provide the details of the Fuel Assembly Guide Tube Fretting Wear Inspection Program for Unit 1.
Attached is the proposed program for your review and approval.
Please provide your response by April 15, 1987.
H you have any questions, please contact Mr. W. F. Quinn of my staff.
Very truly yours, J.
G. Haynes Vice President Nuclear Production JGH/LJM/rw Attachment cc:
0.
M. De Michele E. E.
Van Brunt, Jr.
E.
A. Licitra (w/a)
R.
P.
Zimmerman A. C. Gehr M. Davis
/
87O 3 1 J 005 OGOO0 528 p 87O30b pDR
~DOC~
pDR 8
(
Ql.
I
~
~
ATTACHMENT CEA GUIDE TUBE WEAR INSPECTION PROGRAM GENERAL:
Measurements of C.E.A. guide tube wear will be performed on twenty fuel assemblies, four guide tubes per assembly, by an approved nuclear vendor such as Combustion Engineering, Inc. or Advanced Nuclear Fuels Corp. (formerly Exxon Nuclear).
Details of the proposed test program are as described below.
DESCRIPTION OF TEST METHODOLOGY:
Quantitative measurements of guide tube wear will be performed in the spent fuel pool using accepted nondestructive eddy-current testing techniques developed specifically for this purpose.
The vendor's database of guide tube wear measurement results from other plants and fuel of various designs will be used for comparison and interpretation of test results.
The guide tube wear eddy-current test is performed with a bobbin coil.
The coil is inserted several feet into the top of each guide tube and withdrawn at a constant speed.
Data are collected on the removal stroke.
Volumetric variations in wall thickness are detected through comparison of the guide tube data with data obtained from accurately machined and characterized standards.
Based on both the results of flow tests performed during the design process of the SYSTEM 80 fuel assemblies and on analytical design models, significant wear is not expected in guide tubes of Palo Verde-1 assemblies.
The eddy-current tests to detect wear are expected to provide confirmatory data.
Therefore, these eddy-current tests will be performed in the spent fuel pool as opposed to performing them in the reactor vessel prior to movement to the spent fuel pool.
Performing the tests in the spent fuel pool during the fuel shuffle will conserve valuable critical path time.
SAMPLE SIZE:
Arizona Nuclear Power Project (ANPP) commissioned a statistical study to determine the appropriate sample size, to be performed by Dr. Michael Driscoll Ph.D.,
a consultant in statistics and Associate Professor of Mathematics at Arizona State University.
His study, which is attached, concluded that a sample size of twenty assemblies would be appropriate.
The twenty assemblies would account for the various combinations of C.E.A. finger configuration and for the expected wear distribution range.
As a result of the above considerations, ANPP will perform measurements on twenty assemblies.
I
SELECTION OF ASSEMBLIES TO BE TESTED:
Selection of assemblies will be made so as to thoroughly sample the various C.E.A. finger configurations while concentrating on locations with the highest probability of wear.
The assemblies to be tested include those that resided under both 12 finger and 4 finger C.E.A.'s in Cycle 1.
UALITY ASSURANCE:
The testing will be performed by a fully qualified nuclear vendor.
The vendor's specific QA program for field testing services will be utilized.
REPORT OF RESULTS:
The test program will include a report of test results to be prepared following completion of the examination program.
The report will be used to demonstrate that the guide tube wear is within acceptable
- bounds, and to define whether future testing is necessary.
DSC 2/20/87
1
~
~
I A
,r 1
~
il I
P l
SAMPLIWS TO ASSESS SUIDE ~USE WEAR Report SNFD-ANPP-2, to Nr. David Car leton Ar izona Nuclear Power Project by Michael F. Driscol 1, Ph.
D.
Consultant in Statistics and Computing 38 December 1986 This report concerns issues of sample selection for assessing the extent of guide tube wear in the fuel assemblies of the Palo Verde Unit 1 reactor.
Application-specific information was obtained from Dave Car leto'n in a 4 Dec 86 meeting at ANPP and a 8 Dec 86 letter.
The statistical methodology is based on the literature cited herein.
INTRODUCTION The goal is to determine the extent of wear in control rod guide tubes.
Such wear is due to vibration of the fingers in the control element assemblies (CEA'), vibrat ion being caused by the flow of reactor coolant around and through the fuel assemblies.
Inference about the extent of guide tube wear is to be made via confidence intervals of the form PrCW iB3 l i-a, in which W
with 8
< q 1 is the q quantile (or or 188q-th percentile) in the wear distribution, B is a bounding 'value, and 1-a is confidence level associated with the inference.
The interpretation of (1) is that with confidence 1-N the q quantile in the wear distribution is less than or equal to the value b obtained for B from an actual sample.
This confidence is due to the fact that, prior to sampling, the probability is 1"a that Wq is less than or equal to B.
