ML17263B165

PSA R&D: Changing the Way We Do Business

ML17263B165
Person / Time
Issue date:	09/20/2017
From:	Nathan Siu NRC/RES/DRA
To:
References
Download: ML17263B165 (42)

Text

PSA R&D: Changing the Way We Do Business?*

N. Siu ANS International Topical Meeting on Probabilistic Safety Assessment (PSA 2017)

Pittsburgh, PA September 24-28, 2017

• The views expressed in this presentation are not necessarily those of the U.S. Nuclear Regulatory Commission

2 Introduction

• This talk

- Provides some personal, pragmatic views on R&D intended to support Risk-Informed Decisionmaking (RIDM) in the nuclear industry

- Addresses formulation, practice, and support of R&D;* not a catalog of topic areas

- Presented in two parts

• General R&D context

• Remarks for different PRA communities (developers, analysts, users)

• And about that title

• In the remaining slides, the focus on RIDM support is understood

R&D CONTEXT AND PERSPECTIVES

4 R&D: A Broad Enterprise

• Provides technical advice, tools, and information to meet organizational needs

• Broader than development of theory and methods

• At NRC, includes activities to support

- Decisions on specific issues

- Infrastructure (standards, etc.)

- Understanding (context for decisions)

- Communication with stakeholders Methods, Models, Tools, Databases, Standards,

Guidance, General Knowledge Supporting Analyses Decision R&D

5 with Multiple Stakeholders Developers Analysts/

Reviewers Users 5

6 Personal Background => Perspectives Developer Analyst/

Reviewer User (Advice) 6

7 Dynamic R&D Environment

• Needs

- New events/findings

- New technologies/designs

- Increasing demands on PRA technology

- Risk-informed applications (current plants)

• Resources

- Computational infrastructure

- Demographics

- Budget

8 Dynamic R&D Environment - Applications

9 Dynamic R&D Environment - Demographics 9

10 Dynamic R&D Environment - RES Budget

11 Assumptions We hold these truths to be self-evident

• Risk-related Regulatory R&D (R4&D) has been and continues to be a necessary component of a healthy nuclear enterprise

• R4&D products need to be implemented to affect safety

• There will continue to be challenges in R4&D

• No one community has all of the answers Regulations and Guidance Operational Experience Licensing and Certification Oversight Decision Support (e.g., research, PRA)

R4&D STAKEHOLDERS

13 Discussion Framework Sub-communities that are the focus of remarks

• Developers: academic community

• Analysts/Reviewers:

PRA model builders

• Users: decision makers Developers Analysts/

Reviewers Users

14 Developers

• Typical challenges

- Academic contribution

- Nexus between personal and external interests

- Support (esp. with declining budgets!)

• R4&D solutions include

- Frameworks, methodologies, conceptual demonstrations

- N+1 projects

15 Technology-Driven vs. Issue-Driven R4&D Is it Hammer Time?

• A common (and valid) research strategy

{New Tech*} + {Interesting Problem} => {Research Topic}

• (Of course) beware

- Force-fitting the problem to match the new tech

- Problems not truly requiring the new techs special capabilities

- Problems of tangential importance to major risk drivers

• Technology = {methods, models, tools}

16 Big Data: Prognostics and Reliability An N+1 Example

• Concept: use field data and physics of failure models to anticipate failures and develop prevention strategies

• Formula 1 races

- Heavily instrumented cars

- Engineering models

• Real-time support during race

• Empirically calibrated through testing

• ~100 supporting staff at home office

- Miniscule performance => major effects A. Gilbertson, 2016

17 23 sec video: 2016 US Grand Prix Big Data: Prognostics and Reliability COTA Turn 15 Entry speed ~210 km/h Speed at apex ~84 km/h Braking distance ~30m Braking power ~800 kW Load ~3g Race footage courtesy of A. Gilbertson

18 Big Data: Prognostics and Reliability 120+ sensors (car and driver) 1000-2000 wireless channels Low latency - o(ms) 2 GB/lap, 3 TB/race Thermal cameras during qualification

19 Big Data: Prognostics and Reliability Will it work for R4&D?

• Engineering models

- Empirically calibrated through testing

- Predict performance and wear over time

- Used to develop/modify race strategies

- Models dont cover all factors

• Road debris

• Other drivers

• In a NPP application, what risk-significant failures would be caught by an analogous system?

