ML20078G835
| ML20078G835 | |
| Person / Time | |
|---|---|
| Site: | Byron  |
| Issue date: | 01/25/1995 |
| From: | AFFILIATION NOT ASSIGNED |
| To: | |
| Shared Package | |
| ML19311B718 | List: |
| References | |
| 6424-RP-3580, NUDOCS 9502030290 | |
| Download: ML20078G835 (63) | |
Text
,
~
s..
6424/RP-3580 Probability of Detection by Bobbin Inspection i
P.S. Jackson Tetra Engineering Group,Inc.
Prepared for:
Commonwealth Edison Co.
Duke Power Co.
Electric Power Research Institute January 25,1995 jsA'A888eZi8a86 PDR
6424/RP 3580 ACKNOWLEDGMENTS A number of individuals have contributed to the preparation of this report:
J. Blomgren M. Sears Commonwealth Edison Co.
D. Mayes Duke Power Co.
D. Harris ANACAPA Sciences G. Henry EPRI NDE Center M. Behravesh j
D. Steininger C. Welty i
EPRI l
S. Brown Packer Engineering 1
l A TETRA ENGINEERING GROUP, INC.
iii
6424/RP 3580 L Overview Reliable detection of significant steam generator tube degradation is maintained at high levels to ensure that tubes will provide the required structural integrity for normal operation and postulated accident conditions. A significant indication for outside diameter stress corrosion cracking at tube support plates (ODSCC at TSPs) can be defined as the bobbin coil voltage corresponding to a degraded condition which, due to growth in the course of the next operating cycle, may result in a challenge to the structural integrity of the tube under postulated accident conditions. Therefore, the probability of detecting such indications is of significant interest to utilities and to the NRC as regulators.
Probability of Detection (POD) is defined in this report as an attribute which measures the capability of the eddy current inspection analysis and equipment.
The purpose of this report is to disclose the results of an extensive evaluation of field inspection data which establish the voltage dependence of POD for ODSCC at TSPs. This POD Study applied an approach which differs from classical efforts to characterize POD from first principles (i.e. from destructive examination of pulled tube specimens.) Removal and testing oflarge numbers of tubes from operating steam generators is not practical;it would require a very high cost (upwards of $0.5M/ tube) and would impact in a detrimental manner the reliability and operability of the steam generators.
As an alternative, previously inspected tube-tube support plate intersections were used to prepare a large data base of inspection results for use in establishing a measure of the generic capability of bobbin inspection.
Tne capability for bobbin inspection of potentially degraded tubes is achieved by conformance to industry standards for preparing calibration specimens and by implementation of standardized inspection guidelines These inspection guidelines are utilized by the field inspection teams. Primary and secondary analysts are responsible in the field exams for proper analysis and interpretation of the eddy current data. Their responsibilities include the selection and interpretation of specific multi-frequency mixes to suppress support structure signals, the use of expert judgment in the placement of vector balls to establish the Lissajous pattern and the identification of any indications which may imply degradation of structural integrity. When bobbin indications are detected, confirmatory inspection by rotating pancake coil probes (RPC) are often performed.
ATETRA ENGINEERING GROUP, INC.
Page1
i 6424/RP 3580 Limited efforts to assess the reliability of eddy current inspection in the early 1980's with the technology of that period yielded inconclusive results. Since that time, improvements in inspection guidelines, eddy current technology and analyst training have enhanced the reliability of eddy current inspection. More recently, individual utilities and industry-sponsored programs have been performed to establish the reliability of current technology and inspection guidelines.
A comprehensive study was recently completed by the Electric Power Research Institute on behalf of Commonwealth Edison Co and Duke Power Co. to document the capability of eddy current analysts to detect degradation based on bobbin coil inspection. This " POD Study" involved the development of a significant and large body of field data which were used to prepare an extensive collection of signals from 3/4" tubes with varying extents of outside diameter stress corrosion cracking at drilled hole tube support plate locations.
These bobbin data were obtained from recent steam generator inspection campaigns at the Braidwood Unit 1, Byron Unit 1 and Catawba Unit 1 nuclear plants.
This reoort presents an analysis of the POD Study data and establishes a basis for j
voltage-dependent detection capability for ODSCC at TSPs. The POD Study was not intended to establish POD in the classical sense; rather the objective was to pros ide a measure of the capability of the system which gives a good indication of the reliability of detection of ODSCC at TSPs.
i it will be seen that the dependence on voltage implies that higher voltage indications (i.e. those which exceed the Interim Plugging Criteria (IPC) of 1.0 bobbin volts for 3/4" tubes) are called with almost 100% probability; certainly at a level which exceeds the 60% detection capability currently required by the NRC in IPC submittals. It will also be seen that dual-analyst teams further enhance bobbin detection capability for ODSCC at TSPs to the level where the probability of calling an indication which exceeds the 1.0 volt IPC is >96% and approaches 100% for indications which exceed 2.0 volts.
Appendices to this report discuss the related subjects of repeatability of voltage sizing (that is, voltage sizing uncertainty) and alternate measures of inspection capability based on prior cycle detection.
A TETRA ENGINEERING GROUP. INC.
Page 2
6424/RP 3580
- 2. Detection Capability Assumed by NRC A basic tenet of alternate tube repair criteria for ODSCC at TSPs is that the voltage of an eddy current indication provides a direct measure of tube integrity, in the context of tube burst and/or leakage. This has been confirmed experimentally for both 3/4" and 7/8" tube using >10u laboratory and pulled tube intersections which were subsequently burst and leak tested.
The development of alternate tube repair as a method of steam generator degradation-specific management (SGDSM), is significant in that tube integrity is inferred directly using a parameter that is measured by non-destructive inspection of steam generator tubes; i.e. the bobbin voltage for a tube-tube support plate location with ODSCC.
Bobbin signal voltage provides an integrated measure of tube integrity which incorporates the effects of depth, length, and ligament lengths within the coil field of view. This contrasts with conventional tube repair criteria where depth is the sole basis for repair decisions. Since depth by itself does not provide a direct measure of tube integrity for ODSCC, this has led to the conservative repair of many thousands of tubes throughout the industry.
Current industry practice in the application of IPC ODSCC at TSPs is to scale the frequency distribution of voltages for bobbin indications by a factor, p, which represents the probability that an indication is missed at an inspection.
Specifically, concern is for unidentified, degraded tubes which can leak or rupture under postulated accident conditions.
Historical measures of NDE capability typically were evaluated using POD and sizing statistics which were based on maximum depth (determined from metallography) as the independent variable. Industry tube pull data, acquired under the current generic industry guidelines, have been presented to the NRC Staff during numerous meetings since the time of the SGGP program and have demonstrated an eddy current detection capabilig well in excess of 60% An example is Figure 2-1 which illustrates bobbin detection capability based on destructive samples from pulled tubes with ODSCC at TSPs.
A TETRA ENGINEERING GROUP, INC.
Page 3
'V a
6424/RP 3580 1
0.9
[
Y 0.8
.T
_3 4
] 0.7
[*
h E^
j 0.6 7
[
[
.9
-M Q
e 0.5 e
w w
m e
y 0.4
,4 42 4
9 9
L,
[ 0.3
- f
.(
G i
Q 1}
0.2
_ ~--
2
,,W d
- ~
0.1
-- f l
f hi
~
}
0 20 30 40 50 60 70 80 90 100 Maximum Depth from Metallography (%)
Figure 2-1. Industry Pulled Tube Data Basefor CnSCC at TSPs Several concerns with these data have been identified by NRC, includint;:
- 1. randomness of selected tubes; only those tubes with confirmed indications i
have been pull-d,
- 2. analysts were forewarned that tubes were scheduled for removal and so, may have applied atypical analysis efforts,
- 3. results did not address analyst variability (human factor concerns)
To date, these data have not been accepted by the NRC Staff as defining the inspection capability for ODSCC at TSPs. Instead, the NRC Staff has required that industry assume a value of p=60% in recent IPC submittals and the draft Generic Letter. This value constitutes an average value to use to postulate the number of undetected flaws in accident leakage calculations.
Rnerence 1, " Voltage-Based Interim Plugging Criteria for Steam Generator Tubes" (NUREG-1477), cites this value for the range of crack depths from 20% to 100% through-wall (TW). As a matter of clarification, a 60% POD implies that 40% of all ODSCC at TSP cracks will remain undetected at each inspection independent of the magnitude of the bobbin voltage for each crack.
Reference 2, " Steam Generator Tube Integrity Program / Steam Generator Group Project - Final Project Summary Report" (NUREG/CR-5117) has been cited as the basis for the assumed value of p. It documents the NRC SGTI/SGGP program which consisted of a " mini" round-robin using artificial samples.
ATETRA ENGINEERING GROUP, INC.
Page 4
6424/RP 3580 These were the only data obtained by the NRC which are applicable to ODSCC at TSPs. A small and statistically insignificant number of artificial samples were fabricated and used for round robin studies to provide information on the reliability of eddy cunent techniques. The objective of the NRC effort was to attempt to detect and size stress corrosion cracks "... under simulated service conditions."
