ML19212A735
| ML19212A735 | |
| Person / Time | |
|---|---|
| Issue date: | 07/31/2019 |
| From: | Office of Nuclear Regulatory Research |
| To: | |
| D. Stroup 415-1649 | |
| Shared Package | |
| ML19212A718 | List: |
| References | |
| Download: ML19212A735 (47) | |
Text
NRC-RES Fire PRA Workshop Module IVAugust 5-9, 2019Rockville, MDNRC-RES/EPRI FIRE PRA METHODOLOGYTask 12 -Fire HRA ATHEANA ExampleDetailed Fire HRA Quantification Slide 2Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDSteps in the ATHEANA Process Slide 3Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDSteps 1&2: Objectives of the AnalysisStep 1: Define and Interpret the IssueNeed to identify, model and quantify relevant HFEs forFire PRA sequencesDefined by scope of fire PRA.Step 2: Define the Scope of the Analysis Address human actions needed to prevent core damage in fire induced initiating events and subsequent accident sequences under full
-powerDefined by scope of fire PRA.
Slide 4Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 3: Describe the PRA scenario (nominal context/base case scenario)Initial Conditions: Single unit two loop PWR with two trains of electrical power. Steady state, full power operation.
-No out-of-service unavailability pertinent to this scenarioInitiating Event: Fire in turbine room causes SBOHFE: Operator fails to manually align 115kV (alternate power) power following loss of both buses and EDGs fail to start.
Slide 5Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 3: Describe the PRA scenario (nominal context/base case scenario)Accident Sequence oFire causes reactor trip oReactor trip and Turbine trip successful.
oAFW failed due to the fire.
oPORV spuriously opens due to the fire.
oThe Main Generator breaker opens and the BOP buses are powered through XTF0001 (reverse) and XTF0002.
oEDG B starts and the Engineered Safety Feature (ESF) Loading Sequencer loads onto bus.
oEDG B trips due to fire damage. The ESF Loading Sequencer is still sending a signal to trip the normal and alternate feeder breakers (for EDG protection) to the bus.
oAll diesels failed
-SBO oDC power remains available until batteries deplete. Batteries will last for 4 hrs Operators Success Criteria oLocally trip the alternate feeder breaker by removing power from the ESFLS to remove the trip open signal.
oEnergize XSW1DA or 1DB from the alternate power source.Consequence of failure:Core damage due to stuck open PORV Slide 6Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 3: Describe the PRA scenario(nominal context/base case scenario)Procedures
- oUpon Reactor Trip, enter EOP
-0Step 3 of EOP
-0 verifies that buses are energized. Buses are de
-energized; this will take the operator to ECA 0.0 [Station Blackout Procedure]Step 10 of ECA 0.0 checks that buses 1DB and 1 DA are energized. Both buses are de
-energized; this will take the operator to AOP 304 due to loss of bus with no EDG.Steps 17 and 18 of AOP 304 are the relevant response actions for this HFE:
Slide 7Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 3: Describe the PRA scenario (nominal context/base case scenario)Operator action success criteria: Reset ESFLS to clear trip signal and align alternate power source to XSW1DA. Required Operator Actions:
1.Shift Supervisor directs the Control Room Operator to power 1DA 2.Reset ESFLS to clear trip signal (local action, skill
-of-craft)a)Local Plant Operator, stationed at or near the MCR, gets ESFLS panel key from the MCR and proceeds to the Relay Room b)Dons flash gear c)Opens left cabinet (~2ft from floor) and locally removes power from the loading sequencer d)Alerts Control Room Operator that the trip signal is clear 3.Close Breaker in MCR a)Control Room Operator will ensure BUS 1DA XFER INIT Switch is in OFF b)Close BUS 1DA ALT FEED Breaker c)Verify BUS 1DA potential lights are energized Slide 8Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 3: Describe the PRA scenario (nominal context/base case scenario)Staffing: Minimum staffing of the plant is as follows:*Dealing with high
-level management issues (e.g., communicating with NRC)**Normally outside CR. Will be in CR within 10 minutes of reactor trip.***Daytime onlyShift Manager* (SM)Shift Supervisor (SS) Unit 1Shift Technical Advisor** (STA)Control Operator (OPER1) Control Operator (OPER2) Local Plant OperatorCrew #Auxiliary Operators 3Turbine Hall Operator 2Aux bldg/Water Treatment 2Inside Control Room:Outside Control Room:Control Operator*** (OPER3) Crew composition and titles are plant specific Slide 9Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 3: Describe the PRA scenario (nominal context/base case scenario)Staffing Adequacy:Analysts walked through the scenario, including the parallel use of the fire procedure and confirmed staffing is adequate to perform this function (see table below).
