ML20237A396
| ML20237A396 | |
| Person / Time | |
|---|---|
| Site: | Prairie Island  |
| Issue date: | 08/11/1998 |
| From: | NRC (Affiliation Not Assigned) |
| To: | |
| Shared Package | |
| ML17265A791 | List: |
| References | |
| PROJECT-694 NUDOCS 9808140023 | |
| Download: ML20237A396 (7) | |
Text
__-__ _ _ _ _ _ _ _ _ _ _ _ _ _ - _ _ .__. __
, Staff Evaluation on WOG's Proposed Inspection Proaram for Part-Lenath CRDM Housina lssue
1.0 BACKGROUND
On January 23,1998, a non-isolable reactor coolant pressure boundary leak of 0.26 g.p.m. was discovered in a part-length CRDM housing at the G-9 core location of the Prairie Island Unit 2 reactor while it was operating. Metallurgical evaluation of the failed housing confirmed ultrasonic (UT) examination results that a very deep 360* long, partial through-wall crack was present. The metallurgical evaluation results showed the flaw had been undersized by UT examination results. The failure mechanism was identified as hot tearing associated with the fabrication of the Type 309 austenitic stainless steel (309) weld buttering at the Type 403 martensitic stainless steel (403) forging. Chemical analysis results identified the following contaminants on the fracture face of the failed component: sulfur, copper, boron, and zinc.
Following staff's review of the information provided by the licensee, the NRC formally requested WOG to activate the RRG on February 20,1998. The staff met with RRG on February 27,1998, to discuss this issue. On March 6,1998, RRG issued a letter that requested the affected owners to docket their plans for addressing the issue within 30 days and initiate compensatory measures for RCPB leakage. The options identified by RRG for the plan were:
. Remove the housings and cap the reactor head penetrations
. Perform non-destructive examinations to confirm the absence of any cracking
. Perform additional records search to better identify applicability and obtain other data to confirm the absence of any cracking
. Address the capability of using additional RCS leakage monitoring awareness while the issue is being resolved The NRC found these recommendations an appropriate and acceptable response to the identification of the QA breakdown at the vendor's shop and as suitable corrective actions for the potential very large defects that jeopardize RCS integrity.
To date, affected WOG member utilities have inspected or repaired or committed to inspect or repair 102 CRDM 308/309/403 weldments on 51 assemblies at nine operating plants. There is a total population of 284 welds of the type of interest (i.e. Type 309) in 137 installed assemblies and five spare assemblies at 21 operating plants.
On June 11,1998, representatives of WOG summarized this issue at a public meeting with the staff. Based on the weld inspections, fabrication records review, safety assessment, and statistical evaluation of the inspections planned and performed (assuming no flaws are identified in the planned inspections), WOG concluded that (1) the Prairie Island flaw was an isolated event, (2) there is 95 percent confidence that about 95 percent of the remaining welds do not have flaws, and (3) continued operation of plants will not result in a significant increase in risk.
WOG Indicated it plans to close this part-length CRDM housing issue, if no further unacceptable l flaw is identified in the currently planned weld inspections.
l 9008140023 990811 PDR ADOCK 05000306 P PM ,
9 2
2.0 EVALUATION 1
Materials Enaineerina l
The staff agrees with the industry finding that the significant cracking at Prairie Island Unit 2 was fabrication-related. From the contaminants found on the failed component's fracture faces, it
, appears that the failed component was probably inadequately cleaned prior to weld buttering.
l Some of the elements found on the fracture faces are usually contained in commercial cutting lubricants. The staff agrees that the hot tearing most likely occurred during solidification of the weld butter and that subsequent post-weld heat treatment (PWHT) may have extended the i cracking. Further, the heavy oxide scale found on fracture faces indicates that the open crack was subjected to the high temperatures of the PWHT that are well above plant operating temperatures. Because of the large thermal expansion mismatch betwesn the 309 and 403 materials, care must be taken to minimize solidification cracking. The presence of contaminants, from perhaps residual cutting lubricant, would increase the chances for solidification cracking.