It was necessary to use artificially prepared degraded tubes because of sample inadequacies from the Surry steam generator; the major testbed for the SGGP program. The test set consisted of 15 tubes with laboratory produced stress corrosion cracking, along with one tube which contained a TW stress corrosion crack that was leak-tested previously, one blank tube and an ASME flat-bottomed hole stanard.
These specimens were assembled into a bundle and sent to four firms for examination. Three of the firms were identified as routinely < conducting inservice inspections of steam generators with the caveat ",, the results of this round robin should provide an estimate of field inspections to detect and size stress corrosion cracks."
Sixteen cracks were sectioned with depths ranging from approximately 25% to 100% TW; 12 cracks were greater than 40% TW. Five teams used bobbin coil and nine teams used alternate techniques such as RPC and array probes to inspect i
these artificial samples. Results varied from 5/16 to 11/16 detected. Of particular interest are the concluding remarks in the discussion of the round j
robin results:
" As a final note, the reader should be cautioned that this round robin was obviously not
{
designed to be a definitive study, but rather to indicate trends, Too few specimens were employed to permit firm conclusions about the capability of any given team to detect and size stress corrosion cracks, or to allow team-to-team comparisons."
It is not clear based on this statement how the NRC Staff has concluded any value of detection capability from this program, owing to the limited number of samples which were available and the fact that the technology and procedures employed are now more than 10 years out of date. It must be emphasized that the data of the SGGP program were obtained without the use of inspection cuidelines and by inspection teams who lacked formal training.
j Neither of these data sets can be considered as defining eddy current capability under the SGDSM approach for ODSCC at TSPs. Under SGDSM, NDE performance is defined in terms of the propensity for tube burst or leakage as determined from the correlations between tube burst, leak rate and bobbin A TETRA ENGINEERING GROUP, INC.
Page 5
6424/RP 3580 i
I voltage (Reference 9.) Since degradation is no longer defined in terms of metallographic c:epth or length of a discontinuity, it is necessary to define NDE performance z.s a function of bobbin voltage.
I i
i l
i l
l 1
l l
ATETRA ENGINEERING GROUP, INC.
Page 6 u-
6424/RP 3580
/
- 3. General Considerations 3.1 issues and Notation A number of issues addressed in this report and supporting documents are I
related to the capability of eddy current inspection of steam generator tubes to detect ODSCC at TSPs, including:
+ What is the detection capability for an axial crack?
+ Is the detection capability dependent on oobbin coil voltage?
Are differences between the capability of analysts evident?
+
+ What influencing factors are generic? Which are site-specific?
Does the use of dual-analysts improve the detection capabilit-?
+
If so, is the improvement dependent on bobbin voltage?
What is the analyst sizing uncertainty for indications called by bobbm?
+
The detection capability for an ODSCC indication at a tube support plate is influenced by several factors, including:
+ extent of degradation, through-wall and axial length,
+ presence of interfering signals due to denting or accumulation of elemental copper and tobe support plate distortion and deposits,
+ difficulty in resolving small amplitude indications from NDD When tubes are significantly degraded, these effects are minimal virtually all such indications are detected. When tubes are slightly degraded, these effects reduce the detection capability.
Quantifying the probability of detection for ODSCC at TSPs requires the understanding of basic probability concepts and statistical methods. Standard notation is used to describe more complicated written descriptions of factors which contribute to the overall detection capability.
For example, the expression r[v s r ]is used to describe the probability that a o
variable v has a value less than or equal to some number, Vo. The template form P[...] will be used in this report to specify the probability of the argument specified within the brackets. An example might be the probability that the voltage 'or an indication is less than or equal to 0.5 volts; P{v<0.5].
A physical condition which is dependent on the presence or occurrence of some other situation is termed a conditional probability. For example, the conditional ATETRA ENGINEERING GROUP, INC.
Page 7 u
d 6424/RP 3580 probability of an analyst detecting an indication when it is in fact present is expressed as:
P' analyst detects indication l indication is present in tube' The vertical bar is used to separate the event of interest (in this case detection of the indication by the analyst) from the event it is dependent on (a truly degraded i
tube.)
3.2 Classification of ODSCC by Bobbin Inspection A classical analysis of POD theoretically relates the presence of some' number of l
physical cracks to the proportion of indications which are correctly detected by bobbin coil inspection. Such an approach relies exclusively on the establishment of " ultimate truth" based on known physical defects; a practical impossibility for field defects due to the difficulty, critical path time, radiation exposure and expense associated with pulling tubes.
Instead, the approach pursued in this Study was to determine an alternate measure of inspection capability which reflects the use of industry standard guidelines and procedures for inspection and data analysis.
Bobbin inspection has two objectives with respect to detecting ODSCC indications at TSPs:
- 1. limit the probability of not calling an indication when one is present to some small amount, p (0<p<1),
- 2. limit the probability of calling an indication when one is not present to some small amount, a (0<a<1),
The probability that an indication is called when truly present, p, will be considered the analog to the classical probability of detection (POD). For convenience, this measure of detection capability will also be termed POD although it is determined in a different way than classical POD. The probability of calling an indication is related to the risk of not calling the indication (again, when it is truly present) by the relation: p = 1 - p.
p represents the probability of not detecting the indication and is a related measure of the capability of the inspection process. If callable ODSCC degradation is present in a large number of tube intersections, N, which are inspected and submitted for analysis, it is expected that a fraction 100p% will not be called as indications by bobbin.
A TETRA ENGINEERING GROUP, INC.
Page 8
6424/RP 3580 However, it is expected that the majority of the intersections,100(1-p)%, will be i
called correctly as indications. The discrepancy between N 'and the number called N(1-p), is the number of indications which have been " missed" by the inspection process. It is this number for each voltage interval which is used to adjust the predicted end of cycle (EOC) population of ODSCC to account for undetected indications.
A related situation is where an indication is called by the analyst ODSCC is not actually present at the TSP; this represents afalse call. If a larp; number of non-degraded tube intersections are submitted for inspection, it i, expected that a portion 100a% will be incorrectly classified as containing ir.lications based on an error in interpreting-the signal for the tube location. Sin 6arly, it is expected that a large portion 100(1-a)% will be correctly classified a3 naving no ODSCC l
presenti.
Table 3-1 describes the four possible outcomes of evaluating a potential indication.
Table 3-1 Probability of Classifying Boboin Indications for ODSCC Degradation Analyst (s) Call Analyst (s) Do Not Call ODSCC Deg adation Indication Indication Present 1-p p
Not Present a
1-a 3.3 Detection Capability forIndividuals and Teams Detection capability will be developed first for an individual analyst. Standard industry practice of using two analysts will be evaluated based on the results from individual analysts. Comparisons will also be provided between the l
detection capability of the field analysts (primary, secondary) who called the ODSCC indications in the field and the 12 Qualified Data Analysts who were used in the Study.
Theoretically, the probability of detection by two analysts (call them A and B) is calculated as
' A tube with no degradation is difTerentiated from a tube with an indication which is resolved to an NDD.
A TETRA ENGINEERING GROUP, INC.
Page 9
e l
~
6424/RP 3580 P(detected by A and /or detected by B) = l-P not detected by A and not detected by B'
= 1 -(1 - p3 )(1 - p,)
if the calls by analyst A and B are statistically independent. However, the degree of independence exhibited by the primary and secondary field analysts is known to be somewhat less than the theoretical maximum, as reported in Reference 4,
" Eddy-Current Steam Generator Data Analysis Performance". Similarities in training and accumulated field experience appear to result in correlations between analysts.
As an example, if two analysts are truly independent, the effective POD of the dual-analyst team can theoretically be as high as:
p,m, = 1 -(1.80)(1..80) = 96%
assuming the detection capability of one analyst for a particular voltage size is 80%.
However, the actual improvement will have a probability distribution of values which generally lie between the average individual analyst performance and the theoretical team performance. Generally, the average values of the team performance are closer to the theoretical value than to the average value for the individual analyst.
For indications at or above the repair limit, this improvement will be seen to have less effect because the POD by an individual analyst is very high in that region. It is more significant for smaller indications where the POD for an individual analyst is lower. A model of team performance versus bobbin volts will be presented which enables interpolation for specific bobbin voltage values.
Such a model will be useful to incorporate bobbin POD into burst and leakage analyses for ODSCC at TSPs.
Two additional subjects are addressed in Appendices B, C of this report: voltage sizing uncertainty and a measure of NDE performance which is adjusted to account for cracks which are " born" during the operating cycle.
ATETRA ENGINEERING GROUP, INC.
Page 10
6424/RP 3580
- 4. Detection Capability for Bobbin Inspection This section presents a description of factors which influence bobbin detection capability. An analysis of the POD Study data from Braidwood, Byron and Catawba is presented in Section 5.
A number of factors combine to influence the capability of the inspection program to detect significant degradation for a specific steam generator. Some of 1
these factors are generic to the industry assuming that vendors perform the steam generator inspection in accordance with industry guidelines while others are plant-specific. Plant-specific effects are primarily those factors which relate to local conditions in the steam generator which vary from plant to plant. These effects include the history of operation and water chemistry control, systematic differences in material properties and operating temperatures, extent of sludge accumulation and/or presence of elemental copper and the resultant effects on local noise levels and characteristic signals. Their influences result in potential reductions in the ability to obtain a detectable voltage, relative to the background level of " noise." The measure of detection capability described in this report is based on the tactors described in the following sections.