oAssessment based on minimum staffing situation (i.e., night time). Daytime shifts would have, at the minimum, an additional Control Room Operator
.Crew MemberTotal Available Before Fire# assisting with fire*# Available for EOP actionsRequired for Bus AlignmentShift Manager 1 1 0 0Shift Supervisor 1Directingboth procedures 0STA 1 0 1 0Control Room Operators 2 1 1 1Plant operators 7 4 3 1*This includes members of fire brigade and staff occupied with FPs or otherwise occupied due to the fire Slide 10Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 3: Describe the PRA scenario (nominal context/base case scenario)Timing analysis: Fire ongoing throughout the scenario oDetailed fire modeling shows fire will last approximately one hour90 minutes for the total window (from initiator to core damage) based on a thermal hydraulic run for loss of AFW and a station blackout with one primary PORV stuck openTdelay= 35 min from reactor trip to receiving cue for action (step 17 in AOP 304) oBased on Simulator observation for a similar scenario for SBO it took operators 10 minutes to get through ECA 0.0 step 10Simulation based on non
-fire SBO, so add an 5 additional minutes to account for the initial coordination oBased on operator interviews, estimated additional 20 minutes to reach step 17 of AOP 304 Majority of the steps in AOP 304 are checking indicators, so < 1min per step on averageIncludes 5 minutes to account for AOP/FP meeting to coordinate Slide 11Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 3: Describe the PRA scenario (nominal context/base case scenario)Timing analysis (con't):
- Treqd= 22 min for diagnosis and execution oDiagnosis and SS approval ~2 minutes oThe action to locally remove power from the Train B ESF Loading Sequencer is trained on using Job Performance Measure (JPM) 12654
-Align ALT Feed Breaker. This JPM has a time requirement to be able to complete the local portion of the actions within 15 minutes, and this has been verified by observations of the JPM. The timing starts once the operator is given the instructions to perform this action and ends once the MCR action had been complete (end of step 18). As part of this JPM the operators train on putting on flash gear which is required to locally remove power from the Train B ESF Loading Sequencer. The flash gear is stored in a cabinet at the entrance to the relay room.
oAfter the operators complete the local action they will need to return to the control room to tell the control room operators they were successful.This additional travel time is expected to take 5 minutes. Under ideal conditions the Local Plant Operator could use the phone to call the control room.However, for fire, no cable tracing was performed on the phone lines so the telephones are assumed to unavailable.Radio unavailable during SBO.
Slide 12Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 4: Define HFE and Unsafe ActionsOperator fails to manually align 115kV power OrOperator fails to initiate manual alignmentOperator fails to properly align power HFE UAs OrFailure to locally remove power from ESFLS (step 17)Failure to close breaker in MCR (step 18)
Slide 13Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 4: Define HFE and Unsafe ActionsHFE: -Operator fails to manually align 115kV power (alternate power source) given an SBO.