The CRDM housing is a safety-related Code component that was manufactured under a quality assurance program that was intended to satisfy 10 CFR 50, Appendix B. The cleaning prior to weld buttering was specified in the contrciling procedure. The surface and volumetric examinations performed failed to assure quality in that a component with severe cracking was not rejected. Therefore, it is clear that the quality assurance program broke down for the failed component, in particular, with respect to Appendix B, Criterion IX-Control of Special Processes, in that cleaning prior to welding appears to be not as specified and Criterion X-Inspection, in that !
the examinations performed for the work operation did not identify the unacceptable defect.
The severity of the cracking found in the Prairie Island Unit 2 part-length CRDM was among the worst identified in a safety-related component at an operating nuclear power plant. The cracking found was well in excess of the depth that could be accepted by analysis pursuant to Section XI, IWB 3600. One portion of the 360* circumferential crack was through-wall and other portions of the crack were in excess of the approximately 75% Code maximum flaw depth limitation.
Nonetheless, limit lond fracture analysis was performed by WOG to determine the margins that existed in the flawed component. WOG stated that the average remaining ligament in the {
cracked component from metallography was about 25%. The failure pressure was calculated to j be 2900 psi for the 309 weld. Based on WOG analysis, there was a margin of about 1.3 (2900/2250) to failure for normal and upset conditions; the ASME code-required margin of safety is 2.77. However, the staff was unable to confirm that the average uncracked ligament was !
ES%. It is not clear, from review of the metallography presented (WCAP-15054), if the 25%
average ligament includes a " mixed zone" of small ligaments across the fracture face. WOG stated that additional capacity existed because the actual strength for the 309 weld is about 10%
higher than used in the analysis. Arguments regarding the margins available for component integrity for OBE and SSE loadings are based on a calculation using material allowables higher than the design allowable. Further, WOG argues that since the flawed component passed a hydro test at 3450 psi and crack growth in service has not been identified, a margin of 1.5 is inferred. The staff does not agree that a margin based on pressure alone is indicative of component integrity structural margins. The staff's view is that the actual margin to failure is smaller than claimed, but the margin is essentially indeterminate. In part, this comes from the staff's review of the metallography and from a review of operating history. Prior to the previous refueling no leakage had been reported for the component. A leak in this reactor coolant pressure boundary (RCPB) component was discovered while the plant was operating, and the l plant was taken out of service as required by the technical specifications (TS). To the staff's I
l l
j
. l l
3 l I
knowledge no unusual loadings from transients or other events had occurred prior to the discovery of the leak. The staff understands that some work was done on the reactor head during the last refueling outage and that the head was removed from and replaced on the reactor vessel during the refueling. It is possible that a load from either bumping the head during i movement or when work was being performed was of sufficient magnitude to cause the crack to I open and leak during the subsequent cycle of operation.
The regulations applicable to this issue are as follows:
- 1. 10 CFR 50.55a(g)(4) requires that throughout the service life of a boiling or pressurized water-cooled reuclear power facili:y, components that are ASME Code Class 1,2, or 3 must meet the requirements set forth in the applicable edition and addenda of Section XI for the facility.
I
- 2. 10 CFR 50, Appendix B, Criterion XVI, states that measures will be established to assure that conditions adverse to safety, such as failures, malfunctions, deficiencies, deviations, ,
defective material and equipment, and nonconformances are promptly detected and corrected.
The leaking component at Prairie Island did not satisfy the above requirements in that Section XI l margins to failure were not maintained and there was through-wall leakage in the RCPB. This !
resulted in a reduction in defense in depth since the reactor coolant pressure barrier was j breached. As explained in the following section on the staff's statistical evaluation, the l inspection program proposed by WOG is inadequate.
Statistical Evaluation At the June 11,1998, WOG/NRC meeting, WOG reported that 35 weld inspections were performed, and found no defective welds. Based on this information, the staff performed its I independent statistical evaluation, and found that this inspection would not catch (with 95%
confidence) as many as 21 defective welds in the remaining population of 248 uninspected welds.