4.1 Probability of Detectable Voltage it is assumed that the tubes of interest have an amount of degradation which can be detected under ideal conditions (that is, y mer hardware calibration, conformance to inspection procedure guidelina and low local noise levels.)
Tbc conditional probability that a bobbin voltage will exceed a minimum fareshold value, given that ODSCC degradation is present will be designated PODV. This quantity represents the fraction of degraded tubes which will yield a bobbin coil voltage from the multi-frequency mix which is detectable above the j
local " noise" level. The probability statement for this conditional probability will be expressed as:
PODF = P v 2 vddegradation where vo is a minimum detectable voltage relative to the normal voltage fluctuations associated with " clean" tube. This value is plant-specific and is determined by local conditions and operating history. A high level of i
performance of the inspection system hardware is ensured by the use of qualified probes and by following standard industry guidelines for the calibration, data acquisition and analysis. It is assumed in this analysis that the probability of obtaining a voltage which exceeds the minimum detectable, given that a tube is significantly degraded, is 100%. This is supported by the data presented in Figure 2-1.
A TETRA ENGINEERING GROUP, INC.
Page 11
6424/RP 3580 Thus, hardware calibration and operation by highly skilled and trained personnel is the basis for maintaining a high probability of achieving a detectable voltage.
4.2 Probability of Bobbin Call A second element of the detection capability for ODSCC at TSPs is the ability of the field analyst team to detect an indication that is present in the inspection data. Under the assumption that a callable voltage is contained within the data presented to the analyst (s), the probability that the indication will actually be called depends significantly on the level of analyst training, adherence to standardized inspection guidelines and experience with ODSCC.
This conditional probability that a field analyst calls an indication, given that a detectable voltage is present will be designated POBC where:
POBC = P indication called by BClv 2 vo Much of this report addresses the analysis the extensive field data from the POD Study to quantity this factor.
4.3 Probability ofBobbin Detection The probability that a detectable bobbin indication will, in fact, be detected will be designated bobbin POD which is determined by the product of the factors:
POD = PODV
- POBC Under the assumption that the inspection hardware system has been properly calibrated and is operated in conformance with the standard industry guidelines, this represents a measure of the generic level of capability of the bobbin inspection. As discussed in section 4.1 the capability of the hardware to generate a detectable signal is assumed for the purposes of this analysis to be 1.0; therefore bobbin POD is equal in value to the conditional probability of bobbin detection, POBC. Section 5 of this report provides a complete analysis of the POD Study data.
l A TETRA ENGINEERING GROUP, INC.
Page 12
6424/RP 3580
- 5. Probability of Bobbin Detection This section describes the analysis of the results of the POD Study which was conducted by EPRI on behalf of Commonwealth Edison Co. and Duke Power Co.
A general description of the data set is followed by results from correlation analysis and detailed calculations of bobbin POD by voltage for both individual analysts and dual-analyst teams.
5.1 POD Study Data Set Reference 3 describes an initial evaluation of the POD Study based on the inspection data for 3/4" tube from Braidwood Unit 1, Byron Unit 1 and Catawba Unit 1. The POD Study evaluated an extensive quantity offield data from 818 tubes, which included 5726 tube-tube support plate intersections. Some of these intersections have experienced axial ODSCC. Figure 5-1 summarizes the number of indications which were identified from the complete set of field inspection data.
900 E 800 S 700
,y 600
] 500
- - 400 8 300
.E 200 l 100 0
RPC Not RPC Peer Review Confirmed Confirmed Figure 5-1. Composition of EPRI POD Study Data The POD Study included data from three bobbin indication categories:
confirmed indications, indications not confirmed and indications called by an independent peer review group. Confirmed indications are bobbin indications where field analysis calls were confirmed by RPC analysis. Not confirmed indications correspond to those field calls where the bobbin indications were not confirmed by RPC analysis or were not RPC inspected. Peer review indications correspond to calls by at least one of the Qualified Data Analysts in the Study which were subsequently confirmed by the peer review group.
Peer Review indications are predominantly in the low (less than 1 volt) range, although there are a small number which are somewhat higher. The fact that a small number of these indications are obtained is totally consistent with the A TETRA ENGINEERING GROUP. INC.
Page13 w
M24/RP 3580 concept of POD; that is, varying numbers of undetected indications are expected due to the generally lower POD for indications less than 1.0 volts. No Peer Review indication exceeded 1.5 volts.
Each analyst was required to independently review, evaluate and call each of the TSP data sets. The results of the analyst evaluations of the field data were then tallied according to voltages established by a separate peer review group.
Figures 5-2 to 5-4 provide frequency histograms which indicate the relative number of indications by voltage size in the POD Study data set for each of these groups. It is apparent that the data set contains predominantly indications in the 0.5-1.5 volt range.
Table 5-1 provides a summary of the nunibar of indications in each of the three categories, tabulated by bobbin voltage.
Table 5-1 Comed / Duke /EPRI POD Data Set
~
Voltage RPC Confirmed Not RPC Peer Review Sub-Total Confirmed
<0.25 9
4 24 37
.25.49 121 63 112 296
.50.74 169 83 78 330
.75.99 175 45 31 251 1.00 - 1.49 228 22 6
256 1.50 - 1.99 105 4
0 109 2.00 - 2.49 41 0
0 41 2.50- 2.99 12 1
0 13 3.00 - 3.49 11 0
0 11 3.50 - 3.99 7
0 0
7 4.00 - 4.49 5
0 0
5 4.50 - 4.99 2
0 0
2
>5.00 5
0 0
5 TOTAL 890 222 251 1363 The indications which are listed under the column " Peer Review" represent indications which were called by at least ane of the 12 qualified analysts and which, following subsequent review by the peer review group, were designated as callable indications. It will be seen that the number of implied " misses" by field analysts is in fact consistent with the PODS which were determined in this study.
ATETRA ENGINEERING GROUP, INC.
Page14
I i
8o M
en xg Relative Frequency (%)
o ci 5
5 S
ci 8
8 8
S g
j
~\\
\\
<0.25 i
.25.49 i
.50.74 t
g.
.75.99 9
r 1.00-1.49 1.50-1.99 3
5 D
Q y 2.00-2.49 k
?
f
- ;;M 4,
i E
- 2.50-2.99
=
[.
b R
3.00-3.49 i
k 5m j
9 3.50-3.99
~
2 it g
?
l h
g i
4.00-4.49 l
,y l
- s t
i l
g
-p 4.50-4.99 c,
l l
>5.00 l
_ k_
2 r
.r y.
30
?
m
[
Relative Frequency (%)
5 Ei S
ES 8
8 S
o cn Q
a, N
N j
<0.25.
i
.25.49 i
.50.74p-g
~
3 g-Q
.75.99 w
o P
1.00-1.49
,, -. ~
it ii-
-h n
u 5
1.50-1.99 I l
%g 2
m 8
f:r 4
d
[ 2.00-2.49 '
e:3 o
a.
g i:
e?
2.50-2.99
.. l s
g 3.00-3.49 E-3
-t
{
3.50-3.99 l
9 h
4.00-4.49 6
E.
il 3
s-u l
- i, 4.50-4.99 l O
.[
>5.00
..A.
I b
a T
2
8oya
{
Relative Frequency (%)
b a
m u
u w
s a
2 o
m o
e o
m o
m o
m
~,\\
\\
<0.25 1
1
.25.49
_z, ;.f
?
.50.74
=
L
.75.99 m
E 1.00-1.49 l
,j 7
1-<
i i
2 1.50-1.99 l' M
m I
k
?r I;
A
$ 2.00-2.49 i
te i
o i
?
o
.h l
s-if i
2.50-2.99 1
j.
3.00-3.49 k
4 e
e i
3.50-3.99
.y l
?
j i
l 3
4.00-4.49 l
3 4
t g.
4 4.50-4.99 L
l
>5.00 i l
i j
J l
2 oE o
6424/RP 3580 5.2 Dependence ofPOD on Voltage Industry maintains that analyst capability for detecting an indication which corresponds to significant degradation due to ODSCC at TSPs is proportional to the voltage of the indication. Such a relationship implies that as the magnitude of the voltage increases, the likelihood that an analyst will detect it also increases.
The industry position is based on extensive inspection experience with ODSCC at TSPs at several sites which involved the deployment of calibrated hardware technology and the use of Qualified Data Analysts. Field experience has been confirmed by preliminary efforts to quantify bobbin detection capability in previous test programs (Reference 4.)
The POD Study consisted of the independent analysis by 12 qualified data analysts of the data from the 5726 TSP locations with the distribution of indications which were deccribed in the preceding section. This large number of data points enabled extensive statistical evaluations with definitive results.
i These results were used to establish individual analyst POD estimates using all the data from the three indication categories; graphs and tables are provided by voltage interval in Appendix A. A general trend of increasmg POD is observed with increasing voltage.
)
Table 5-2 lists the average, median, mimmum and maximum percent detected by the 12 analysts for each voltage interval based on only the RPC confirmed indications. These results indicate very high percentages detected for all voltage intervals. Figure 5-5 indicates the discrepancy between the observed number of
" missed" indications and the number predicted when POD is calculated from -
only the RPC confirmed indications. Thus, the results based exclusively on RPC confirmed indications appear overly optimistic.