In order to draw such an inference, certain guide tubes need to be selected and measured "for wear.
This report addresses the questions of (1) among what guide tubes selection should be made and (2) how many guide tubes should be selected.
1.
DEFINITION OF THE SAMPLE This section considers from what guide tubes selection should bo made.
The nature of the
~oulation and the
~sam le are discussed, and the underlying assumptions are specified.
(continued)
0 n
r I
NFD-ANPP-2 (Guide Tube Wear) page 2
- 1. 1.
The Population Since the problem of guide tube wear is a question of the engineering design of the~'fuel assemblies, the population is a conceptual
- one, namely the collection of all the guide tubes that could have been or could be used in CEA-finger locations of the Unit 1 reactor.,This population will be called the population of ~n ered guide tubes.
Part of this conceptual population has real existence, namely the set of fingered guide tubes which have actually been used.
The Unit 1 reactor has 748 CEA fingers, grouped into 48 full-length 12-finger CEA',
28 full-length 4-finger CEA',
and 13 part-length 4-finger CEA'.
Physical considerations gustify the assumptions (1) that guide tube wear under part-length fingers is no greater than that under full-length fingers and (2) that the environments exper ienced by the guide tubes under the 688 full-length fingers are equivalent.
- 1. 2.
The Random Sample order to control extraneous sources of variability (such as manufacturing differences and rate of coolant flow),
the sample to be selected from the 688 full-length finger -locations is to be selected at random.
The specific fuel assemblies discharged at the end of,a given cycle are those which have had maximum use in the reactor, that is, those whose fingered guide tubes have experienced maximum wear.
Furthermore, the assemblies discharged at the end of any cycle are spread throughout the core in a radially uniform fashion, so their guide tube wear is representative of that throughout the entire cor e.
It is therefore appropriate to select the random sample from among the fingered guide tubes occurring at full-length finger locations within the discharge assemblies.
1.3.
First Use The impending assessment of guide tube wear will occur at the end of cycle one of the Unit 1
reactor.
At this
- point, 69 of the 241 fuel assemblies in the core are to be discharged.
Of these
'69 so-called "A"
assemblies, 12 hold no fingers, 13 hold 4'part-length fingers, and 45 hold 4 full-length fingers.
So the sub-population to be sampled from consists of 183 (full-length) fingered guide tubes.
Under the assumption that. the guide tubes at different finger locations experience similat environments, it is reasonable to do the sampling by fuel assembly rather than by guide tube, that is, to randomly select k
(say) of these 45 assemblies and measure the wear in the n ~ 4k fingered guide tubes they contain.
2, CHOICE OF SANPLE SIZE This section considers the question of how large a
random sample is
- needed, (continued)
t
38 Dec 86 b1FD-RNPP-2 (Guide Tube Wear) page 3
2.1.
Nonparametr ic Intervals for a Single Quantile In this approach, the "only assumption made about the 'istribution of wear is that tpe cumulative distr ibution function is continuous.
The form of the confidence interval
<1) is (2)
Pr(W
<Wn 3
1 a
q in which W
denotes the largest obser vat ion in a
random sample of n
(n) observations.
Such a
statement is sometimes called a tolerance interval rather than a confidence interval, in order to emphasize its interpretation as an interval
(-n>, W
)
- which, with probability 1-a prior to sampling,
<n) contains at least proport ion q of the populat ion of measurements rather than
'ts interpretation as an interval
(-n>>W "
3 which contains the q quantile W
(This distinction is semantic.)
The issue is to determine how large a sample must be taken to insure that with confidence at least 1-a the largest measurement in the sample is exceeds 188q percent of the measurements in the population.
It can be shown that for (2) hold with given values of q and a, the sample size n must satisfy the condition q
< a The smallest such sample size is clearly the smallest integer for which n
(log a)/(log q)
- where, as usual, log denotes logarithms (to any convenient base),
Details of the derivation of this sample size rule may be found, for example, in Conover (1988, sec.
- 3. 3).
Under the assumption that all 4 guide tubes will be measured in any fuel assembly in which one guide tube is measured, equation (3) gives (4)
, k
<log a)/(log q
)
as the number of assemblies required.
For example, 6 assemblies are needed to be 98><
sure that the
~ 988 quantile is smaller than W
~
As a fur ther iIlustrat ion of the use of <4), the values 1-a ~.988 q
k 1-a ~.958 q
k 1-a ~.975 1-a ~.998
.988 6
. 958 12
. 975 23
~ 998 58
. 988
. 958
. 975
. 998 8
o 988 15
. 958 38
. 975 75
. 998 9
18 37 92
. 988
. 9S8
. 975
. 998 11 23 46 115 might indicate that a sample of 18 assemblies is marginal and that a
sample of 28 is acceptable.