20 Big Data: Prognostics and Reliability Some Old But Interesting Events Year Plant Feature 1975 Browns Ferry Cable fire affects multiple units 1985 Hatch HVAC water falls into MCR panel; SRV cycles, sticks open 1993 Cooper External flood, one evacuation route blocked 1997 Fort Calhoun Steam line rupture, intermittent electrical grounds 1999 Blayais High wind and external flood affect multiple units, site access 2001 Maanshan High energy arc fault, station blackout

21 Big Data: Prognostics and Reliability Multi-Unit Precursor Events Date Plant Description 6/22/82 Quad Cities LOOP, Maintenance 8/11/83 Salem LOOP, Clogged screens 7/26/84 Susquehanna SBO, Bkr mis-aligned 5/17/85 Turkey Point LOOP, Brush fires 7/23/87 Calvert Cliffs LOOP, Offsite tree 3/20/90 Vogtle LOOP, Truck hit support 8/24/92 Turkey Point LOOP, Hurricane 12/31/92 Sequoyah LOOP, Switchyard fault 10/12/93 Beaver Valley LOOP, Offsite fault 6/28/96 LaSalle Trip, Foreign material in SW Tunnel 6/29/96 Prairie Island LOOP, High winds Date Plant Description 8/14/03 6 Sites LOOP, NE Blackout 6/14/04 Palo Verde LOOP, Offsite fault 9/25/04 St. Lucie LOOP, Hurricane 5/20/06 Catawba LOOP, Switchyard fault 3/26/09 Sequoyah LOOP, Bus fault 4/16/11 Surry LOOP, Tornado 4/27/11 Browns Ferry LOOP, Winds/tornadoes 8/23/11 North Anna LOOP, Earthquake 3/31/13 ANO LOOP/Trip, Load drop 4/17/13 LaSalle LOOP, Lightning 5/25/14 Millstone LOOP, Offsite fault

22 Big Data: Prognostics and Reliability How Do Things Fail? (Service Water) the station declared all Core Standby Cooling Systems (CSCS), Emergency Core Cooling Systems (ECCS),

and Diesel Generators (DG) inoperable due to foreign material identified on the floor of the service water tunnelAlthough the systems were declared inoperable, they were available. The foreign material was an injectable sealant foam substance which had been used in the Lake Screen House (LSH) to seal water seepage cracks. (LER 373/96-008) manual reactor shutdown conservatively initiated... due to concern for the safety and well being of a diver working in the Unit 2 circulating water pump house discharge piping communications with one diver was [sic] lost and the retrieval efforts by a second and third diver were initially unsuccessful in reestablishing contact.. The plant equipment and systems worked as designed. The divers were unharmed. (LER 266/00-001)

23 Big Data: Prognostics and Reliability An Alternate/Complementary Line of R4&D?

• Searching is fundamental to PRA:

- First question of risk triplet: What can go wrong?

- PRA Procedures Guide and ASME/ANS PRA Standard

• Sparse data, beyond design-basis concerns

=> imagination needed

• Operational experience can fuel, temper, and support imagination

- Massive, unstructured databases

- How to better use?

Investigation Committee on the Accident at Fukushima (7/23/2012): TEPCO lacked a sense of urgency and imagination toward major tsunami, which could threaten to deal a fatal blow to its nuclear power plants.

E. De Fraguier, Lessons learned from 1999 Blayais flood: overview of EDF flood risk management plan, U.S. NRC Regulatory Information Conference, March 11, 2010.