-HFE defined as part of previous steps of Fire HRA process (Identification and Definition) but unsafe actions must be defined here if applicable.Cues:Multiple Indications of Loss of Buses1DA and 1DB with EDG not Available. SS makes call to power 1DA after buses have been inspected.AOP-304, Step 17: Locally remove from the Train A ESFLS (Local, Skill-of-Craft action).AOP-304, Step 18: Energize XSW1DA from the normal power source (MCR, proceduralized action):
-Ensure BUS 1DA XFER INIT Switch is in OFF
-Close BUS1DA ALT FEED Breaker
-Verify BUS 1DA potential lights are energized Slide 14Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 4: Define HFE and Unsafe Actions
- Unsafe Actions:
oControl room crew actions:
1.Fails to initiate manual alignment (EOO) 2.Fails to close breaker in MCR (to properly align alternate power) (EOC) a)Fails to recover from EOC (long time window, immediate feedback)oLocal operator actions:3.Fails to locally remove power Train A ESFLS (only credible failure mode is EOC) a)Fails to recover from EOC (with no local feedback available)
Slide 15Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDPotential Failure Modes and RecoveryUnsafe actions:
1.Control room crew fails to initiate manual alignment (EOO):Given the nature of the action and the training, it is unlikely that the crew will skip either step 17 or step 18, but it is possible that sufficient distractions (and other factors elongating the timeline) exist that the crew could fail to complete the action in time 2.Control room crew fails to close breaker in MCR (to properly align alternate power) (EOC)This unsafe action is not considered further because there is a very high potential for recovery, e.g., -Good cues for recovery
-Long Time Frame (35 minute time available for recovery)
-Fire extinguished by this point in time Slide 16Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDPotential Failure Modes and Recovery (cont.)Unsafe actions (continued):3.Local operator fails to locally remove power Train A ESFLS (only credible failure mode is EOC), AND3a.Local operator fails to recover from EOC (with no local feedback available)EOC: -Well proceduralized/skill
-of-craft step with good training
-EOC failure modes may include: Open wrong switch (fail local action)
-Diagnosis is largely performed by CR operators; plant operators must simply execute the required actions and report back to CR (for purposes of coordination)Recovery of EOC: In this case, there is no feedback available to the local operator that the wrong action was performed. Clear indications in the MCR that the ESFLS signal has not been cleared; the local operator will not get this feedback until he returns to the MCR to report back. After being notified that the wrong action has been performed, the local operator must return to the location of the ESFLS switch.
Slide 17Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDSteps 5-8: Understanding the Context (Iterative Process)Step 5: Identify Potential VulnerabilitiesStep 6: Search for Plausible Scenario VariationsStep 7: Evaluate Potential to RecoverStep 8a: Create Operational Story/StoriesStep 8b: Numerical Assessment Slide 18Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDGroup ExerciseBreak into groups and identify factor that could:
-Create potential vulnerabilities in the crew's ability to respond to the scenario(s) of interest and increase the likelihood of the HFEs or UAs-Failure modes (i.e., how can the scenario go wrong?)
-Lead to variability in crew response
- You may want to consider the following
-Division of Labor/Workload
-Procedures
-Training-Complexity
-Environment
-Special Requirements (e.g., keys)
-Stress due to Fire
-Communication
-Crew Coordination
-Variations in Timing
-Variation in Crew Characteristics Slide 19Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDGroup Exercise (2)Which factors are drivers? [Error Forcing Contexts]
-Note: Normally this would be done with the input of those knowledgeable of the plant and crews (e.g., operators, trainers) and any assumptions would be verified against the plant's operations Slide 20Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDPotential VulnerabilitiesTraining: Operators trained on procedures, including applicable alternative actions. Non
-fire SBO scenarios are common in training and "Align ALT Feed Breaker" is a Job Performance Measure which is trained on bi
-annually. Annual training on Fire Procedures. Trained as crew on SBO, not single operator. Fire Procedure training may not include doing the procedures in parallel.Parallel Procedures: The fire is ongoing during this scenario, so a portion of the staff will be unavailable to help with the EOPs as they will be in the fire procedures. Through operator talk
-throughs verified that adequate personnel are available for the necessary actions in this scenario. While operators will be going through two procedures in parallel (FP and EOP), the relevant steps of the FP have been examined and do not conflict with the EOP actions. While the Control Room Operators will be operating in parallel, the Shift Supervisor's attention will be split and he is a key decision point at several places in the procedure.
