development of pipework system failure...

Development Of Pipework System Failure Rates

Where do the numbers come from and why

should we believe them?

By:

Karl N. Fleming, President

KNF Consulting Services LLC

Presented to:

CRA’s UK’s 5th Probabilistic Safety Analysis & Human Factors Assessment Forum

What lies behind the numbers? 17-18 September 2014

Scope of PRA Issues

n  Pipe failure rate issues in PRA n  Loss of coolant accident initiating events n  Internal flooding and high energy line

breaks n  Risk-informed inservice inspection n  Generic safety issue for emergency cooling

system blockage

PSA 2013 2

Approach to Solving Issues

n  Technical approach to piping reliability n  Sources of pipe failure data n  Treatment of uncertainty n  Key results and insights

PSA 2013 3

Two types of pipe failure data:

n  Type 1 Databases of events involving failure or degradation of piping system components with information on causes, failure modes, corrective actions etc.

n  Type 2 Estimates of piping system failure rates and rupture frequencies for PRA development and risk-informed applications

PSA 2013 4

Sources of Type 1 Failure Data n  EPRI-SKI Collaboration

n  SKI 96:20 Bush and Chockie records of events involving pipe failures

n  EPRI RI-ISI Project n  EPRI TR-111880 estimates of pipe failure rates and rupture

frequencies for RI-ISI

n  PIPExp Comprehensive database of worldwide NPP service experience developed by Bengt Lydell

n  OECD/NEA Pipe Data Exchange – international effort to collect and analyze pipe failure data

PSA 2013 5

PIPExp Database n  Evolved from EPRI-SKI collaboration n  Continuously updated since 1994 n  Summary available in “PIPExp-2014 High-Level

Summary of Database Content as of June 30, 2014, Sigma Phase Inc.”

n  Currently has nearly 10,000 pipe failure records n  Non-through wall defects (e.g., cracks and wall thinning) n  Small leaks resulting in piping repair or replacements; n  Leaks; n  Severance (i.e., pressure boundary failure due to external impact); n  Rupture (significant structural failure).

PSA 2013 6

PIPExp Plant Level Data Summary Showing Aging Effects

PSA 2013 7

PIPExp Distribution of System and Failure Mode

PSA 2013 8

Technical Approach To Pipe Failure Rate Estimation

PSA 2013 9

March 2011 PSA 2011 10

Pipe Rupture Frequency Model

ikt

M

kxikik

M

kikxixt IFRP

ii

∑∑==

==11

}{λρρ

ρixt = total rupture frequency for component i for rupture mode x at plant age t

ρikx = rupture frequency of component i due to damage mechanism k for rupture mode x

λik = failure rate of pipe component i due to damage mechanism k Pik{Rx|F} = conditional probability of rupture mode x given failure for pipe

component i and damage mechanism k Mi = Number of different damage mechanisms for component i Iikt = Age and Integrity management factor for component i and

damage mechanism k ;this factor adjusts the rupture frequency to account for plant age t and integrity management strategy which may be different than the components in the service data.

March 2011 PSA 2011 11

Markov Model Background n  Markov Model originally developed for EPRI RI-ISI

Program n  Applied to 26 plant specific RI-ISI programs in U.S.

and South Africa n  Applied to PBMR to support new ASME Code

development for in-service inspections n  Applied in NUREG-1829 LOCA frequency update n  Currently being applied to address CANDU feeder

pipe cracking issue n  Recently applied to LWRs to guide efforts to reduce

internal flood and HELB contributions to CDF

March 2011 PSA 2011 12

MARKOV MODEL OF PIPE ELEMENT

S

F

L

R

φ ω

µ

λ

ρF

ρL

S

F

L

R

SS

FF

LL

RR

φ ω

µ

λ

ρF

ρL

Pipe Element States

S – success, no detectable damageF – detectable flawL – detectable leakR - rupture

State Transition Rates

φ – flaw occurrence rateλ – leak failure rateρF – rupture failure rate given flawρL – rupture failure rate given leakω – repair rate via ISI examsµ – repair rate via leak detection

March 2011 PSA 2011 13

Modeling Impact Of NDE Inspections

n  Capture by ω: the repair rate for flaws

where: n  PFI = probability that segment element with flaw will be

inspected n  PFD= probability that flaw is detected given inspection n  TI = mean time between inspections n  TR = mean time to repair after detection