(continued)
II t
I(
P
~
38 Dec 86 MFD-ANPP-2 (Guide Tube Wear) page 4
2.2.
Nonparametric Intervals for Several Quantiles The approach given in the previous section controls the confidence level at 1-a for earth interval est imate pr oduced, that is, for the interval estimate of a single given quantile.
It does not control an overall confidence level that is, one applying to several quantiles at once as is necessary if the intent is to describe the overall behavior of the target distribution.
Such simultaneous inferences can be made by forming confidence bands.
Again, the only assumption needed about the distribution of wear is that the cumulative distribution function is continuous.
In the present
- context, the goal would be to produce a lower confidence
- band, that is, a non-negative monotone non-decreasing function L(w), with the property that Pt ( L(w) I FW(w) for all w
3 I 1-<<,
where FW is the unknown cumulative distribution function of the guide tube wear distribution~
Note that the onl~ pertinent values of w in (5) are the or dered values W
W
~
~
~
W obtained fr om a
random sample of (1) size n.
Confidence bands can be found by several methods.
One approach is based on the Kolmogor ov-Smirnov distributions and is discussed in standar d textbooks, for example, in Conover (1988, sec.
- 6. 1).
The lower confidence band is obtained as (6)
L (w)
~ minC8, S (w) -R (1-<<, n) )
in which S(w) denotes the empirical distribution function S(w)
~
CO of sample values i w>/n.
The correction term,
)<,
depends both on the chosen confidence level 1-a and on the sample size n.
Values for selected and 1
< n
<48 ar e given in Conover (1988, Table A14).
For n
> 48, use the approximation k ( l-a, n)
= c (1-a ) /C n+4 (n/18) >
in which the values of c(1-a) are
- 1. 87,
- 1. 22,
- 1. 36,
- 1. 52, and
- 1. 63 corresponding, respectively, to the values
.988,
.958,
.975,
.998, and.995 of 1-<<.
Using n ~ 4k as before, the corrections in (6) are 1-a ~.988 1-a
~a 958 1-a ~.975 1
a
~a 998 n
k 5
.232 18
. 165 15
. 135 28
. 118 5
.265 18
. 189 15
. 154 28
. 134 5
.294 18
. 218 15
. 172 28
. 149 5
.329 18
. 235 15
~ 192 28
. 167 (continued)
38 Dec 86 MFD-ANPP-2 (Guide Tube Wear) page 5
which again might indicate.that a sample of 18 assemblies is marginal and that a sample of 28 is acceptable.
r Somewhat sharper methods for setting confidence
- bands, due to Sandford (1986),
use corrections in (6) which guarantee that (5) holds but also control each of the so-called trespass probabilities Pr(:
FW(W J
)
(
L(g) 3.
(
- However, these methods require either tables of the incomplete beta function or extensive computation.
- 2. 3.
An Extr erne-Va lue Appr oach Detailed information about maximum guide tube wear can be obtained by employing results from the theory of extreme values.
The literature on this topic is extensive.
Some introductory references include David (1978, sec.
- 9. 3),
Johnson and Kotz (1978, chap.
21),
Kendal l and Stuart (1969, chap.
14),
and Oliviera (1983).
The kind of'istribution appropriate for the maximum guide tube wear depends on the behavior in the wear distribut ion.
Results in Mann and Singpurwal la (1982) show that if the so-cal led parent distribut ion is exponential,
- gamma, Weibull, or
- normal, then the maximum wear W "
in a random sample of size n will, as n approaches infinity, follow an type I, or
- Bumbe1, extr erne-value distribut ion.
The difficulty is to determine how large n must be for the asymptotic behavior to come into play.
Determination of the sample size needed cannot be readily made unless the parent wear distribution is assumed to be of a
specific parametr ic type
- and, even
- then, the desired analysis can be computat ional ly tedious or intractable.
David (1978, p.
289) reports that the tendency toward the asymptotic form is reasonably fast in the exponential case but exceedingly slow in the normal case.
(Note that these disparate results hold for parent distributions which are known to produce the same type of asymptotic extreme-value behavior.
)
Gumbel (1958, p.
'222) says that agreement with the asymptotic form in the normal case is good for n as small as
- 188, except in the tails.
Of course, the distribution of guide tube wear cannot be exactly normal, because it is bounded below.
On the other hand, it cannot be exactly exponential because, if it were, the process of guide tube wear would have no memory or, equivalently, new guide tubes would be as good as used guide tubes.
It seems reasonable in this context to suppose that the actual wear distribution lies somewhere between the exponential and normal distributions, a realm which is in a sense bridged by the gamma and Weibull distr ibutions.