Slide 21Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDPotential VulnerabilitiesComplexity:Local action to remove power from ESFLS is a simple, skill-of-craft action.Environment:
oAvailability and Accessibility: Given location of fire and layout of plant, the relay room is accessible and there is no degraded environment (e.g., no smoke) in the relay room or en route to the relay room.
oVisibility: Given a SBO event, lighting will be significantly reduced (i.e., flashlights and/or emergency lighting). Training discusses these conditions.
oHeat/Humidity: Normal -fire effects do not reach this area, however, after some time (>action window) there could be a rise in temperature due to SBO.*Stress due to Fire: Some stress due to on
-going fire and related distractions.
Slide 22Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDPotential Vulnerabilities
- Communications: Communication lines impacted by SBO (no radios) and landlines potentially impacted by fire (no cable tracing). Timeline adjusted appropriately.
oPrevious steps in the ECA/AOP (e.g., local actions such as step 13) might cause delays due to extra time required for communication, delaying the cue (step 17).
oGenerally, local plant operators have to travel back to MCR to report
- Efficiency of crew coordination:
oCrew variations that could result in variability in the time to perform actions and effectiveness of communication back to control room.
oToo much focus on fire.
o"Weaker" crews.
- Special Requirements:
oOperators will need key to access relay room; all doors locked on loss of power.
oChange in security configuration due to SBO may require operators to take a different pathway or some doors which would otherwise be open may now be closed and locked. Not all operators have all keys.
Slide 23Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (6 Steps Overview)1: Discuss HFE and possible influences / contexts using a factor "checklist" as an aid2: Identify "driving" influencing factors and thus most important contexts to consider (e.g., operational story)3: Compare these contexts to other familiar contexts and each expert independently provide the initial probability distribution for the HEP based on a common calibration scale.4: Each expert discuss and justify their HEP5: Openly discuss opinions and refine the HFE, associated contexts, and/or HEPs (if needed)
-each expert independently provides HEP (may be the same as the initial judgment or may be modified)6: Arrive at a consensus HEP for use in the PRA Slide 24Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (Operational Story)
- Not limited to one operational story, particularly if the analysts have identified multiple credible contexts [EFCs] that need to be examined separately.
- A full operational scenario description, or "operational story," including accident progression and as many "bells and whistles" as are reasonable, such that operator trainers can "put themselves into" scenario.
-In quantification, you will be asking them, "what would your crews do in this situation?" The resulting operational scenario description may include:
-Additional plant conditions that will need to be quantified as part of the HFE (unless accident sequence analyst wants to revise event trees or fault trees).
-Distinctions on timing of plant behavior (that might need to be addressed as part of the HFE, unless logic is revised).
-Instrument or indication issues (including failures) that will need to be reflected (for fire, might be explicitly part of PRA model, or may not).
-Different possible procedure paths or response strategies that operators might rationally take.-Reasons why operators might take different procedure paths.
-Credible recovery actions
.
Slide 25Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (Operational Story, UA1)Operator Fails to Initiate Manual AlignmentPossible factors/sub
-scenario to explore with experts in:Staffing variations: can be two sub
-cases if large impact on crew performance
-Night time, minimal staffing (2 Control Room Operators)
-Day time, normal staffing (3 Control Room Operators)Crew variations, such as these two extremes in possible timing outcomes:
-Methodical crew that is good at taking time to work through the procedures and talk through potential conflicts. The crew works well as a team and rely on each other a lot. Training is done as a team on both the non
-fire SBO procedures and the fire procedure, so the Control Room Operators are a bit slower in working through their respective procedures when they are done in parallel, depending heavily on the Shift Supervisor for coordination, OR
-Aggressive crew, good at planning ahead, working fairly autonomously but coordinating when needed. Efficient at parallel procedures. Weak team members, i.e., OPER1 is struggling to keep pace with the rest of the team. There may or may not be an OPER3 that is available to look at boards and help with EOPs and/or FPs.
Slide 26Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (Operational Story, UA1)Operator Fails to Initiate Manual AlignmentPossible factors/sub
-scenario to explore with experts:Variations in SS experience, command & control style, & so forth, e.g.,-SS's first actual fire and, because it is a fairly big fire, he gets very focused on fire and becomes less cognizant of timeline or becomes a bottle neck for key decisions.