( )ω =+

P PT TFI FD

I R

March 2011 PSA 2011 14

Modeling Impact Of Leak Tests

n  Capture by µ: the repair rate for leaks

where: n  PLD= probability that leak is detected given inspection n  TI = mean time between inspections n  TR = mean time to repair after detection

µ =+

PT T

LD

LI R( )

March 2011 PSA 2011 15

BWR Recirculation Pipe LOCA Frequency from NUREG-1829

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

5 15 25 35 45 55

Plant Age (Years)

BW

R R

ecirc

ulat

ion

Pipi

ng L

OC

A F

requ

ency

/yea

r

No ISI/No Leak InspectionNo ISI/ Leak Inspection 1/Refueling OutageNo ISI/ Leak Inspection 1/WeekISI/Leak Inspection 1/Refueling OutageISI/Leak Inspection 1/Week

March 2011 PSA 2011 16

Application to Justify On-line Leakage Monitoring for PBMR

0

0

1

10

100

RIM Case 1: T= 27yrs -LWR RIM

RIM Case 2: T= 27yrs - No

RIM

RIM Case 3: T= 40yrs -LWR RIM

RIM Case 4: T= 40yrs - No

RIM

RIM Case 5: T= 40yrs -

Primary NDE

RIM Case 6: T= 40yrs -

OLLD - NoNDE

RIM Case 7: T= 40yrs -OLLD -

SecondaryNDE

RIM Case 8: T= 40yrs -OLLD -

Primary NDE

Plant Age (T) and RIM Strategy

Inte

grity

Man

agem

ent F

acto

r

LWR RIM:RIM = reliability and integrity management100% Leak Tested @90% POD every 1.5yrs50% NDE @50% POD every 10 yrsPOD = Probability of detectionNDE = Non-destructive examination

PBMR RIMOLLD - On-line leak detection with 90% POD in 24hrsSecondary NDE - 50%POD every 12yrsPrimary NDE - 90% POD every 12 years

PBMR RIM CasesLWR RIM Cases

Pipe Failure Rates For Internal Flooding PRA

PSA 2013 17

Guidance Available for IFPRA n  Guidelines for Performance of Internal Flooding

Probabilistic Risk Assessment* 1019194 Final Report, December 2009

n  Pipe Rupture Frequencies for Internal Flooding Probabilistic Risk Assessments* Revision 3 3002000079 Final Report, May 2013

n  ASME/ANS PRA Standard RA-Sb-2013, Part 3** * Available from EPRI ** Available from ASME or ANS

Scope of Pipe Failure Rate Cases in EPRI Report 3002000079

SystemReactor

Type Type ASME ClassNominal

Pipe SizesClass 3

Non-SafetyClass 3

Non-SafetyClass 3

Non-SafetyClass 3

Non-SafetyClass 3

Non-SafetyClass 3

Non-SafetyNormal System

Protected Against Water HammerSIR Outside Cont All N/A Class 3 4,10,24 (>2)

CCW and CST All N/A Non-Safety 24 (>2)BWR N/APWR N/ABWR N/APWR N/ABWR N/APWR N/ABWR N/APWR N/A

Piping Non-SafetyExpansion Joints Non-Safety

10,24 (>2)Non-Safety

BWR

PWR

2,4,10,24

Fire Protection All Non-Safety 4,6,24

HP Steam

LP Steam

Ext Steam

Service Water

Circulating Water All >24

River

Lake

Sea

River

Lake

Sea

FWC

Failure Rates And Rupture Frequencies For Different Piping Systems

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

PWRFeedwater

FireProtection

>6"

Circ Water >24"

PWRCondensate

BWR SeaSW Class 3 >

10"

ECCS Class2 >10"

BWR LakeSW Class 3 >

10"

CCW Class 3>10"

BWR Class 1Pipe toNozzle

Failu

re R

ate,

per

line

ar fo

ot-y

ear

All Failure ModesCatastrophic Failure Modes

Non Safety Class Safety Class

Pressure Boundary Failure Modes Considered in IFPRA n  Spray Events

n  Can screen out leaks < 1gpm n  Initial flood rates from 1gpm to 100gpm n  Typically within drain and sump capacities n  Damage is typically localized (within∼10ft)

n  Flood Events n  Would also involve sprays n  May involve pipe whip effects n  Initial flood rates from 100gpm to 2,000gpm (or lower limit based on system capacity) n  Typically in excess of drain and sump capacities n  Flood volume and damage dependent on time to isolate, flood propagation paths, etc.