The inspection program proposed by WOG is inadequate from a statistical point of view, it does not provide adequate confidence that appropriate corrective actions are taken to ensure the Type 309 weld buttered 403 components manufactured at Royal Industries satisfy the applicable regu;ation, specifically the required margins for structural integrity. In order to attain this goal, it ,
would be necessary to demonstrate with 95% confidence that there are no flaws remaining in l the uninspected welds. Even if no defective welds are found in the 66 additional welds which '
are to be inspected, and accounting for the 35 welds already inspected, what can be demonstrated with 95% confidence, is only that there are less than seven defective welds in the j remaining population of 182 uninspected welds. l 1
l I
4 The staff has the following additional comments on WOG's statistical evaluation presented at the 5/6/98 meeting (Statistical Evaluation viewgraphs):
Pa,qq,3:
a The inspection results to date are not "1 flaw in 36 welds inspected" but rather zero flaws in 35 welds inspected. The one flaw found was not the result of a random inspection and should therefore not be counted.
Pace 8: 1. Frcrn the '82' Sub-lot column, it appears that p s .0271 with at least 95%
assurance. However, there are two problems with this conclusion. First, from page 6, the value .0271 is the posterior mean. However, the Perdue-Abramson method does not use mean values. It uses the posterior distribution which is
~
given on page 6. From this distribution, the probability c p s .0278 is .235+.637
= .872. Thus, p s .0278 with only 87 percent assurance (not with 95 percent as .
claimed by WOG). The only statement that can be made with at least 95% (
assurance (actually,100%) is that p s .0729.
- 2. From the "309" Sub-lot column, it appears that p s .0278 with at least 95% ,
assurance. Because this population is described by the prior distribution, from j page 6 the probability that p s .0278 is .185 + .630 = .815. Thus, p s .0278 with j only 81 percent assurance (not with 95 percent as claimed by WOG). j Risk Assessment I'
WOG risk analysis is the product of three numbers:
. the probability that a reactor will have at least one flaw, the frequency of operational events that might cause the flaw to fail catastrophically enough to create a LOCA, and
. the probability of failing to mitigate the LOCA. ,
)
The staff disagrees with the Westinghouse analysis on the first two values.
l The Westinghouse analysis uses a probability of '~0.05 for a flaw to exist," presumably in a j single plant. However, that is apparently taken from its statement that there is 95% confidence that the whole population of welds is less than 5% flawed. For a PRA, the appropriate value is the probability that one or more of the 8 to 10 welds in the plant is flawed. To determine that probability properly, the mean or "best estimate" value of the flaw rate should be used, not the {
95th percentile value. Assuming that the inspection of the sample of 101 welds is completed without discovery of another severe flaw, there is 50% confidence that the rate of flaws in the remaining population is less than 0.7/101=0.0069; so there is 50% confidence that there are no more than 1.26 flaws in the remaining uninspected population of 182 welds. Together with the known flaw, that makes a total flaw occurrence rate of 2.26/284= 0.008 for the whole population.
At this rate, the probability that one of the welds in a plant will be flawed is between 0.062 for 8 welds and 0.12 for 16 welds.