{
A TETRA ENGINEERING GROUP, INC.
Page 18 i
y m%
C D "
M E $.
M P o
m M
vi o m
$ C-O O
Ul N E
m u
o y
y a
e o
U-A 09 O
m R-.
mC. 21
" Missed" Indications u
w x e y
g o
a a 3
s e o
0 3
2 E
o g
Oc q
o o-o 5 e e
e A
e vi m
m m
Q m
n N
A m
to o
N 3
g m
s ye m
9 9
9 9
9 9
9 m
yo 6
O O
O O
O O
O y
2 o,,
o nu e
~
M 8
3' o z
<0.25 D
g o
5 IQ b
o s
2 a
4o R.
a O
o e m 25^49 C
9 C
o-a" 9 9 *.*.N Y' M N L
m 3
s-e x a
D t
o m N e o c
A o
e-
~o e
~
n h
E.
Ok 8
.50.74 g
3-D g
m o
2 at t
=
3 n
tn a,
c 8_
1.[
Q 8
.75.99 E Np Y
c n-m.
E D
>e
^
i a
.a
.a a
- 10. m o o o o e e e ao 8
3 g 3
3 C-
=
< 1.001.49 9 P P P
.N
- 91
- E.,
O 02 o e
o_
o o o o A
w vi e w
E @m o
s g
gN 4
^g 1.50-1.99 k 8
w ". <
3 e.
- =
o x,
g m
mu u
>2.00 h.
O O$
R C
9 O
o c.
a o
- 3 Y
o e
e e e
co N w o
EU a
E P
5 5
M P P P 3
e 4m g-
>3 00 g:
0 2
- ' N W
3 B<
=
y gE D-Q o
o m
a m
n.
v 5%
25.
- c. -.
2 5
3 R3 e.
c D:
1 m c.
o 3
n a
.a a
.a i
a
- T. p
- O E.
o o m
o O o e o e o l
o --
O0 P P o
A o w o o
P 9 9.* P O-t Q
l
.M c-cr B ;
o o o x
e a
3
-e w
m$
b h5 5
W
+e m:
82 e
a a-e l
l l
6424/RP 3580 One alternative is to consider all indications (RPC confirmed, not RPC confirmed or not RPC inspected and " Peer Review) as the complete set of indications which should be called by a " perfect" analyst. Statistics for bobbin POD under this assumption are provided in Table 5-3. Significantly lower fractions detected are apparent in the low voltage range. Figure 5-6 compares the predicted number of " misses" using this alternate assessment of POD with the observed number. The agreement is somewhat closer, although overly conservative in the sense that significantly larger numbers of indications are predicted to be missed than were actually found by the Peer Review group.
Table 5-3 POD for the 12 Study Analysts - All Indications 3 Voltage Average (%)
Median (%)
Minimum Maximum
(%)
(%)
<0.25 35.6 37.8 10.8 51.4
.25.49 56.6 55.2 42.9 66.2
.50.74 70.9 70.6 58.2 81.5
.75.99 82.0 81.3 74.1 90.4 1.00 - 1.49 92.8 93.0 85.2 98.0 1.50 - 1.99 99.4 100.0 96.3 100.0
>2.00 99.3 99.4 97.6 100.0
>3.00 100.0 100.0 100.0 100.0 Figure 5-7 provides a graph of the probability of detection for an individual analyst versus the bobbin voltage.
i
' Data set includes: RPC confirmed indications, indications not confirmed by RPC and peer review indications.
A TETRA ENGINEERING GROUP, INC.
Page 20
6424/RP 3580 s
120.0 100.0
~ ' ~ ~
~
~
~~~
~~~~~ ~ ' ~
80.0
- - - ~ ~
-~
60.0
~
~"
~
~ ~ ~
40.0. --
~
~
- -~
j g Observed '
20.0 0.0
.x 1
e D m= D zzaw= w
=
A 4
4 i
i
?
^
Bobbin V Its Figure 5-6. Predicted and Observed Number of Peer Review Indications (All Data) 100 %.
~
~
90 %
80 %
~ ~ ' ~ ~
~
~
g 70%
~
~
~
s 60%
~ ~ ~ ~
~
~
~
J
^
e gg, 40%
~
^
3gy
{
20%
~
,g 0% -
s s
s s
Z T
T o_
9 Bobbin Volts Figure 5-7. Probability of Bobbin Call vs. Voltage A1ETRA ENGINEERING GROUP, INC.
Page 21
6424/RP 3580 1
5.3 Comparison ofField and Study Analysts Reference 3 reportr three comparisons of the results obtained by the original (primary and secoadary) field ana.lysts for the POD Study data and the corresponding resn!ts obtained by the 12 analysts. These results will be summarized in this section.
The first comparison is of the average fractions of indications detected (RPC confirmed, Not RPC Confirmed, or not RPC'd and Peer Review). A similar performance level is anticipated if the Study analysts are truly representative of the capability of field analysts. The results are listed in Table 5-4 and provided in graph form in Figure 5-8. A correlation coefficient of.949 indicating very good agreement was reported in Reference 3. Figure 5-8 provides a graph comparing the field analysts capability with the capability of the set of analysts used in the Study. No bias of test analysts outperforming field analysts or vice-versa is indicated.
Table 5-4 Study and Field Analyst Capability for Detecting Bobbin Indications i
(RPC Confirmed, Not RPC Confirmed, Peer Review)
Voltage Average of Study Average of Analysts (%)
Field Analysts j
i
(%)
<0.25 82.4 94.4
.25.49 91.9 79.8
.50.74 95.6 87.0
.75.99 97.0 91.7 1.00 -1.49 96.9 96.1 1.50 - 1.99 99.7 99.0
>2.00 99.2 98.2 Not RPC 66.9 80.6 Confirmed Peer Review 17.0 16.1 A TETRA ENGINEERING GROUP, INC.
Page 22 l
6424/RP 3580 100 % -
.r 90 %
Study Analysts,
< 70%
- -.J Ideal g
--.-~
j60%
N 50%
1B j40%
1; o 30 %
- __ _.u.
8 520%
- - ~ _
E' 10%
0%
0%
20%
40%
60%
80 %
100 %
Fraction Detected by Field Analysts Figure 5-8. Comparison ofStudy and Field Analyst Capability The second case considered is when all indications (RPC Confirmed, Not RPC Confirmed and Peer Review) are combined. The results (average fraction detected) are listed in Table 5-5. Figure 5-9 provides the results in graph form.
Very good agreement is again indicated by the correlation coefficient of 0.999 for this tabulation of the data.
Table 5-5 Comparison of Study and Field Analyst Capability (AllIndications)
Voltage Average of Study Average of Analysts (%)
Field Analysts
(%)
<0.25 35.6 35.1
.25.49 56.6 53.5
.50.74 70.0 69.8
.75.99 82.0 82.0 1.00 -1.49 92.8 93.6 1.50 -1.99 99.4 99.1
>2.00 99.3 97.6 A TETRA ENGINEERING GROUP, INC.
Page 23
I 6424/RP 3580 100 %
_I._ __
g 90%
f f~~.- ~ Study ^5aiy~sts~
,[
< 80%
Ideal g
m E 70%
_ + -
E 60%
8 a%
50 c
.j o
$40%
30 %
30%
40%
50 %
60%
70%
80 %
90 %
100 %
Fraction Detected by Field Analysts Figure 5-9 Cornparison of Study and Field Analyst Capability (All Indications)
The third case is a comparison of the fraction of all indications detected by analyst with the corresponding number of overcalls (Peer Review indications.)
Table 5-6 presents the results of this comparison, tabulated by analyst. A significant trend of increasing fraction detected with increasing numbers of overcalls is indicated in Figure 5-10, consistent with the correlation coefficie,c 0.88 reported in Reference 3.
w ATETRA ENGINEERING GROUF, NC.
Page 24 t
6424/RP 3500 Table 5-6 Correlation of Fraction Detected and Overcalls by Analyst Analyst Number of Overcalls Total Fraction Detected (Confirmed by Peer (AllIndications)
Review) 1 378 79.6 2
708 84.1 3
255 73.7 4
385 76.4 5
462 77.0 6
430 83.9 7
604 82.2 8
165 74.8 9
566 82.8 10 157 71.6 11 86 68.2 12 296 70.3 Eimary 333 75.5 Field Secondary 367 76.8 Field 90 %
80 %
=
j70%
_ _ p'
=
60%
e Predicted g
g 50%
~
c f40%
2
~...._;-
k 30%
~
i 0
200 400 600 800 Indications Confirmed by Peer Review i
Figure 5-10. Fraction Detected vs. Indications (by Voltage) Conprmed by Peer Review A TETRA ENGINEERING GROUP. INC.
Page 25
6424/RP 3580 Variability in analyst to analyst capability is significant in the voltage region below 1.0 volts as indicated in Tables A-1 through A-4. Statistical tests such as conventional x contingency table analysis confirm the existence of statistically 2
significant differences. At higher voltages, the individual analysts achieve very high detection capability as discussed in previous sections of this report and there is much less variability between analysts. This is confirmed in the discussion of dual-analyst team capability in Seedon 5.5.