n  Major Flood Events n  Initial flood rates > 2,000 and up to system capacity n  May involve pipe whip effects n  Well in excess of any drain or sump capacity outside containment n  Flood volume and damage also dependent on time to isolate, flood propagation paths,

etc. n  High Energy Line Breaks

n  Pressure and temperature effects n  Steam jet and pipe whip effects n  Inadvertent fire protection (sprinkler) operation n  Other effects of flooding

Pipe Failure Rates for Screening Analysis

PSA 2013 22

BWR River Site Class 3 Service Water Results – Baseline Results for 24” Pipe

1.0E-12

1.0E-11

1.0E-10

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

0.01 0.10 1.00 10.00 100.00

X, Equivalent Break Size (in.)

Freq

uenc

y of

Rup

ture

Siz

e G

reat

er th

an o

r Equ

al to

X (e

vent

s pe

r RO

Y-ft.

)

95%tileMean50%tile (Median)5%tile

March 2011 PSA 2011 24

Impact of Design Features to Reduce FP Susceptibility to Water Hammer

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

0.01 0.10 1.00 10.00 100.00

X, Equivalent Break Size (in.)

Freq

uenc

y of

Rup

ture

Siz

e G

reat

er th

an o

r Equ

al to

X (e

vent

s pe

r RO

Y-ft.

)

Current Study w/ WHCurrent Study no WHEPRI 1013141 FP NPS > 10"

Spray1-100gpm

Flood100-2,000gpm

Major Flood> 2,000gpm

Unfavorable Trend in FP Performance Since 2004

March 2011 PSA 2011 25

Impact of Integrity Management Strategies for Fire Protection Piping

1.0E-09

1.0E-08

1.0E-07

1.0E-06

1.0E-05

1.0E-04

0.01 0.10 1.00 10.00 100.00

X, Equivalent Break Size (in.)

Freq

uenc

y of

Rup

ture

Siz

e G

reat

er th

an o

r Equ

al to

X (e

vent

s pe

r RO

Y-ft.

)

Current Study w/ WHCurrent Study no WHEPRI 1013141 FP NPS > 10"Current Study No WH + Yearly Leak TestCurrent Study No WH + Quaterly Leak Test

Plant Level Flood Frequencies Added in Revision 3 of Failure Rate Data Report

n  Plant level flood events identified in U.S. service data from 1980-2011 encompass about 3,500 reactor years

n  Flood events analyzed by system, building, event type, cause, flood rate, flood volume, and event description

n  160 flood events identified in four major event types n  Pressure boundary failures (PBF) including HELB, piping and

component failures n  Maintenance induced floods not involving PBF n  Spurious fire protection system actuation n  Design deficiency

Distribution of Flood Events by Flood Type

Risk Evaluation Of LOCA Debris Blockage Of ECCS

(GSI-191)

PSA 2013 28

GSI-191 Modifies LOCA Success Criteria n  Swedish BWR event in 1980’s identified new failure mechanism

involving LOCA induced debris formation and flow blockage n  Normal PRA LOCA success criteria

n  based on capability to control inventory and remove heat via different systems

n  Independent of break location n  3 to 5 LOCA size categories (e.g. small, medium, large) from 0.5 in. n  break sizes greater than about 6in. have same criteria

n  GSI-191 PRA LOCA success criteria n  Debris formation highly dependent on break size, location, and

orientation within containment relative to insulation distribution n  Debris formation not significant for break sizes less than ≅10in. n  need to consider a continuous break size distribution up to and

including a DEGB at each location within RCS pressure boundary PSA 2013 29

Evolution of LOCA Frequency Estimates n  WASH-1400

n  Derived from gas-pipeline data and expert judgments to credit ASME nuclear piping codes

n  Used for all PRAs from 1975 to mid-1990s

n  NUREG/CR-5750 n  Incorporated some nuclear power plant service data n  Used for PRAs until NUREG-1829

n  NUREG-1829 (NUREG/CR-6928) n  Expert elicitation informed by service data and fracture mechanics n  Basis for most current PRAs

n  GSI-191 n  New requirements for more refined location and break size

dependent LOCA frequencies PSA 2013 30

LOCA Frequencies Objectives n  Incorporate insights from previous work on LOCA frequencies n  Characterize LOCA initiating events and their frequencies with

respect to: n  Specific components, materials, dimensions n  Specific locations n  Range of break sizes n  Damage / Degradation mechanisms and mitigation effectiveness

n  Quantify both aleatory and epistemic uncertainties; augment with sensitivity studies

n  Support interfaces with other parts of the GSI-191 evaluation n  Support submittal and RAI responses

