ML18141A395
| ML18141A395 | |
| Person / Time | |
|---|---|
| Issue date: | 05/21/2018 |
| From: | Ali Rezai NRC/NRR/DMLR/MPHB |
| To: | |
| Rezai A, NRR-DMLR 415-1328 | |
| References | |
| Download: ML18141A395 (7) | |
Text
Report Models Inputs Uncertainties V&V Convergence Sensitivity studies Input Importance Other Risk arguments MRP-105 Partial Partial Partial Partial No Partial No No MRP-116 Partial No Partial Partial No Partial No No MRP-362, Rev. 1 Yes Yes Partial Partial no Yes Partial No MRP-335, Rev. 3-A Partial Partial Partial Partial No Partial No No MRP-395 Partial Partial Partial Partial One paragraph Yes, but limitations No Yes, but not acceptable (1) ignore leakage through weld (2) ignore BAC Are these topics sufficient relative to staff position on PFM, Yes, No, Partial Point of note: Looking at MRP-395 RPVH probabilistic model - Section 4 points you to MRP-335R1 and MRP-375 for details of model. MRP-375 has a description of the model in appendix A, but says its only adapted from MRP-373, which is a butt weld model for MSIP - MRP-335R1 says the RPVHPN model is derived from DM model in appendix A of MRP-335. Where do i find the full details of the code used for MRP-395 and RPVHPN
Comment Models Inputs Uncertainties V&V Convergence Sensitivity studies Input Importance Quantification Other Risk arguments 1
Unclear how deterministic models compare to probabilistic models No discussion on why certain distributions were chosen Some discussion on Weibull and crack growth uncertainty Details of MRPERCRD unclear - Staff requested user manual but it was not
- provided, Never mentioned Limited cases done based on variations in Weibull, crack growth, inspection and temp Never determined what's driving the problem directly Never mentioned 2
pc-Crack used for deterministic and MRPERCRD used for probabilistic - same models?
Why are some constant and some random?
Most uncertain parameters have standard Dev, but no discussion of basis No mention of validation or verification, QA, etc. Some benchmarking and case studies presented - basis for inputs unknown The reasoning for the choices of parameters modified in unclear 3
Report says, built-in crack inspection POD curves - what are they, what are the basis?
Report states user has many options for inputs, but gives no basis for chosen inputs Chpt 7 and 8 on PFM results never mentions uncertainty NRC tried to confirm results with in-house P-CRDM code, but struggled to get similar results due to lack of understanding of MRPERCRD tech basis MRP-105 Comments
Comment Models Inputs Uncertainties V&V Convergence Sensitivity studies Input Importance Quantification Other Risk arguments 1
SRRA basis was used - NRC SE on topic
- full distribution of CGR was used Attempts were made to calibrate to VC Summer - 7.06 initiation adjustment factor - whats the impact of this assumption Basis for distributions never discussed Some benchmarking against VC summer done No discussions Studies on ISI accuracy and frequency - what about drivers to the problem?
None Uses RG1.174 acceptance criteria with a CCDP value of 0.003 - no basis 2
WRS model assumed constant - some discussion on WRS, but no idea what value was used in the analyses - Table 3-1 points back to MRP-112 - Figure 3-5 from that doc has some WRS - was that used?
Table 3-1 and 4 inputs - not clear the basis for distributions or values for inputs - example chose "Hoop/Axial fatigue stress range -
points to MRP-109, 112 - after searching through those documents, no idea what values were used Only discusses uncertainties in initiation CGR, geometry, but not WRS, loads, other material properties - no idea what values were chosen for analysis Code changed from SRRA since SRRA had no PWSCC initiation model what other changes were made? PRAISE was mentioned many times, was that used also? how was that V&V?
Three constant WRS cases run changed leak probabilities by three orders of magnitude. But no real discussion No other risk arguments (Defense in depth, SF, monitoring) per RG1.174 are included 3
Initiation model similar to that in WCAP-14572 Supp 1 (and WCAP-14901 Rev 0 -
I think this is Weibull), but that says.. In common with the pc-PRAISE code, Supplement 1 to WCAP-14572 does not address fatigue crack initiation except In an indirect manner by conservatively assuming that initiated cracks are present at the beginning of plant operation. The limitations of this approach to fatigue crack initiation are addressed below. The effects of crack initiation can conservatively be estimated by assuming one flaw per weld at the start of plant operation.
