ML24303A222
| ML24303A222 | |
| Person / Time | |
|---|---|
| Issue date: | 10/29/2024 |
| From: | Christopher Hunter NRC/RES/DRA/PRB |
| To: | |
| References | |
| Download: ML24303A222 (1) | |
Text
Various CCF Topics Christopher Hunter Performance and Reliability Branch Division of Risk Analysis Office of Nuclear Regulatory Research October 30, 2024, Public Meeting
Key Messages
- The NRC agrees with NEI and industry stakeholders that ECAs should be as realistic as possible and not overly conservative.
- The NRC uses a data-driven approach to represent CCFs in a transparent and scrutable manner.
- However, the NRC acknowledges that existing CCF parameters and modeling have greater uncertainties when compared to most of the other SPAR model parameters.
2
NRC Approach for ECA Activities
- NUREG-2225, Basis for Treatment of Potential CCF in the SDP, (ML18033A152) provides the following guidelines:
- Credit for CCF Defenses Typical defenses are already reflected in the existing CCF data.
Rare and exceptional circumstances where unique CCF defenses are not reflected in the current CCF parameters can be considered (see Deviations of Key Principles).
o The NRC would need additional information about how RMAs at plants with RICTs to determine if and how such measures should be considered in ECAs.
- Elimination/Reduction of CCF Impact due to Extent of Condition Demonstration of functionally redundant components does not eliminate or reduce the potential for CCF (see Key Principle #3).
Given the existing alpha factor values (usually in the low 10-2 range or lower), it is expected that redundant components would be available.
- Piece-Part and Causal-Level Arguments The SDP assesses the licensee performance deficiency (PD) at the proximate cause level (see Key Principle #1).
The licensee PD could be manifested as different piece part failures due to the same cause.
3
Recent CCF Activities
- The NRC has completed work in calculating additional CCF parameters for base SPAR model and potential use in ECAs:
- CCF parameters derived from component-specific priors.
- Causal alpha factors CAFM.
- Cross-unit CCF modeling and treatment is also being explored to improve realism and reduce conservatism.
4
Component-Specific Priors 5
=
Background===
- The current CCF parameters (i.e., alpha factors) are calculated using a generic prior consisting of all CCF events from all component types (INL/EXT-21-62940).
- While the use of this generic prior provides a stronger prior by leveraging all available data, it combines failure data across dissimilar component groups.
- This results in CCF parameters being affected by dissimilar components.
Applying a generic prior to CCF component types that have not experienced many failures may result in a conservative estimation of CCF potential.
- The prior uses data for 1997 through 2015.
- The Bayesian update is performed to calculate component/
failure mode CCF parameters.
Uses a rolling 15-year rolling period; current parameters use specific CCF data from 2006 through 2020.
6
Component-Specific Priors
- Reducing the effect of the dissimilar components on the CCF parameter is desirable.
- INL was able to calculate priors for the following component groups (INL-EXT-21-65527) using the using CCF data from the 1997-2015 period:
EDGs Pumps Valves Strainers Other Components
- Results show some significant differences with the existing prior.
The priors for valves and strainers increase, while the priors for EDGs, pumps, and all other components decrease compared to the generic prior.
Components that fall in the other component category are still affected by dissimilar components.
7
Prior Comparisons Group Size Alpha Factor 2015 Generic Priors 2015 Pump Priors 2015 Valve Priors 2015 Strainer Priors 2015 EDG Priors 2015 Other Priors Mean Mean
Mean
Mean
Mean
Mean
2 2
2.1E-02 1.1E-02 -45% 3.5E-02 73% 9.7E-02 373% 7.9E-03 -61% 9.3E-03 -55%
3 2
1.4E-02 8.6E-03 -40% 2.7E-02 87% 5.1E-02 251% 5.2E-03 -64% 6.7E-03 -54%
3 4.7E-03 2.0E-03 -57% 6.5E-03 39% 2.2E-02 370% 2.2E-03 -54% 1.2E-03 -74%
4 2
1.4E-02 8.0E-03 -41% 2.4E-02 74% 4.9E-02 260% 4.0E-03 -70% 7.3E-03 -46%
3 4.4E-03 2.4E-03 -46% 9.0E-03 108% 1.7E-02 298% 2.1E-03 -52% 1.1E-03 -75%
4 2.5E-03 9.5E-04 -62% 3.0E-03 20% 9.9E-03 295% 8.0E-04 -68% 1.6E-03 -38%
8
CCF Parameter Comparison 9
Template Name CCCG Alpha Factor 2020 CCF Parameters Component-Specific Priors 2020 CCF Parameters Generic Priors Difference a
b Mean a
b Mean Mean EDGs Fail to Run 3
2 2.38E+00 3.12E+02 7.58E-03 2.47E+00 2.23E+02 1.10E-02
-31.1%
3 3
8.26E-01 3.13E+02 2.63E-03 7.83E-01 2.24E+02 3.48E-03
-24.4%
EDGs Fail to Load/Run 3
2 1.27E+00 2.84E+02 4.45E-03 1.36E+00 1.95E+02 6.90E-03
-35.5%
3 3
3.20E-01 2.85E+02 1.12E-03 2.77E-01 1.96E+02 1.41E-03
-20.6%
EDGs Fail to Start 3
2 1.26E+00 3.78E+02 3.33E-03 1.35E+00 2.89E+02 4.66E-03
-28.5%
3 3
8.20E-01 3.78E+02 2.16E-03 7.77E-01 2.89E+02 2.68E-03
-19.4%
Service Water Strainers Plug 3
2 2.62E+00 3.56E+01 6.85E-02 2.80E+00 8.13E+01 3.33E-02 105.7%