31 PSA 2013

Risk Informed GSI-191

Pipe Rupture Model

PSA 2013 32

n  Total LOCA frequency of a given break size x is the sum of frequencies over a set of homogenous weld categories i, x treated as a continuous function, mi is the number of welds in the category, ρix is the rupture frequency per weld.

F(LOCAx ) = miρixi∑

Pipe Rupture Model

n  Rupture frequency for category i and size x is product of three terms: failure rate, conditional rupture size probability, and integrity management factor

PSA 2013 33

ρix = λikP Rx Fik( )k∑ Iik

Impact of Inspections on Weld Failure Rates

PSA 2013 34

Model Characteristics n  Failure defined as events involving repair or replacement n  All failures are assumed to be pre-cursors of pipe ruptures n  Failure rates based on Bayes’ analysis of pooled vendor

specific service data using accepted RI-ISI methodology n  Failure rates are conditional on the susceptibility to damage

mechanisms using RI-ISI damage mechanism criteria n  CRPs derived from expert elicitation inputs to NUREG-1829 n  Integrity management factors from Markov model from RI-

ISI program n  Epistemic uncertainties via Monte Carlo or via log normal

formulas for product terms

PSA 2013 35

Weld Categories

n  Service data indicates vast majority of pipe failures occur at or near welds

n  Weld locations provide convenient roadmap where pipe failures can occur

n  Weld categories define by: n  System (e.g. hot leg, cold leg, surge line..) n  Pipe Size n  Applicable susceptible damage mechanisms

determined via (EPRI) RI-ISI program

PSA 2013 36

Example Weld Categories

PSA 2013 37

System  Case   System   Component  

Case  Weld  Type   Applicable  DM  

STP  Total  No.  of  Welds  

Pipe  Size  (in.)  

DEGB  Size  (in.)  

1   RC  Hot  Leg  1A   B-­‐F   SC,  D&C   4   29   41.0  1B   B-­‐J   D&C   11   29   41.0  1C   B-­‐J   TF,  D&C   1   29   41.0  

2   RC  SG  Inlet   2   B-­‐F   SC,  D&C   4   29   41.0  

3   RC  Cold  Leg  

3A   B-­‐F  SC,  D&C  

4   27.5   38.9  3B   B-­‐J   4   31   43.8  3C   B-­‐J  

D&C   12   27.5   38.9  3D   B-­‐J   24   31   43.8  

4   RC  Surge  

4A   B-­‐F   SC,  TF,  D&C   1   16   22.6  4B   B-­‐J  

TF,  D&C  7   16   22.6  

4C   BC   2   16   22.6  4D   B-­‐J   6   2.5   3.5  

Failure Rate Development n  Pooled data for reactor vendor population

n  Westinghouse PWR data for STPEGs n  CE data for Calvert Cliffs

n  Separate failure rates for each of 40 to 50 weld categories n  Start with broad priors with means anchored to industry data

and RF =100 n  Bayes updates for different hypotheses about weld counts and

DM susceptibilities based on RI-ISI evaluations n  Mixture distributions to combine Bayes’ Updates n  Same methodology as developed for EPRI RI-ISI program

PSA 2013 38

Conditional Rupture Size Probabilities n  Reverse engineered from NUREG-1829 expert

elicitation inputs for reference systems (hot leg, cold leg, surge line, HPI line)

n  Inputs aggregated at system level using geometric mean method

n  Results for target LOCA frequency distributions converted to CRP distributions for reference systems

n  CRP distributions used as priors in Bayes’ update using industry data with zero ruptures and many failures.