(In fact, results cited in Oliviera,
- 1983, indicate that for moderately sized samples the Weibull distributions will provide a
better appr ox imat ion to the extreme value behavior than will the Gumbel distr ibut ions. )
The conclusion here must be that the sample size needed to assure applicability of the asymptotic distribution of maximum wear lies somewhere (continued)
4 C
l]
J II ij t
1 pl
,, II
NFD-ANPP-2 (Guide Tube Wear) page 6
between those required for exponential parents and those required for normal parents.
- Again, it appears that a sample of 18 assemblies (48 fingered guide tubes) would be mar'ginal and that a sample of 28 assemblies (88 tubes) would be acceptable.
(However, several such samples would be needed, in order to estimate the location and scale parameters in the limiting Gumbel distribution.)
- 2. 4.
An Extreme-Value Approach with Simulation A reliable analysis of maximum guide tube wear might be obtained from a
sample of fewer than 28 fuel assemblies by applying the methodology described in a recent report by Combustion Engineering
<1985),
as corrected by Driscoll (1985).
The thrust of this approach is to obtain a fitted distribution for guide tube wear and then use simulation techniques to simulate the behavior of maximum wear.under this distribution.
The approach can be outlined as follows.
1)
Select a random sample of k = 18 (say) fuel assemblies to obtain n =
4k measurements of guide tube wear.
2)
Perform a
Kruskal-Wallis test
<samples of size 4
from several populat ions) to veri fy the homogeneity of guide tube wear across fuel assemb1 ies.
3)
Assuming that the hypothesis of homogeneity is not rejected, pool the n
wear measurements and form a 95)( lower confidence band on Che empirical distribution function (as in section 2.2 above)
~
4)
Fit a gamma or Weibull distribution to the n pertinent points in this lower confidence
- band, using a
least-squares or method-of-moments fitting criterion, thereby obtaining a
conservative functional description of the distribution of guide tube wear.
If a good fit cannot be
- obCained, randomly select an additional k
(say) fuel assemblies to obtain additional wear measurements, and return to step (2)
~
5)
Use computer simulation techniques to repeatedly (1888
- times, say) generate samples of size 688 (or 748) from this fitted distribution and, for each such
- sample, find the sample maximum.
6)
Fit a type I extreme-value distribution to this sample of 1888 maxima and calculate representative tail quantiles therein to assess the maximum guide tube wear occurring among the fingered guide tubes.
This appr oach would obviously require intensive computation.
- But, since Combustion Engineering already has most (if not all) of the required programming in place, this may not be obgectionable.
The advantage of the appr oach is that it could reduce the number of discharge fuel assemblies
'rom which measurements of guide tube wear are needed.
'continued)
l t
NFD-ANPP-2 (Guide Tube Waar) page 7
FINAL RENARK Note that the second step of the method in section 2.4 can be applied no matter what methodology is used.
Performing this Kruskal-Wallis test would serve as a check on the assumption that the environments experienced by fingered guide tubes are independent of fuel assembly location.
REFERENCES
"'"el'" "
York: Wiley and Sons.
Rival a
on o(>'>a)o Y~de ED)ts
)
Rnd 8 Fuel Rod Qydroqen C~otent, Document Number CE NPSD-283-P9 dated 18 Jan 85, Nuclear Power
- Systems, Combust ion Engineering, Inc., Windsor, CT.
- David, H.
A.
(1978),
Order Statistics, New York> Wiley and 'Sons.
Dr(moo)i, Michael F.
((985>,
Barns tGD'Diseuse CE )(GPD-ERG-E, Col>au)tant's Report Number NFD-ANPP-1, dated 14 Feb 85,
- Tempe, AZ, Gumbo),
E. J.
()985>, Qtat let los o~ E~xemes, New York: Columbia
(>n)varsity Press.
Kendall, Naurice G.
and
- Stuart, Alan (1969), ~ ~Deed Ibe~
gf Publishing.
e
- Johnson, Normal L.
and
- Kotz, Gamuel
()978>,
C~o~)u~o
()JL((var settee Distributio s-i, New York: Wiley and Sons.
- Nann, Nancy R.
and Sing pur wa 1 1 a, Nozer D.
(1982),
"Extreme-Value K
(Vol. 2),
Kotz9
- Samuel, Johnson, Normal L.,
and
- Read, Campbell
- III, eds.,
New York: Wiley and Sons.
- Oliviera, J.
Tiago de'. (1983),
"Gumbel Distributions, "
pp.
552-558 in Normal L.,
and
- Read, Campbell III, eds.,
New York: Wiley and Sons.
- Sandford, Nartin D.
(1985),
"Nonparametr ic One-Bided Confidence Intervals for an Unknown Distribution Function Using Censored Data, "
Technometrics, 27, 41-48.
Report prepared by:
P4dd dJeC~
'(yy~+~
Nichael F. Driscoll, Ph.
D.
Consultant in Statistics and Computing
~
~
1