-SS calm under stress and has no problem coordinating the two procedures. Team is working at a fairly fast pace and multi
-tasking well (e.g., dealing with distractions), but working at the top of their capacity. Timing Variations:
-Delays in previous steps due to combination of radio unavailability and operators having to "hunt down" appropriate keys due to change in security configuration for SBO.Other:-Fairly significant fire (lasts 60 min), so there are many distractions (e.g., failed indicators and/or spurious indicators not directly relevant to this HFE, but may take time/attention away from operators)
-End of shift fatigueOverall, explore what factors (e.g., "slow crew" and other delays), result in crew missing timeframe to take action.
Slide 27Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (Operational Story, UA3/3a)Fail Local ActionPossible factors/sub
-scenario to explore with experts:Unsafe action #3 (EOC):Training of non
-fire SBO only; JPM timing based on average crew time, but accounts for many Local Plant Operators to be available to help with the procedure. With only two Local Plant Operators available for the EOP/AOP, the operator in question may be fatigued from rushing around and performing the higher workload.Timing Variations:
-Delays in previous steps due to combination of radio unavailability and operators having to "hunt down" appropriate keys due to change in security configuration for SBO.Given fast pace and general stress, the Local Plant Operator may feel rushed and open the wrong switch Slide 28Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (Operational Story, UA3/3a)Fail Local ActionPossible factors/sub
-scenario to explore with experts:Unsafe action #3a (Failure to recover EOC):Staffing:-Variations in staffing not applicable to this failure mode (i.e., 2 or 3 CROs)
-2 Local Plant Operators available for assistance with this actionRecovery includes:
-Diagnosis of problem (good cues); 5
-10 minutesClear indications in the MCR that the ESFLS signal has not been cleared.
-Action time (including travel time)20-25 minutes because, while OPER1 knows right away that the ESFLS switch has not been cleared, he has to wait until the Local Plant Operator gets back to re-dispatch him to perform the local action. Need to account for travel time and time to perform the local andMCR actions.Fire is extinguished at this point.Adequate time for recovery 35 minutes required compared to the nominal 55 minutes available.
Slide 29Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDLogic of Failure ModesOperator fails to manually align 115kV power OrOperator fails to initiate manual alignmentOperator fails to properly align power OrFailure to locally remove power from ESFLS (step 17)Failure to close breaker in MCR (step 18)EOOEOCPlant layout examined in closer detail and contribution due to EOC considered -4 even discounting recovery of local action Slide 30Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDQuantifying Unsafe Action #1 (EOO)Driving factors:
-Slow crew-Excessive travel time for local actions extends timeline
-Mismatch between training (heavy interaction as crew) and reality (relatively autonomous, especially with minimum staffing)
-Distractions and stress due to fire
-SS is a funnel point for decisionsStaffing identified as a driver, so can split this scenario into 2 contexts:
-2 Control Room Operators available (Minimal Staffing): 33%
-3 Control Room Operators available (Normal Staffing): 67%
- Given "slow and careful" crew, they are unlikely to make a mistake in the action, but may come close to missing the action time window (see next slide).
- "Nominal" case accounted for by shape of the distributions
-If heavily weighted to left, positive or nominal factors more likely; having the right combination of "driving" factors is less likely Slide 31Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDTiming VariationsTiming is a driving factor in the Operational Story
-Would ask "experts" to develop a more detailed analysis of potential variations in timing (e.g., more explanations, more developed description of possible scenario variations, detailed histogram of probability of timing for both arrival at Step 17 and performance of required actions)
-Might separate HFE into two or more separate HFEs to address different timing for different scenariosVariations in timing due to factors discussed earlier:
-Could there be variations in the scenario (e.g., additional minor distractions in working through procedure?"Experts" estimate minor variations:
10-15 additional minutes to get to critical procedure step
-Could there be variations in the time to perform (especially with different crews, availability of equipment, communication)?"Experts" estimate minor variations:
5-10 additional minutes to perform critical procedure stepsOverall, could reduce time for recovery to as little as 8 minutes. This, however, does not jeopardize the timeline for the actions themselves.