PSA 2013 39

Example Failure Data Query

PSA 2013 40

Summary of Component Exposure Estimates W-PWRs

PSA 2013 41

System  Case   System   Component  

Case  Weld  Type  

Best  Estimate  

Upper  Bound  

Lower  Bound  

1   RCS  Hot  Leg  1A   B-­‐F   21,732   24,147   12,074  

1B,  1C   B-­‐J   32,297   36,221   24,147  2   RCS  SG  Inlet   2   B-­‐F   12,074   12,074   12,074  

3   RCS  Cold  Leg  3A   B-­‐F   22,315   24,794   12,397  3B   B-­‐J   123,764   177,279   99,177  

4   RCS  Surge  4A   B-­‐F   3,914   3,914   3,914  4B   B-­‐J   27,007   54,013   13,503  4C   BC   7,828   7,828   7,828  

5   PZR  5A–5D   B-­‐J   351,127   496,158   286,245  5E–5G   B-­‐F   19,083   19,083   19,083  

6   SB   6A–6B   B-­‐J   744,237   1,144,980   366,394  

7  SIR  Lines  Excl.  Accumulator   7A–7L   B-­‐J   590,797   637,190   507,518  

SIR  Accumulator  Lines   7M–7O   B-­‐J   175,067   277,693   132,810  

8   CVCS  8A–8D   B-­‐J   562,348   627,324   403,018  8E,  8F   BC   81,393   90,797   58,332  

Total  Estimated  Weld-­‐Yrs   2,774,983   3,633,494   1,958,513    

Mean Failure Rate Results for STP Class 1 Components

PSA 2013 42

Example Results – Hot Leg B-F Weld at RPV Nozzle

PSA 2013 43

Epistemic Uncertainties for Individual Welds –Case 1B

PSA 2013 44

New Paradigm for LWR PRA LOCA Frequencies

n  Historically LOCA frequencies only dependent on reactor type gross size range (e.g. small, medium, large) with minor differences on size ranges

n  Historically PRAs have relied on generic estimates from a single source of data

n  GSI-191 identifies the need to develop plant specific LOCA frequencies

PSA 2013 45

Plant to Plant Variability

n  Among Westinghouse PWRs there are 2 loop, 3 loop, and 4 loop designs

n  CE PWRs use different piping materials and different weld types; 1/3 larger hot legs

n  All plants have different numbers of welds and different weld categories incl. pipe sizes

n  Strong evidence to suggest plant to plant variability in LOCA frequencies

n  Lack of variability in PRAs due only to single source of generic data PSA 2013 46

Aggregated LOCA Frequencies are Plant Specific!

PSA 2013 47

PRA  LOCA  Category  

WCGS   STPEGS  Unit  1  

Break  Size  Range  (in.)  

Mean  Frequency  (/rx-­‐yr)  

Break  Size  Range  (in.)  

Mean  Frequency  (/rx-­‐yr)  

Small  LOCA  0.375  -­‐  .50  

1.71E-­‐04  Not  Modeled  

.50  -­‐  2.0   .5  -­‐  2.0   3.54E-­‐04  

Medium  LOCA   2.0  -­‐  6.0   9.38E-­‐06   2.0  -­‐  6.0   2.01E-­‐05  

Large  LOCA   >  6.0   5.77E-­‐07   >  6   2.29E-­‐06  

Comparison STP Pipe Induced Mean LOCA Frequencies with NUREG-1829

PSA 2013 48

Comparison of Calvert Cliffs, STPEGS, and NUREG-1829

PSA 2013 49

Contributions to LOCA Category 6 (> 31”) Frequency

PSA 2013 50

GSI-191 Results Summary n  Bottom-up LOCA frequencies for STP comparable to NUREG-1829 for pipe

induced LOCAs; n  Bottom-up LOCA frequencies for Category 6 for Calvert Cliffs appear to be

significantly smaller than for STP and NUREG-1829; no hot leg or SG inlet B-F welds in CE PWRs

n  Large variability in LOCA frequencies for different weld types; more than 3 orders of magnitude variation in mean failure rates

n  Uncertainties in local LOCA frequencies much larger than those for total LOCA frequencies; this result is magnified when debris-induced failures only occur in limited number of locations

n  After applying this method to STPEGS, Calvert Cliffs, and Vogtle there is strong evidence for plant to plant variability in total LOCA frequencies

PSA 2013 51

Summary

n  Pipe failure rate estimates involve large uncertainties; uncertainty modeling and quantification is a must

n  PIPExp database and OECD pipe data exchange provide excellent Type 1 sources

n  Uncertainties in estimating component exposure (“success data”) addressed

n  Experience and expertise required to interpret data, make sensible database queries, and apply failure rate models

PSA 2013 52