3 3
1.34E+00 3.69E+01 3.51E-02 1.33E+00 8.27E+01 1.58E-02 122.2%
2 3.41E+00 6.94E+01 4.69E-02 3.35E+00 1.06E+02 3.06E-02 53.3%
4 3
8.07E-01 7.20E+01 1.11E-02 7.03E-01 1.09E+02 6.42E-03 72.9%
4 4
2.31E-01 7.26E+01 3.18E-03 2.95E-01 1.09E+02 2.69E-03 18.2%
Batteries Fail 3
2 1.30E+00 1.93E+02 6.66E-03 9.76E-01 7.74E+01 1.25E-02
-46.7%
3 3
2.09E-01 1.94E+02 1.08E-03 2.77E-01 7.81E+01 3.53E-03
-69.4%
2 4.44E-01 2.51E+01 1.74E-02 4.69E-01 3.54E+01 1.31E-02 32.8%
HPI Pumps Fail to Run 2
2 7.33E-01 6.11E+01 1.19E-02 7.19E-01 4.10E+01 1.72E-02
-30.8%
4.16 KV Breakers Fail to Close 4
2 1.56E+00 2.68E+02 5.80E-03 1.26E+00 1.48E+02 8.42E-03
-31.1%
4 3
2.35E-01 2.69E+02 8.73E-04 4.02E-01 1.49E+02 2.69E-03
-67.5%
4 4
3.32E-01 2.69E+02 1.23E-03 2.31E-01 1.49E+02 1.55E-03
-20.6%
Status
- NRC is evaluating the replacement of existing CCF parameters derived from the generic prior with those derived using component-specific priors in the base SPAR models in the next routine data update in 2026.
- In the interim, NRC analysts can replace the existing CCF parameters in assessments where CCF is a significant contributor to risk.
- A spreadsheet with all CCF parameters calculated using the component-specific priors, including a comparison with the existing CCF parameters, is located on the Operating Experience Results and Databases public webpage.
10
Application of the CAFM 11
=
Background===
- The CAFM was (partially) developed to allow the CCF impact to be focused more on the proximate cause.
- The causal alpha factors may be higher or lower depending the cause and component/failure mode.
- However, the proximate cause categories used in the CCF data collection process do not always align well with licensee PDs being evaluated in the SDP.
- The causal alpha factors are available for all components (INL/RPT-23-72728).
- Calculated using causal priors for the five CCF cause groups (component, design, environmental, human, and other/unknown).
Uses the same data periods for the priors (1997-2015) and Bayesian update (2006-2020) as the existing CCF parameters.
12
CCF Cause Grouping Scheme CCF Cause Group Failure Cause Likely PD Category Component Internal to component; piece-part Setpoint drift Age or wear Design Construction installation error or inadequacy Design and Engineering Design error or inadequacy Manufacturing error or inadequacy Environment Ambient environmental stress Extreme environmental stress Internal environment Human Accidental human action Corrective Action Program Maintenance Management Oversight Operations Procedures Inadequate maintenance Human action procedure Inadequate procedure Other/Unknown State of other component Other 13
Status
- Use of the causal alpha factors has some potentially limitations.
- Causal alpha factors include data of dissimilar components and failure modes, which increases the uncertainties.
- Failure causes may map to several different causal categories, increasing uncertainty in the estimates.
- Use of failure data may not accurately reflect the coupling strength for certain causal factors
- RES is continuing to evaluate the development and use of causal alpha factors.
- In the interim, it is believed that the CCF parameters derived from the component-specific priors provide a better estimate of potential CCF if an alternative approach is used.
- Future guidance within the RASP Handbook is needed on when and how to apply the causal alpha method becomes a recommended approach.
14
Cross-Unit CCF Modeling 15
=
Background===
- This modeling is typically limited to EDGs and service water pumps.
The SPAR models do not capture other intersystem CCFs due to the complexities and data limitations, which is nonconservative.
- Cross-unit CCF can be the dominant risk contributor in some event assessments.
- Cross-Unit CCF modeling can increase the existing uncertainties associated with CCF model or create new issues
- The existing CCF data is collected on a per unit basis.
However, the data can be reviewed to determine if duplicative CCF events for similar components occur close in time at multiple units at the same site.
- Additional upward mapping of CCF events increases the uncertainty with the CCF calculations.
- The modeling also results in additional CCCGs, which the analyst must ensure does not result in double counting.
16
Status
- NRC is evaluating options for cross-unit CCF modeling to reflect differences in CCF coupling and eliminate overlapping CCCGs in NRC event assessments.
- Need to develop the appropriate guidance and determine where it will reside.
- In the interim, analysts can perform sensitivity calculations using factors to represent increases CCF likelihood of key cross-unit components given a failure on the other unit.
- Need to determine if this should be limited to shared systems or systems that can be cross-connected only.
- Further research is needed to determine best way to perform these types of sensitivity calculations.
17