5.4 95% Confidence AnalysisforIndividual Analyst POD Lower 95% confidence limits on individual analyst POD as a function of voltage were determined by applying the standard expression (References 5,6):
x P.95 = (x + (n - x)F )
where x is the number of indications detected out of a total of n for a specified voltage interval i and F is the 95% value of the F probability distribution with 2(n-x+1) and 2x degrees of freedom, respectively. The results are presented for an individual analyst in Figure 5-11. The uncertainty in POD as measured by the difference between the average and the lower 95% confidence limit is seen to be a decreasing function with increasing voltage.
l ATETRA ENGINEERING GROUP, INC.
Page 26
6424/RP 3580
~
90%.
~ ~ ~ DAverage
~
4 g Lower 95% CL 7
80%.
f 70%
~ ' ' ~ ~ ~ ~
~
~
~
~
z 8
60 %
5 0
50 %
~
h 40%
~'
-~
~~
~~
~~
~
~~
~ ~ '
=
]
30%
~
~
~
~
~
.o
[
20 %
10 %
~
~
~
~
~
0% -
~
~
<0.25
.2 5-
. 50-
.75-1.00-1.50-
>2.00
>3.00
.49
.74
.99 1.49 1.99 Bobbin Voltage Figure 5-11. Lower 95% Confidence Lirnitsfor Bobbin POD (Single Analyst) 5.5 Bobbin Detection Capability for Dual-Analyst Teams Theoretically, the POD for dual-analyst teams should be significantly improved relative to that achieved by individual analysts provided that the calling of indications by the two analysts is statistically independent.
The theoretical POD for two independent analysts with equal capabilities is:
Fwu = 1 -(1 - P)(1 - P)
I However,it was observed in a previous study of eddy-current data analyst performance (Reference 4) that as analysts become better trained and more experienced, they tend to call indications similarly, the consequence being that their performance becomes more and more alike. This leads to a reduction in the benefit which can be achieved due to the correlation of analyst calls, but still provides improvement relative to the average individual PODS.
The POD Study data were used to calculate the effective performance of dual analysts by a procedure which consists of determining all combinations of selecting 2 analysts from the 12 analysts that were used in tne POD Study.
A TETRA ENGINEERING GROUP, INC.
Page 27
6424/RP 3580 j
j The number of " pseudo-teams" is calculated in terms of the standard combination notation.
i
'12' 12!
12 *11 = 66 Y=
=
=
(2s 2!!0!
2 That is,66 pseudo-teams of dual-analysts were formed combining the individual scoring for each voltage interval. An indication was considered detected by the pseudo-team if either or both individuals had called it. All indications in the POD Stu ly (RPC confirmed, not RPC confirmed or not RPC'd and Peer Review) were used in the dual-analyst evaluation.
The results were tabulated for all 66 pseudo-teams in terms of the average, standard deviation, minimum and maximum fractions detected by the " teams" These results are presented in Table 5-7 by voltage. Figure 5-12 provides a comparison of the average POD for dual-analysts and the average POD for individual analysts based on the POD Study. The significant trend of increasing POD with increasing bobbin voltage is evident in both cases. Also, the relative improvement offered by the dual-an alysts is seen to be greatest for the lower voltage indications.
Table 5-7 Bobbin POD for Dual-Analyst Teams Voltage Average for ideal for Average for Standard Mirumum (%)
Maximum (%)
Indwidual(%)
Dual-Team pod study Deviation (%)
(%)
" Teams"(%)
<0.25 35 6 57.0 46.0 6.0 32.0 61.0
.25. 49 56.6 81.5 67.0 5.0 56.0 76.0
.50.74 70.9 91.5 79 5 4.H 67.0 86.0
.75. 99 82.0
%.8 87.4 3.3 80.0 93.0 1.00 -1.49 92.8 99.5 96.2 2.0 91.0 98.4 1.50 *1.99 99.4 100.0 W.9
.03 98 0 100.0
>2110 W.3 100.0 99.8
.04 98.8 100.0 ATETRA ENGINEERING GROUP, INC.
Page 28
100 % -
1 90% -f '
~~~~
~
80%-
~~
~
70% ---
~
~-
e t
~~
i 60% -
t
..L l
3
-~--
50% -
O Dual-Analysts j
~~~-
"O " ^"#'I 40% -
.h 30% -/
'~~
0 20% -
/
~ - -
10% -
[
...)..
p y
_ y 0% --f4
<.25
.25.49
.50.74
.75.99 1.00-1.49 1.50 - 1.99
>2.00 Bobbin Volts POD.XLS: 2 Anatysts All V Figure 5-12. Comparison off.1rerage) Team andIndividual Capabilitv 1126195 5:33 PM
6424/RP 3580 Figures 5-13 through 5-19 provide a comparison of the results described previously for an individual analyst, the theoretical improvement which requires statistically independent calls by the analysts and the distribution of pseudo-team results based on an analysis of the POD Study data.
Several important trends are indicated by these graphs. First, at low voltages the dual-analyst detection capability is distributed more or less symmetrically between the capability of an average individual analyst and that of an ideal team.
Generally the average dual team result is somewhat closer to the value of the ideal team than the individual, although there are infrequent occurrences where a particular " team" performance is equal to, or just slightly less than the average individual capability.
This is expected since all combinations of analysts have been used and there is a small probability that two of the lesser capable analysts will be paired as a team.
j As voltage increases, the average dual-analyst capability becomes more like that of the ideal team; this is particularly evident in Figures 5-17 through 5-19 for j
indications which exceed the IPC of 1.0 volts for 3/4" tube.
l l
l l
A TETRA ENGINEERING 3ROUP, INC.
Page 30
100.0 % -
O
^
90.0% -
80.0% --
70.0% -
60.0% -
D g
50.0% -
T2 Analysts - POD Study o
7 2
01 Analyst-Average u.
40.0% -,
E Analysts - Ideal 30.0% -'
20.0% - '
10.0% -
0.0% ^~'"*T'"' "'~7b
y
'""T "i
'==f E9S " 9 " " "~i'* ~"^{"
0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 Fraction Detected POD.XLS: 2 Anal. <.25 Figure 5-13. Dual-iinalyst Detection Copability (<.25v) 1126195 5:33 PM
100 % -
O 6
90% -
80% -'
70% s 60% --'--
i6.
5 50% - '
E2 Analysts-POD Study
- s I
{
D1 Analyst-Average u.
~ ' " '
40% -
30% --'
V 1
20% -. '
10% - '
I": -
' "7 S
"""i'""~ " i
~"'i ~ """f 0%--^*"'""T"
'"T "7
"7 f
0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Fraction Detected POD.XLS: 2 Anal. <.50 Ingure 5-14. Dual ctnalyst Detection Capability (.25.49v) 1126195 5:33 PM
100 % -
O 4
90% --'
80% d' 70% -/
_ 60% -
N 5
50% -
l 52 Analysts - POD Study 3g 01 Analyst-Average 40% s'
" '#8 ~ *'
30 % s 20%-
10% -
0% *'"*~7" "'7
- 9 TV
~~ "[
"7
_= i
~~"""~"""f' O.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Fraction Detected I
i i
POD.XLS: 2 Anal. <.75 Figure 5-15. Dual <tnalyst Detection Capability (.50.7k) 1126/95 5:33 PM
100 % -
O 6
a it 90% --
- f. -
D s
- 1 80% -
ej 70% -I
.)
- s 60% -/
I U
?
g 50% -/-
((.
52 Analysts - POD Study a
01 Analyst-Average e
2 u.
y O2 Analysts-Ideal 40% --
+
30% -/
4 F
fi
.a te z
20% --/
f..
10% -/
p,.m
, men)
) ym-. -
rawvn-
~n=.~,8,,.
w,.,,g,,.ww-ryy,
-- i
^
0%
0.75 0.8 0.85 0.9 0.95 1
Fraction Detected POD.XLS: Anal. 2 < 1.00 Figure 5-16. Dual-Analvst Detection Capability (.75.99r) 1/26/95 5:33 PM
100 % -
O C
90% - '
80% - '
~
70% - '
~
60% --
~
D g
50% ---
52 Analysts-POD Study 3
{
01 Analyst-Average 1
40% -'
AnalyMs - Meal 30% --.'
20% -
i 10% --'
0%
~ '* T
"'T
> '7
^~~T
~"~ T
- ~T
?""
?