24 Big Data: Prognostics and Reliability Mr. Watson come here, I want you."

• Advanced knowledge engineering technologies can help

• IT challenges include:

- Faulty data

- Embedded structure

- Machine learning

- Speed

• PRA user challenges include:

- Use case identification, specification

- Working with IT: develop, test solutions ICA 2.2 Intelligent Personal Assistants Watson

Searcher, Explorer

Aide, Oracle

Tool, Toy

Servant, Partner

25 Analysts/Reviewers

• Typical challenges

- Need near-term solutions: heavy time/budget pressure

- Huge problem size and complexity

- Multiple technical communities/cultures

- State of technology: Too much/little diversity, Holes

• PRA solutions include

- Tried and true; reluctance to try new approaches

- Engineering judgment

- Completeness uncertainty Reverse ageism in the PRA community

26 Dynamic PRA: What and Why

• Literal view: explicit treatment of time

• More broadly: explicit treatment of dynamic phenomena (simulation-based)

• Potential advantages include

- Direct ties to other science/engineering models (with usual V&V)

- Elimination of intermediate modeling approximations

- Natural language for interdisciplinary enterprise

- Consistent with external tech world

- Most direct approach to some tough problems

- Academically rewarding

27 Dynamic PRA: Why Not

• Dynamics not key issue for many problems

• Models can be complex:

- Resource-intensive (construction, validation, computation, analysis/sensemaking)

- Inscrutable (at least to practitioners)

- Vulnerable to sub-model applicability limits

- Massive output, but information?

• Long gestation, in early phase of maturity

- Starting to expand from academic centers

- Few real applications of full-power tools

- Reward system has likely inhibited search for simple but practical applications The Aldemir Tank

28 Dynamic PRA: An Opportunity Missed?

• Object-oriented simulation: long history, fully developed general technology

- Operations Research and System dynamics

- Military simulations and supporting tools

- Considered but rejected for general PRA quantification (aleatory uncertainty)

- Highly limited, demonstration-oriented NPP PRA applications (power recovery)

• Basis for advanced Vulnerability Assessment tools

• Well-suited for FLEX?

29 Fire PRA: From Research to Application

• Following 1975 Browns Ferry fire

- NRC supported fire PRA R&D at UCLA and RPI

- Methodology and tools used in industry-sponsored PRAs

- Approach used in NUREG-1150

- Formed basis for guidance

- Currently continuing improvements on piece parts

• Success factors

- Real problem

- NRC and industry involvement

- Researchers directly involved in application

30 Users

• Decision maker challenges include

- Managing/leveraging resources

- Prioritizing needs and activities

• Short-vs. long-term

• Organizational

• Multiple stakeholders

- Difficult decisions

• Technical complexity

• High uncertainty, diverse views

• Multi-variate

• A key adviser challenge: communication Aleatory

31 Communication Challenges: Model Uncertainty

32 Communication Challenges: Model Uncertainty

• Model uncertainty is important and real

• Some R4&D questions

- Ensemble better than best estimate + sensitivities? Under what conditions?

- How important is it to understand the technical reasons for model-to-model variability?

- Ensemble doesnt necessarily capture (revealed) reality: what to do when all options have significant costs and potential consequences?

If anything on these products causes confusion, ignore the entire product.

http://my.sfwmd.gov/sfwmd/common/images/weather/plots.html

33 Communication Challenges: Gatekeeping

• Pre-Fukushima WGRISK report on external hazards PSA

- Varied country responses

• Some treatment for internal events (CCF, LOOP, LOHS)

• Research on some hazards (seismic, typhoon)

• Some with no special considerations

- Topic, findings, or recommendations not provided in Conclusions

• Omission: lack of actionable message

• R4&D question: When (under what circumstances) should we boost the signal? How? [see Blayais]