No actual inputs described anywhere in the document. Points to a variety of references, but no clue what actual input values were used MRP-116 Comments
Comment Models Inputs Uncertainties V&V Convergence Sensitivity studies Input Importance Quantification Other Risk arguments 1
Toughness, resistance, strength with aging models well defined.
Details of J-integral model in appendix Inputs were very vague in R0, but improved after much discussion with NRC Model uncertainty not addressed Not discussed, but alternate calculations were performed by Emc2 during development of N838 not discussed Sensitivity studies on toughness, aged strength, etc in Appendix not really discussed, but only a few main inputs not discussed 2
Could have looked at alternate models - was done by Emc2/NRC Basis for most distribution inputs and fits in appendix Input uncertainty addressed Code based on praise, which has been benchmarked with xLPR -
but no idea of V&V for modified code Sampling and probabilistic structure of code is not documented Most came after discussions from NRC/Emc2 3
Input modified after discussions with NRC/
Emc2 Emc2 also did sensitivity studies to build confidence MRP-362 Rev1 comments
Comment Models Inputs Uncertainties V&V Convergence Sensitivity studies Input Importance Quantification Other Risk arguments 1
No modeling of boric acid corrosion was used.
Wall thicknesses were not varied.
No uncertainties were given for the final results Verification and Validation discussed for some modules and inputs No discussions Some sensitivity studies provided None Uses RG1.174 acceptance criteria - no basis 2
WRS model did not cover the range of possible weld residual stresses.
An unusual temperature distribution was used, with a normal distribution over the range of temperatures as opposed to bimodal temperature distribution over hot and cold leg temperatures.
Uncertainties in several inputs were not quantified.
No independent peer reviews of the codes were performed No other risk arguments (Defense in depth, SF, monitoring) per RG1.174 are included 3
Initiation models use Weibull distributions from previous initiations. This does not take the changes in the numbers of susceptible heads into account nor the effects of site-specific issues such as heat-to-heat variability or surface conditioning in crack initiation.
Some effort at validating inputs, with notable gaps MRP-335 R3 Comments
Comment Models Inputs Uncertainties V&V Convergence Sensitivity studies Input Importance Quantification Other Risk arguments 1
Built on MRP-105 model with known outstanding NRC concerns about the MRP-105 model.
Crack growth rate uses a log-triangular distribution to reach high confidence in covering the all of the mean crack growth rates, but then uses a "local crack growth rate variability" term (fudge factor) to somehow modify the distribution. Then there is a claim that all laboratory data points are included. This does not show the distribution of rates, how they are sampled or the impact of the local crack growth rate variability term and how it allows coverage of all crack growth rate data.
Address uncertainty entirely through the Weibull intercept parameter. However with the variations in the Weibull parameter, there was no discussion on how uncertainty is carried forward through the analysis.
Benchmarking to MRP-105, a known probabilistic analyses for which the NRC did not find acceptable, is not acceptable.
One paragraph on convergence work with one figure.
PLUS-Evaluated different Weibull crack initiation parameters Truncation used with no basis provided on CGRs 2
Model built on MRP-335 Rev.1 which had not been reviewed and approved by the NRC. Also includes assumption that bare metal visuals would address any boric acid corrosion activities, and therefore do not need to be modeled. An assumption the NRC has consistently stated as being unacceptable.
Use of a variable term for POD of an initial missed flaw had a limited if any discussion.
However previously noted as being a key to obtain results that required volumetric inspections on a 2.25 RIY basis.
Some sensitivity study variability was limited. For example for what value of UT or BMV POD where the results show an unacceptable leakage or ejection rate would show how much margin is available to address POD uncertainty. However by just varying 10% does not give an effective sensitivity analysis for this parameter.
Plant experience has demonstrated that there is a low probability of leakage overall considering the possibility of through-weld cracking.
Furthermore, any leaks that might occur due to through-weld PWSCC that is not detectable via the periodic volumetric examinations of the nozzle tube are expected to be relatively small.
Such small leak rates are unlikely to be sufficient to produce the amount of local cooling necessary for substantial boric acid corrosion to occur. NRC - There is no mechanism that prohibits cracks from growing from the weld material into the nozzle. 1/3 of the leaks were from the weld material.
MRP-395 & 375 comments
CCDP of 1 assumed