0.905 0.915 0.925 0.935 0.945 0.955 0.965 0.975 0.985 0.995 Fraction Detected POD.XLS: 2 Anal. <1.50 Figure 5-17. Dualvinalyst Detection Capability (1.00 - 1.49v) 1126/95 5:33 PM
I 100% -
F 90%-
4 r
i.-
80%-
70% --
t 60%-
~
~
u g
50%
E2 Analysts-POD Study
- g 01 Analyst-Average E
40% -
^
s
+
30 % 4 '
20% -
/
10% --'
~ h ~w y ~=e g~: vmm, m ~w 7 m.,
yw=, c - ~n e enn evm r
0.98 0.98 0.985 0.99 0.995 1
Fraction Detected POD.XLS: 2 Anal. < 1.99 Figure 5-18. Dual,1nalyst Detection Capability (1.50 - 1.99v) 1/26195 5:33 PM
100 % -
O 6
90%
~
l
[*
80% -'
70% -
i; 4
60% -
at e
3 g
50% -/
E2 Analysts - POD Study
- s
=
{
01 Analyst-Average u.
7 02 Analysts-Ideal 40% --
30 %
L-
/
20% -'
4 10% -/
~
s f,8D 'j & Y'Df#
- "TI'
- $@f f
$,j5%*?7 T[7IKTT'
' N 7D ?4* 7IT"C'ITD@'D*'YW" ?
l 7
f 0%
i i
i i
i
-- i 0.98 0.98 0.985 0.99 0.995 1
Fraction Detected POD.XLS: 2 Anal. >2.00 Figure 5-19. Dual,tnalyst Detection Capability (>200v) 1126/95 5:33 PM
6424/RP 3580 5.6 Model of Dual-Analyst Bobbin POD Detection capability for ODSCC at TSPs has been shown to depend on bobbin voltage. The results obtained from the POD Study were presented in the preceding sections in histogram form for voltage intervals. The results can be applied in that form; that is, by using a constant POD value for any bobbin indication which falls in the corresponding voltage interval. Or equivalently, a probability model for bobbin detection capability can be fit to the POD results to allow accurate interpolation for specific bobbin voltages. Such a model will be useful if the treatment of bobbin POD is fully incorporated in the accident leak rate and probability of burst analyses typically performed as part of a plant's IPC submittal.
Many functions are candidates for representing the smooth relationship between POD and voltage. One which was considered is related to the probability model used for probability of leakage in the ODSCC leak rate methodology (Reference 9, Appendix G); a variant of the logistic function (Reference 10.) The logistic function is a common model for engineering processes which possess an asymptotic behavior. The following voltage-dependent form of the logistic was considered:
1 POD =
1 + c"""'b."'I The particular form considered is useful for representing probabilities in terms of a variable (in this case bobbin voltage.) This three-parameter model was fit to the average dual-analyst results from the POD Study as reported in the previous section. The coefficients solutions were:
a = -1.842 o
a = -2.950 i
a = 0.788 This model of dual-analyst detection capability provides an accurate and convenient method for calculating a value of POD to apply to a specific voltage indication, as indicated in Figure 5-20. Confidence limits for the logistic POD model can be calculated via the same approach as is described in Reference 9 for the probability of leakage model in the ODSCC leak rate methodology.
A TETRA ENGINEERING GROUP, INC.
Page 38 l
6424/RP 3580 100 %.
90% _ _
g 80% ---
70%.__ _
60%-- _
Q 50% -
o b 40% _ _-
E Average Dual-Analyst POD 30% _ _ _
Log. tic Model is m
,2 20%.
~.
n 1.
10%.._ - _.. - _ - _ - -. _. - _ _ _. _
0% -
~ ~.
0 0.5 1
1.5 2
2.5 Bobbin Volts Figure 5-20. Logistic Model of Dual-Analyst Bobbin POD 5.7 Applicability of Results to 7/8" Tube with ODSCC at TSPs The results developed in the preceding sections are equally applicable to 7/8" diameter Alloy 600 tube. This applicability is based on the conventional procedure for scaling optimum inspection frequencies for differing wall thickness of like material tube (Reference 11.)
The conventional inspection frequencies for 3/4" tube are a mix of 550/130 KHz; likewise, the conventional frequencies for 7/8" tube are a mix of 400/100 KHz.
These specific frequency pairs were purposely selected to provide the optimum frequency for detectability. The optimum frequency for a specified wall thickness is obtained by evaluating-10p F, =
2 I
p = 98 (resistivity ofInconel 600) t = wall thickness i
i A TETRA ENGINEERING GROUP, INC.
Page 39 l
4 ^
a 4
6424/RP 3580 The optimum divider frequency is 1/4 the primary frequency. The optimum 4
inspection frequencies are calculated to be 408KHz for 7/8" diameter tube with 0.049" wall and 530 KHz for 3/4" diameter tube with 0.043" wall thickness.
i r
1
+
\\
i t
i I
I A TETRA ENGINEERING GROUP, INC.
Page 40
6424/RP 3580
- 6.
SUMMARY
OF RESULTS This report has established the generic voltage dependence of detection j
capability for ODSCC at TSPs in steam generators with drilled tube support l
plates via eddy current inspection by bobbin coil. The approach described in this report was to develop a measure of overall inspection capability, as opposed to a classical approach to establishing POD. Statistical analyses were performed to quantify voltage dependence, to evaluate the uncertainty in POD for specified voltages and to investigate potential variability between analyst calls for the low voltage regime. These analyses confirm the voltage dependence of bobbin detection capability and provided sufficient information to characterize uncertainties associated with eddy current inspection.
The values of bobbin detection capability presented in this report provide substantiated alternatives to the conservative NRC assumption that p=60% for all ODSCC at TSP indications. Bobbin detection capability exceeds 60% for all indications greater than 0.5 volts. Bobbin detection capability exceeds 95% for indications larger than the IPC repair limit of 1.0 volts. The POD for large indications (exceeding 2.0 volts) is essentially 100%.
9 As such, it is appropriate that utilities take credit in plant submittals for the high probability of detecting indications which, under postulated accident conditions, may be structurally challenging either from a tube burst or leakage perspective.
This credit is based on the presumption that future inspection activities will be performed in accordance with standard Industry Guidelines.
The relationship between bobbin POD and RPC POPCD is discussed in Appendix C. It was concluded that the POPCD approach provides an alternative, plant-specific measure for evaluating NDE effectiveness, albeit one which is directly influenced by the rate at which new degradation is forming.
Such a plant-specific approach complements the generic results developed in the POD Study.
A basis was described for applying the results of the POD Study to plants with 7/8" diameter steam generator tubes so long as an inspection frequency mix of 400/100 KHz is applied. Therefore, the results presented in this report may be considered industry-generic for plants with primarily axially-oriented ODSCC at drilled tube support plate locations.
The bobbin POD results should be incorporated with the methods for calculating the probability of tube rupture and accident leakage. Preliminary analysis indicates a significant benefit - as much as an order of magnitude in the predicted ATETRA ENGINEERING GROUP, INC.
Page 41
6424/RP 3580 probability of tube rupture and EOC leakage can be attributed to the use of the industry-generic, voltage-dependent POD.
l 4
i A 7ETRA ENGINEERING GROUP, INC.
Page 42
6424/RP 3580 I
- 7. References
- 1. " Voltage-Based Interim Plugging Criteria for Steam Generator Tubes",
NUREG-1477 (draft), June 1993.
- 2. R.J. Kurtz, et. al., " Steam Generator Tube Integrity Program / Steam Generator Group Project - Final Project Summary Report", NUREG/CR-5117, May, 1990.
- 3. D.H. Harris," Assessment of Eddy-Current Data Analysis of ODSCC in Steam Generator Tubes", December 9,1994.
- 4. " Eddy-Current Steam Generator Data Analysis Performance", EPRI Report TR-102549, June 1993.
- 5. W.J. Conover, Practical Nonparametric Statistics,2nd Ed., John Wiley and Sons,1980.
- 6. G.J. Hahn and W.Q. Meeker, Statistical Intervals: A Guide for Practitioners, John Wiley and Sons,1991.
i
- 7. SPSS for Windows, Version 6.1, SPSS, Inc.1993.
- 8. M. Sears, Commonwealth Edison, private communication, December 9,1994.
- 9. "PWR Steam Generator Tube Repair Limits - Technical Support Document for Outside Diameter Stress Corrosion Cracking at Tube Support Plates",
EPRI Report TR-100407, Revision 1, August 1993.
- 10. D.W. Hosmer and S. Lemeshow, Applied Louistic Renression, John Wiley and Sons,1989.
A TETRA ENGINEERING GROUP, INC.
Page 43
6424/RP 3580 Appendix A - Bobbin Capability for Individual Analysts 1
1 1
l A TETRA ENGINEERING GROUP, INC.
Pagel
Table A-1 Indications Called by Qualified Analysts (<0.25v )
Analyst Sample Size RPC Confirmed Peer Review Not RPC Confirmed p
1 37 7
4 1
32.4 %
2 37 3
1 0
10.8 %
3 37 7
2 2
29.7 %
4 37 8
6 2
43.2 %
5 37 9
7 3
51.4 %
6 37 7
5 1
35.1%
7 37 7
6 2
40.5 %
8 37 8
5 2
40.5%
9 37 8
3 4
40.5%
10 37 8
5 1
37.8%
II 37 9
3 2
37.8%
12 37 8
2 0
27.0 %
1 Table A-2 Indications Called by Qualified Analysts (0.25v - 0.49v)
Analyst Sample Size RPC Confirmed Peer Review Not RPC Confirmed p
i 296 107 14 22 48.3 %
2 296 89 12 26 42.9 %
3 296 102 12 30 48.6 %
4 296 118 23 36 59.8 %
296 120 26 50 66.2 %
6 296 106 22 26 52.0 %
7 296 112 15 35 54.7 %
8 296 116 2v 29 55.7 %
9 296 116 21 54 64.5 %
10 296 116 25 52 65.2 %
11 296 117 37 41 65.9 %
12 296 115 14 33 54.7'/o A TETRA ENGINEERING GROUP, INC.