Slide 32Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (Numerical Assessment)Combining Multiple ContextsOnly one dominant UA, so this formula simplifies to:Two distributions need to be estimated oMinimal Staffing oNormal Staffing
- Only one distribution will be estimated here for illustrationj i j j i i S EFC UA P S EFC P SHFE P),l(*)l()l()(i i S EFC UA P SHFE P),l()l(
Slide 33Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (Calibrate Experts)CircumstanceProbabilityMeaningOperator(s) is "Certain" to fail1.0Failure is ensured. Allcrews/operatorswould not perform the desired action correctly and on time.Operator(s) is "Likely" to fail~0.55 out of 10 operators would fail. The level of difficulty is sufficiently high that we should see many failures if all the crews/operators were to experience this scenario.Operator(s) would "Infrequently" fail~0.11 out of 10 would fail. The level of difficulty is moderatelyhigh, such that we should see an occasional failure if all of the crew/operators were to experience this scenario.Operator(s) is "Unlikely" to fail~0.01 1 outof 100 would fail. The level of difficulty is quite low and we should not see any failures if all the crews/operators were to experience this scenario.Operator(s) is "Extremely Unlikely" to fail~0.001 1 outof 1000 would fail. This desired action is so easy that it is almost inconceivable that any crew/operator would fail to perform the desired action correctly and on time.Note: These values are meant as calibration points, not discrete values. The 1E
-03 values is not meant to be a lower bound.
Slide 34Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (Numerical Assessment)Very structured, facilitator led, expert opinion elicitation process oleads to consensus distributions of operator failure probabilitiesConsiderations in elicitation process (covered in NUREG
-1880): oForming the team of experts (include experts familiar with important relevant factors during fire conditions, operator trainers, etc.)
oControlling for biases when performing elicitations oAddressing uncertaintyDistribution characteristics:
othe 99th percentile is the HEP for the worst coincident (but not too unlikely) set of negative influences representing a very strong EFC othe 1st percentile is the HEP for the best coincident set of positive influences representing a weak EFC (actually a very positive context odependency considerations embedded ouncertainty distribution explicitly consideredFor this illustrative example an HRA SME was used to derive the HEP; this would not normally be sufficient for an actual quantification.
Slide 35Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: QuantificationA tip for expert elicitation facilitators:
-In order to get "experts" to better access their knowledge (i.e., not just remember recent history), you can use examples from real events (i.e., "stories") to illustrate how operators can do "surprising" things (but for good reasons.You know that you've succeeded in getting access to this deeper knowledge when the "experts" start exchanging stories (e.g., "do you remember when 'Charlie' -.?" "I can remember a time or two kind of like that-.")
Slide 36Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 8: Quantification (Bases for Consensus Distribution)Bases for Consensus Distribution:
oUnder normal circumstances, the action is "Extremely Unlikely" to fail, but the shortened time frame due to no radio communication in combination with potential coordination complications from the fire may produce some difficulties for the crews.Holistically, on average the action was determined to be "Extremely Unlikely" because actions are well trained, proceduralized/skill
-of-craft, long timeline, a high potential for recovery and cues are clear so little potential for confusion or mis
-direction.Probability capped at 1E
-04Worst case falls between "Unlikely" to fail and "Infrequently" fails because even in the worst case they still have buffer time.Tails: effectiveness of crew collaboration, specifics of timingPercentilesAnalyst 1 st 10 th 25 th 50 th 75 th 90 th 99 thLarry0.000010.00010.00070.0010.0050.0070.01Moe0.00010.00030.0010.0050.0070.030.07Curly0.000010.000050.00070.0030.0050.010.05Consensus 1E-04 3E-04 1E-03 3E-03 5E-03 1E-02 5E-02 Slide 37Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDStep 9: Incorporating HEP into PRA
- When quantifying a scenario with multiple contexts, need to combine weighted distributions. Discrete distributions can be combined using a convolution:
-Recommend using a statistical software package (e.g., Crystal Ball)
- Depending on the PRA needs, you may:
-Provide the entire consensus histogram as your answer.