l l
l l
l
I 5
( The Westinghouse analysis uses "-1 E 1 E-05/ year abnormal event frequency" as the frequency of occurrence of events that might cause catastrophic failure of a flawed weld in these CRDM housings. That is apparently based on the facts that the flawed housing survived a 3450
, psi hydrostatic test after fabrication and was analyzed by Westinghouse to be capable withstanding an operating basis earthquake. However, for reasons addressed elsewhere in this review, the staff is not confident in that part of the Westinghouse analysis. It is known that the flawed weld survived 23 years of service at Prairie Island Unit 2, so one estimate is that the frequency of occurrence of events that would cause weld failure is probably less than 1/23 years =
4.3 x 10 2/ reactor-year. Since none of the operating PWRs have experienced pressure transients exceeding 3450 psi or earthquakes exceeding the magnitude of operating basis earthquakes in i
- approximately 1500 combined years of operation, one could estimate the occurrence rate of events that could fail this flawed weld as less than 1/1500 years = 6.7 x 10"/ reactor-year. This value is just inside the upper range suggested by Westinghouse. However, it is not clear that the flawed weld at Prairie Island actually would have survived all of the operational occurrences experienced at the other PWRs to date. Although corrosive degradation of the weld during its service life was not evident, it was observed to begin leaking noticeably during the current cycle operation. Some sort of stress imposed during the outage is suspected of producing the leak that occurred later, although no actual stress inducing occurrence was noted. However, other plants have experienced such events as moderate earthquakes, cable snags and impact loading while moving the upper heads and other loads, it is not clear how these occurrences at the other plants would have affected the flawed CRDM housing. Degradation during an outage may potentially make the flawed weld more susceptible to failure during operational events.
The Westinghouse analysis used a "LOCA CCDP ~ 1E-02 -1E-04." The staff agrees that the probability is in this range for failing to prevent core damage, given a LOCA of this size. The ,
staff's analysis uses a value of 1 x 10' for the conditional core damage probability due to small to I medium LOCAs. This is consistent with the results of a variety of NRC and industry PRAs. fl 4
Combining the staff's factors provides a range from 5 x 104 to 4 x 10 / year over which the staff's confidence varies from good to poor that the core damage frequency due to failure of a flawed CRDM housing has been bounded. The staff's range generally overlaps and slightly exceeds the i upper part of the range suggested by Westinghouse, which is "1E-06/yr to1E-10/yr."
3.0 CONCLUSION
The staff has concluded the following:
The leaking component at Prairie Island did not meet the current regulation. In order to assure that the remainder of the population of the 309 weld buttered 403 components manufactured at Royal Industries have the required specified margins for structural integrity, ano to satisfy applicable quality assurance requirements, corrective actions are necessary to provide a high confidence that the deficiencies revealed by the discovery of the weld flaw at Prairie Island Unit 2 did not result in a similar CRDM housing weld flaw at another facility.
l l
l
6 The inspection program proposed by WOG for the Type 309 welds is inadequate from a statistical point of view. As stated in the staff's statistical evaluation, WOG's current inspection plan is inadequate to catch (with 95% confidence) as many as six defective welds in the remaining population of 182 uninspected welds, even if no additional defective welds are found in the sample. There would be only 36% confidence that no defective weld remains in the uninspected population. Dased on the staff's evaluation a combination of inspection or repair of 100% of the 309/403 partial length CRDMs is appropriate.
The level of risk associated with this issue at plants which have not yet inspected similar welds may be small (i.e., the CDF increment is in or below the mid-10 4/ reactor-year range). Considering this level of risk, the staff has concluded that it is not appropriate to require immediate action that would subject plants to additional startup and shutdown activities. The staff considers it more prudent to implement the necessary inspection or repair during the next refueling outage. This should allow planning and qualification of inspection and repair methods that will minimize personnel exposure and best integrate with other refueling activities.
i l
l l
- r
, - Westinghouse Owners Group Project No. 694 cc:
Mr. Nicholas Liparulo, Manager Equipment Design and Regulatory Engineering Westinghouse Electric Corporation
. Mail Stop ECE 4-15 P.O. Box 355 Pittsburgh, PA 15230-0355 Mr. Andrew Drake, Project Manager .
. Westinghouse Owners Group Westinghouse Electric Corporation Mail Stop ECE 5-16 P.O. Box 355
' Pittsburgh, PA 15230-0355 Mr. Jack Bastin, Director Regulatory Affairs Westinghouse Electric Corporation 11921 Rockville Pike Suite 107 Rockville, MD 20852 Mr. Hank Sepp, Manager Regulatory and Licensing Engineering Westinghouse Electric Corporation PO Box 355 Pittsburgh, PA 152304355 l