Page A-1
Table A-3 Indications Called by Qualified Analysts (0.50v - 0.74v)
Analyst Sample Size RPC Confirmed Peer Review Not RPC Confirmed p
1 330 157 7
37 60.9 %
2 330 146
'I 39 58.2 %
3 330 158 6
48 64.2 %
4 330 167 19 63 75.5 %
5 330 168 13 80 79.1%
6 330 157 12 47 65.5 %
7 330 158 13 56 68.8 %
8 330 163 20 56 72.4 %
9 330 169 14 78 79.1 %
10 330 167 21 81 81.5%
11 330 166 18 77 79.1%
12 330 162 Ii 48 67.0 %
Table A-4 Indications Called by Qualified Analysts (0.75v - 0.99v)
Analyst Sample Size RPC Confirmed Peer Review Not RPC Confirmed p
1 251 164 3
19 74.1%
2 251 163 4
21 74.9 %
3 251 165 2
25 76.5 %
4 251 171 6
33 83.7 %
5 251 173 10 44 90.4 %
6 251 170 5
26 80.1 %
7 251 170 1
36 82.5 %
8 251 171 4
26 80.1 %
9 251 174 4
42 87.6 %
10 251 174 8
42 89.2 %
11 251 173 7
37 86.5 %
12 251 168 1
28 78.5 %
A TETRA ENGINEERING GROUP, INC.
Page A-2
~
i Table A-5 Indications Called by Qualified Analysts (1.00v - 1.49v)
Analyst Sample Size RPC Confirmed Peer Review Not RPC Confirmed p
1 256 209 0
9 85.2 %
2 256 219 1
11 90.2 %
3 256 213 0
13 88.3 %
i 4
256 225 2
17 95.3 %
i 5
256 227 1
22 97.7 %
6 256 217 l
12 89.8 %
7 256 223 0
16 93.4 %
8 256 222 2
13 92.6 %
9 256 228 1
20 97.3 %
10 256 228 1
22 98.0 %
11 256 222 1
18 94.1%
12 256 219 0
16 91.8 %
Table A-6 Indications Called by Qualified Analysts (1.50v - 1.99v)
Analyst Sample Size RPC Confirmed Peer Review Not RPC Confirmed p
I 109 105 0
2 98.2 %
2 109 103 0
2 96.3 %
3 109 104 0
4 99.1 %
4 109 105 0
4 100.0 %
5 109 105 0
4 100.0 %
6 109 105 0
4 100.0 %
7 109 104 0
4 99.1 %
8 109 105 0
4 100.0 %
9 109 105 0
4 100.0 %
10 109 105 0
4 100.0 %
1i 109 105 0
4 100.0 %
12 109 105 0
4 100.0 %
i A TETRA ENGINEERING GROUP, INC.
Vase A 3
Table A 7 Indications Called by Qualified Analysts (>2.0v)
Analyst Sample Size RPC Confirmed Peer Review Not RPC Confirmed p
i 84 82 0
1 98.8%
2 84 83 0
0 98.8 %
_, 3 84 83 0
0 98.8%
4 84 83 0
0 98.8%
5 84 83 0
1 100.0 %
6 84 82 0
0 97.6 %
7 84 83 0
1 100.0 %
8 84 83 0
1 100.0 %
9 84 83 0
1 100.0 %
10 84 83 0
1 100.0 %
i1 84 83 0
1 100.0 %
12 84 83 0
0 98.8%
Table A-8 Indications Called by Qualified Analysts (> 3.0v) i Analyst Sample Size RPC Confirmed Peer Review Not RPC Confirmed p
I 30 30 0
0 100.0 %
2 30 30 0
0 100.0 %
3 30 30 0
0 100.0 %
4 30 30 0
0 100.0 %
5 30 30 0
0 100.0 %
6 30 30 0
0 100.0 %
7 30 30 0
0 100.0 %
8 30 30 0
0 100.0 %
i 9
30 30 0
0 100.0 %
10 30 30 0
0 100.0 %
11 30 30 0
0 100.0 %
12 30 30 0
0 100.0 %
A TETRA ENGINEERING GROUP, INC.
Page A-4
100.0 % /
90.0% -
W y
60.0% -
E o
E 50.0 %
e hff E
4 0
/e T*
M E
Analyst 1 Analyst 3 Analyst 5 Analyst 7 Analyst 9 Analyst 11 roo.xLS: Detection Percent (<.2Sv)
Figure A-1. Conditional Probability ofDetection by Analyst ( <0.25 volt) 11419 5 4:58 PM
OQ M
m 8
U u
u a
ui o
-4 ce co P
P P
P P
P P
P P
g h
h h
h h
h h
h h
g i
I I
I I
I t
]
?
k
\\
\\
\\
\\
\\
\\
\\
S 1
Analyst 1 @$%%7dd@T#$;l 7
~4 l
l 9
Analyst 2 a
Analyst 3 M f " !4'* * "J E Q $j]
~
[
Analyst 4 3Q$y!jQETR%Egif$%@$R7 KS I
g.
g g[
t h
l I
f Analyst 5
~
- 7^ M9WW7# # 4W "^
Analyst 6
$$j$*d/2Q "J
k Analyst 7 3.
g 0
g.
g o.1 Analyst 8 MM$ig@J{$$$$@)MS i
P
[Y f
Analyst 9 W%3QWyJ7M@@FM :N^+.' A I
I t
i a
^^T"!Y7FY" 4."$3 P*W* %($ '
[
Analyst 10 g 2
s i
i Analyst 11 7
- 7 0 5 7 E P i" T *" " '# C M-
- ; yl Analyst 12 W$}fy '%Tj;$${$
n s
8, E
9-
~
100.0 % /
4 p
.~
p
{:.
IJ XV
- }
g 60.0% -
50.0% -
1 hk fjfj hk fI,-
i b
G~
~
[5!
k
{
@ij m
en tg 30.0% -
I 20.0% - d Analyst 1 Analyst 3 Analyst 5 Analyst 7 Analyst 9 Analyst 11 POD.XLS: Detection Percent (.50.74v)
Figure A-3. Conditional Probability ofBobbin Deiection by Analpt ( 30
'4 volts) va:gs a-s8 pu 4
m
.m
--- m ua t-e ew-.
--W N
l 1
i 100.0 % /
i U
G 70.0%
__ [
[
,f y<
~
~-
o:
30.0%
'g nalysti Analyst 3 Anafvst5 Analyst Analyst 9 Analyst 11 POD.XLS. Detectson Percent (.75-99v}
Figure A-4 Conditionall'robability ofBobbin Dereaion by Analyst (. 7.5 - 99 volts) tratus s.ss pu
-=
a l
Detection Percent for Field Indications, No-Confirms (RPC) and Peer Review Indications 100.0 % /
9 90.0 %
80.0 %
l
~
l f.
1, i
30.0% -
nalyst1 Analyst 3 Analyst 5 Analyst 7 Analyst 9 Analyst 11 pod.XLS. Detectron Percent t100-149v)
Figtore A-3 Conditional Probabilstr of Bobbin Detection by Analyst (1.00 - 149 voltst 1ratgs a sa pu
]
.;i
=
i l
i t
l 2
i t
l 100.0 % /
n l
90.0% -
c
~
i a
1
[
15 l
j
??
80.0% -
}
U l
c-70.0% -
3; I"d l
60.0% -
i d
~
50.0% -
l l
l
-f 4
..)
40.0%
l$
I Y
o s
s.
{
- g p
- 4 p
f t
30.0% -
.'y 8
e d#
l-
-'5 20.0% - d Analyst 1 Anaryst3 Analysts Analyst 7 Analyst 9 Analyst 11 l
l l
I POD.XLS: Detection Percent (1.50-199v)
Figuire A-6 Contlitional Probability of Bobbin Detection by Analyst (1.50 - l 99 voltsi statg5 a 58 pu t
4 i
i 100.0 % /
Q:
'ji Y;
l 90.0 %
6!
Y_
'l
,w t-e-
80.0 %
.e 4;
w 70.0% -
d
~
5^
~
i e
60.0% -
~
+
+
,N f.;'
{'
l 50.0 %
2
- t. h.
g s
40.0%
M. ^;
L:,
~
s.
o g.-
9 7;t q
e
.s..
'/
h.E
,; l' b
~
30.0% -
+.
4
- 1. 9 y,-
20.0% M W
D Analyst 1 Analyst 3 Analyst 5 Analyst 7 Analyst 9 Analyst 11 POD.XLS: Detection Percent (>2 Ov)
Figure A 7 Conditional Probability of Bobbin Detect lon by.-Inalyst P 2 0 volts) 11stes a ss Pu
E i
d 100.0 % /
g 90.0% -
4 e
3 Y-
<a 5-s 80.0% -
{-
.+'
~
t-70.0 %
~
[r i
60.0 %
t i.
50.0 % -
-i:
t, <
,a
<rz 40.0% -
e e
+
4
?-
1 '-
y, L:
y.
G.