-Need to develop a mean value for the distribution using a software tool (e.g., Crystal Ball).
- NUREG-1880 provides some guidance and cautions on the development of mean values.
Slide 38Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDWhat if-What if communication was not impacted, how would the analysis change?What if there were not clearly enough people to complete the actions, how would the analysis change?What if the operators had to take a detour that comes close to the fire?
Slide 39Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDScoping Analysis of Fire + SBO Slide 40Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDReview of HFEInitial Conditions: Single unit two loop PWR with two trains of electrical power. Steady state, full power operation. Night shift with minimal staff onsite.
-No out-of-service unavailability pertinent to this scenarioInitiating Event: Fire in turbine room causes SBOHFE: Operator fails to manually align 115kV (alternate power) power on loss of both buses and EDGs fail to start.
Slide 41Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDMinimum Criteria 1.Procedures
-Plant procedures covering each operator action being modeled-Support both diagnosis & execution of the action 2.Training -on the procedures and the actions 3.Availability and Accessibility of Equipment
-Key to ESFLS Panel needed, but available in MCRLocal action (step 17) is skill
-of-craft; MCR action (step 18) well proceduralized.Regular training on non
-fire SBO, including alternative actions. Training on FPs.Key to ESFLS Panel needed, but available in MCR. Flash gear needed, but available locally.
Slide 42Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDFeasibilityTiming analysis
- o T sw: Assume 90 minutes for the total window (IE to core damage) based on a thermal hydraulic run for loss of AFW and a station blackout with one primary PORV stuck open.
o Tdelay= 35 min from reactor trip to receiving cue for action (step 17 AOP 304) o T cog+ T exe= 22 min for diagnosis and executionFeasible? Yes time available (90 minutes) is greater than time for action (55 minutes).
Slide 43Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDTime Margin T o Tdelay = 35 minInitiating eventCuereceivedCrewdiagnosiscompleteActioncompleteAction nolongerbeneficial Tavail = 55 min Tcog + Texe = 22 min Tsw = 90 min%150%100*22 22 55%100*)()( Margin Timeexecog execogavail T T T T T Slide 44Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDAssessing Key Conditions & PSFs within the Scoping Flowcharts
- How well the procedures match the scenario
- Response execution complexity
- Timing of cues for the action relative to expected fire suppression time
- Action time window
-Short time window = 30 minutes or less
-Long time window = greater than 30 minutes
- Level of smoke and other hazardous elements in the action areas
-Need for special equipment (e.g., SCBA)
-Impairment of vision or prevention of the execution of the action*Accessibility Slide 45Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDHFE BreakdownOperator fails to manually align 115kV power Or HFEFailure to locally remove power from ESFLS (step 17)Failure to close breaker in MCR (step 18)INCREXCRWhile the HFE can be broken down into multiple steps (INCR and EXCR), because this is defined as one HFE (based on the fact it is one diagnostic step), we will quantify this HFE using the EXCR tree because it is more conservative.Quantified at this level Slide 46Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDSearch SchemeGo to EXCRScoping Analysis:
- Define HFE:Failure to locally remove power from ESFLS (step 17). This includes both the diagnosis and the execution.
- Does it meet the minimum criteria? Yes 1)Procedures are available 2)Training is performed on the procedure 3)The key to the Relay Room is determined to be accessible
- Is the action Feasible?Yes 1)Demonstrated sufficient time to perform action
- Selection Scheme
- 1)D1: Entry criteria are met 2)D2: command and control in MCR 3)D3: primary cues/instrument not spuriously affected by fire 4)D4: procedures match the scenario 5)D5: some actions within MCR, but key actions outside MCR, so use EXCR tree 6)D6: procedures available/skill
-of-craft 7)GO TO EXCR TREE Slide 47Fire HRA -ATHEANA ExampleFire PRA Workshop 2019, Rockville, MDHFELocal Action
-D22: Fire is ongoing
-D26: Area accessible and no fire in vicinity.
-D27: Time window is greater than 30 min (90-35=55min).-D33: High complexity in execution due to multiple step/locations
-D37: No smoke.
-Time Margin >100%