34 Communication Challenges: Unintended Messages How not to start an R4&D program ID Fire PRA Issue ID Fire PRA Issue I1 Adequacy of fire events database P1 Circuit interactions I2 Scenario frequencies P2 Availability of safe shutdown equipment I3 Effect of plant operations, including comp measures P3 Fire scenario cognitive impact I4 Likelihood of severe fires P4 Impact of fire induced environment on operators E1 Source fire modeling P5 Role of fire brigade in plant response E2 Compartment fire modeling R1 Main control room fires E3 Multi-compartment fire modeling R2 Turbine building fires E4 Smoke generation and transport modeling R3 Containment fires H1 Circuit failure mode and likelihood R4 Seismic/fire interactions H2 Thermal fragilities R5 Multiple unit interactions H3 Smoke fragilities R6 Non-power and degraded conditions H4 Suppressant-related fragilities R7 Decommissioning and decontamination B1 Adequacy of data for active and passive barriers R8 Fire-induced non-reactor radiological releases B2 Barrier performance analysis tools R9 Flammable gas lines B3 Barrier qualification R10 Scenario dynamics B4 Penetration seals R11 Precursor analysis methods S1 Adequacy of detection time data R12 Uncertainty analysis S2 Fire protection system reliability/availability O1 Learning from experience S3 Suppression effectiveness (automatic, manual)

O2 Learning from others S4 Effect of compensatory measures on suppression O3 Comparison of methodologies S5 Scenario-specific detection and suppression analysis O4 Standardization of methods 42 = 1 [year]

From: N. Siu, J.T. Chen, and E. Chelliah, Research Needs in Fire Risk Assessment, NUREG/CP-0162, Vol. 2, 1997.

35 Closing Remarks

• R4&D has helped change the way we do business

• Continuous improvement + changing times => some suggestions

- Developers: Dig a little deeper

- Analysts/reviewers: Give em a chance

- Advisers: Be like Sherlock Theres gold in these hills

36 Give Em a Chance [2]

Grizzled Vets Young Pups

ADDITIONAL SLIDES

38 Dynamic R&D Environment - Industry Needs*

• Finding Closure

• Risk-Informed SSC Categorization (50.69)

• Risk-Informed Completion Times (TSTF-505)

• Fire PRA Realism

• Methods Vetting

• Aggregation

• Realism in Reactor Oversight Process

• FLEX in Risk-Informed Decision Making

• From public meeting of NRC and Industry Risk-Informed Steering Committees, April 13, 2017 (ML17104A315)

39 0

500 1000 1500 2000 2500 3000 3500 4000 4500 0

200 400 600 800 1000 1200 1400 NRC Budget (FY 1976 - FY 2017)

Actual ($M)

Fiscal Year Budget ($M)

Data from NUREG-1350 (NRC Information Digest)

Staff (FTE)

40 Big Data: Prognostics and Reliability How Do Things Fail? (Power) both units tripped automatically from 100% power following a Loss of Offsite Power (LOOP) event.

The event began when a fault occurred internal to a current transformer associated with one of the switchyard power circuit breakers. A second current transformer failure, along with the actuation of differential relaying associated with both switchyard busses, cleared both busses and separated the units from the grid... The root cause analysis determined that certain switchyard relay tap setting changes were not implemented. (LER 413/06-001)

Units 1 and 2 received an automatic reactor trip on reactor coolant pump (RCP) buses undervoltage. A loss of Common Station Service Transformer (CSST) C caused a loss of power to two unit boards on each unit that feed RCPs The cause of the bus fault was determined to be degraded bus bar insulation and water intrusion into the CSST D secondary bus duct. (LER 327/09-003)

41 Historical View on Searching PRA Procedures Guide, NUREG/CR-2300 (1983)

The search for dependent failures should be performed as described in Section 3.7 and incorporated as appropriate into the plant and system models.

A preliminary systems analysis can thus be a vital step in the search for initiators, helping to ensure completeness in the definition of accident sequences.

Another approach is to more formally organize the search for initiating events by constructing a top level logic model and then deducing the appropriate set of initiating events.

A systematic search of the reactor-coolant pressure boundary should be performed to identify any active elements that could fail or be operated in such a manner as to result in an uncontrolled loss of coolant.

a more formal search and documentation of all elements that depend on input from another source beyond the identified system boundary may be appropriate.

42 External Flooding: A Really Big Picture

• Sparse data and concerns with extrapolation => mechanistic analysis

• Daunting scale

- Regional analysis

- Human actions

- Besides flooding level: duration, debris, dynamic forces, warning time

- Multi-site impacts

• How can R4&D help?

- Multiple, heterogeneous evidence sets =>

uncertainty

- Demonstration of validity of more restricted representation