'+
30.0% -
E
[
i
,},
Ij i
20.0% - I-E
-A l
Analyst 1 Analyst 3 Analyst 5 Analyst 7 Analyst 9 Analyst 11 i
k POO.xLS: Detecten Percent (>3 0 v)
Figure A_8 Conditionall'robability of Bobbm Detection by.4nalyst (> 3 0 volts) 11419s 4.s8 Pu l
r
I 6424/RP 3580 l
i Appendix B - Bobbin Voltage Sizing Uncertainty The POD Study Data were also used to evaluate the uncertainty associated with analyst placement of vector balls when sizing the voltage for an indication. Table B-1 provides a summary of the bobbin sizing uncertainty data (Reference 12) as a function of increasing voltage. A general trend of decreasing voltage size variability is exhibited with increasing voltage call; this indicates that the analysts are more likely to size an indication with a similar magnitude as the "true" voltage becomes large.
Figure B-1 provides a graph of this trend as measured by the median, minimum and maximum values of the 12 Study analysts for each voltage interval. Figure B-2 provides a graph of the same RMS uncertainty measures, but grouped by 0.5 i
volt intervals. Again the trend of decreasing RMS sizing uncertainty with increasing voltage is evident.
j i
?
r I
t i
A TETRA ENGINEERING GROUP, INC.
Page B-1
n Table B-1 RMS Bobbin Voltage Sizing Uncertainty Study Analyst Flei!d Analyst Volts i
1' 2I 3:
4 I SI 6'
'7' 8[
9' 10 11 12 Primary iSecondary
< 25 0.31:
0.56i 0.26 0.25:
0.20:
0.30:
0.45:
0.26i 0.26' O.29:
0.24:
0.20 0.13i
~.ib!
.lBi 0.15I ~ 6.iDI U~13 ~~ U.15I ~ ~ ' 0.22
~I
~6705 ~ d.15'
.68 0.iS 50.74
. f [0.1i} ~ 0.i1[ d.iO{ 0.12 }.i3'!~0.i7!~.5{l~D.14 0
d.'i4' 25.49
'D.iI!
0.14[
6.17[ 0.iS[ '6.1[ f 0.}1j _6.ii 0.i5i_' _, ^6.i4
.75.99 O.16 0.14; 0.15 0.14; 0.13!
0.17!
0.18:
0.15!
0.14-0.18 0.14!
0.16 0.15' O.23 i.00 -1.49
' _' 23[ 0.2iI d.i9f 0.25{[ __0.15]d.23[ d.2[ 6.15{6.i6(('612 D.T9i}~651.i5 0.25 ~ 1.50 - 1.99. 0.18' O.12: 0.15! 0.11 l 0.10'.
- 6. 5!
d.iii b.65I 0.05I 605!"D~64! 6.07 D. ~~~ .i5 0.17; 0.19: 0.06; 0.12; 0.10 0.09I 0.11 0.16' O.17 0.06! 0.05I 6.05i .0 -2.49 d.17' D.26i ^ 5.50 .95 .16I d.iO[ _ U,05' . 5{ D.13 ' D.69 .08 I I d.8I '.8! 6.i6' ~655I D. 3 ~ 6. 5 ~ ~~ .i ~ 3.60-3 45' f l0.i1[ 0.16! 0.05l 0.11l [0.iif _ U.i1] 0.i9i .11{0.i1{ Dj6} ]~ 6.05[ ~'_dj08 ~. ' O. i i . [ 6.~06 ~ 3.50 - 3.99 i 0.04: 0.04: 0.01: 0.04l 0.04l 0.06 l 0.04 0.04 j 0.04; 0.04 0.01 j 0.01 0.16 0.05 4.0 -4.49 j ~ D.d5{ U.64 0.02[ 0.04 [ 0.64[ U.64} 0.04{ d.d4{ _6.04{ ".64[_U.02" 6.02 . 5 '_ ] U.55 I 4.50 -4.99 j _0.04;. 0,04}__0.00 0.04[ 0.04j 0.04; 0.04[ 0.04[_ 0.04! O.04! 0.00 0.00 0.11 0.25 ~ 1 > 5.00 .}___0.24[_0.24; 0.00 0.24 l 0.24l 0.24{ 0.24j 0.24[ 0.24[ 0.24j 0.00; _0,.00 _. 0.20! 0.20 Not Confirmed 0.28: 0.15I 0.isi 0.21} 0.21: 0.24i 0.28; 0.21: 0.24! 0.22: 0.16 0.19 0.24i _0.26
I i RMS Voltage Sizing Uncerts.inty o o o o o o 8 s E O o <.25 + -~~- 8 .25.49 + >--a t .50.74 +> u l j .75.99 + >- e i 3 I 5 1.00 -1.49 + >-e l 9 .~ k 1.50 -1.99 + a kn? d 2.00 -2.49 + - > --E i g E n 8 m g o 2.50 - 2.99 + >~--- a 5 i Q l 5 3.00 -3.49 +- a te oO' O-3' 3.50 -3.99 + - >- a w 4 %m 4.00 -4.49 +- pm 4.50 -4.99 o-a >5.00 m 4 i Not Confirmed +- c 5 I 1 d !> E + iE E E !; g R 5'l' ) is g','. e i
r o 50.00 4 i 1 i t 40.00 2 -j -{- i i l e l i i I i -- -- - 30.00 + - e l i x Min (%) i x Max (%) at 20.00 1 e Weighted (% Volts) ~ 1 i i
- -Predicted x
i i i i i j 1 5 10.00 1 - - - - - - - f- - - - - - - - --- - - - j-- -- -j-- -- t r i i i i t x l R x 0.00
t--------7---
-- - p --- - ---
- - p' - - - - -; - - - -It i i l i i l i i i 1 l t l -10.00 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 Bobbin Volts Figure B-2. Grouped RMS Sizing Uncertainty vs. Bobbin Volts
~ l I l Appendix C - Prior Cycle Detection ard RPC Confirmation C.1 Probability ofRPC Confirmation of Bobbin Calls l An alternate measure of the capability of the eddy current inspection is the extent to which bobbin indications are confirmed by rotating pancake coil (RPC) l inspection. The tube location of a bobbin indication is often inspected by RPC to l confirm and size the degradation. This aspect of the inspection capability goes l beyond the simple detection of an indication to obtain a more detailed l description of the indication. Differences exist between the detection threshold l of bobbin and RPC; indications called by bobbin are not always confirmed by RPC, whereas RPC confirrned indications generally were also called by bobbin. i The probability that an indication is confumed by RPC, given that significant l degradation was present and it was called based on bobbin is designated Puc j _i where: ) Pm. = P indication confirmed by RPCl degradation present andindication called by BC Related to this quantity is the (conditional) probability of an NDD,5 ven a call by i t bobbin, as it follows that not all bobbin indications are confirmed as indications l by RPC. Since a bobbin indication will be resolved to either a confirmed l indication by RPC or an NDD, the following relation results i Pym = P[ indication resolved to NDD by RPCl indication called by BC) l = 1.- PRPC In practice, there is a complication because not all bobbin indications are always i RPC inspected. So, estimates of the probability of an RPC confirmed indication may be biased by the number of indications which are actually RPC inspected. This is why this study did not rely on the results of RPC confirmed indications alone; additional indications based on bobbin which were not RPC confirmed, or which were not RPC inspected were also included as were indications confirmed l i by the peer review group. ?i C.2 Probability of Prior Cycle Detection h Bobbin POD is the quantity which is required for defining the population of undetected indications for accident leakage and probability of tube rupture analyses under IPC/ APC. While a POD for RPC inspection coula be defined as l the product of the bobbin POD and the probability of confirmation by RPC, this l l Par,e C-1 ATETRA ENGINEERING GRGUP, INC.
F approach is likely to underestimate the capability of the bobbin-based detection approach due differences in the sensitivity to local noise levels. Therefore, RPC detection thmsholds may differ from those of bobbin inspection and so may not always confirm the presence of an indication called by bobbin when one is l actually present. Nevertheless, it has been proposed (Reference 7) that a measure termed POPCD (probability of prior cycle detection) be considered as an alternative to bobbin POD. This alternate measure relates current bobbin detected and RPC con 6rmed indications to the fraction which were reported as indications at the prior cycle (inspection.) A high value of POPCD is also an indicatior of an effective inspection method as well as a low incidence of new indications. A low value of POPCD corresponds to either an ineffective inspection method or a high incidence of new indications. POPCD is defined as: =N POPCD(Y) = mygg +my ny where:
- 1. nv is the equal to the sum of the numerator and the new indications RPC confirmed in the latest inspection, muw.
- 2. mv is the total number of prior cycle reported indications confirmed by RPC:
confirmed in latest insepetion plus confirmed and repaired in prior inspection. The quantities mv, nv ud mevare related by: ny = fny +(i-f)ny = my +mygg wherefis the fraction of all RPC confirmed from the current inspection which r were conhrmed in latest inspection plus confirmed and repaired in the prior inspection. l From this relationship it follows that POPCD reduces to: I"' POPCD(V) =- =l fnv +(1-I)ny It can be concluded that POPCD, while not a measure of detection capability,is - still useful as a measure of the incidence of new (RPC confirmed) indications from cycle to cycle. Page C-2 ATETRA ENGINEERING GROUP, INC. .}}