Quantitative Risk Assessment of Underground Pipelines with a … · 2019-08-09 · Quantitative...

www.cfertech.com

Quantitative Risk Assessment of Underground Pipelines with a Focus on Probability Estimation

4th International Forum on

Industrial Pipeline Management

Mark Stephens C-FER Technologies

Kaohsiung, Taiwan - August 9, 2019

1

www.cfertech.com

About C-FER

QRA of Underground Pipelines, Aug 9, 2019,

Kaohsiung, Taiwan 2

• Independent engineering, consulting and testing services

• Not-for-Profit subsidiary of Alberta Innovates

• Self-sustaining, fee-for-service operation

• In operation since 1984

www.cfertech.com

Making Pipelines Safer

• Large-scale Destructive Testing

• Technology Evaluation

• Advanced Design Methods

• Quantitative Risk Assessment

• Risk and Reliability Based Integrity Management


Kaohsiung, Taiwan 3

www.cfertech.com

Why Consider Risk?

• Can provide an objective and consistent basis for assessing, managing and demonstrating pipeline performance

• Actions based on risk considerations reflect both likelihood of failure and potential consequences

• Applicable to new design and integrity management of existing pipelines Presentation focus on application to integrity management


Kaohsiung, Taiwan 4

www.cfertech.com

Why Quantitative Risk?

• Available approaches

Qualitative methods

• Characterize risk by a descriptive value or index

• Best suited to threat identification and risk ranking

Quantitative methods

• Characterize risk as the expected value of a measurable consequence parameter

• More objective basis for decision making

Better suited to determining if a system is safe enough

Better suited to determining what actions are required and when


Kaohsiung, Taiwan 5

www.cfertech.com

North American Perspective

High profile incidents have led stakeholders to demand better pipeline performance

Quantitative analysis methods

are being embraced to more

objectively, consistently and

defensibly address pipeline risk


Kaohsiung, Taiwan 6

www.cfertech.com

Quantitative Risk Analysis (QRA)

• Given the basic definition of risk

• QRA requires

Quantitative failure probability or failure frequency estimates

• Distinction by failure cause to identify dominant threats and targeted remediation action

• Distinction by failure mode (leak versus rupture) to acknowledge consequence differences

Quantitative failure consequence estimates

• Consideration of public safety impact for toxic or flammable gas and liquid lines

• Consideration of environmental impact for lines transporting persistent hazardous liquids

• Consideration of financial impact direct and indirect costs


Kaohsiung, Taiwan 7

Risk = Probability x Consequence

www.cfertech.com

Probability Estimation

• Quantitative analysis options

Statistical methods

• Estimates developed from evaluation of past pipeline performance

• Various historical incident data sets available

Reliability-based methods

• Estimates developed from pipeline and ROW attributes

• Various approaches available including structural reliability methods


Kaohsiung, Taiwan 8

www.cfertech.com

Statistical Methods

• Approach Collect historical data on previous pipeline failures

Use historical data as basis for probability estimates

• Data sources Proprietary operator data

Industry data in public domain

• US: US Department of Transportation (USDOT / PHMSA)

• Canada: National Energy Board (NEB) | Canadian Energy Pipeline Association (CEPA)

• Europe: European Gas Pipeline Incident Group (EGIG) UK Onshore Pipeline Operators Association (UKOPA) CONservation of Clean Air and Water in Europe (CONCAWE)


Kaohsiung, Taiwan 9

www.cfertech.com

Example – Statistical Approach

Calculate annual probability of corrosion rupture for section of natural gas pipeline

• Given incident database Data on 100,000 km of gas transmission lines

50 corrosion failures recorded in last 5 years

10% of corrosion failures were ruptures

• Solution Annual failure rate = number of incidents / system exposure = 50 / (100,000 x 5) = 1 x 10-4 per km-yr

Annual rupture rate = 0.1 x 1 x 10-4 per km-yr = 1 x 10-5 per km-yr

• Key assumption Historical average is representative of pipeline performance in the future

• Possible concerns How are failure rate estimates affected by:

• pipeline age and condition

• operating stress level

• integrity management actions taken


Kaohsiung, Taiwan 10

www.cfertech.com

Comments on Statistical Methods

• Advantages Simple

Credible (based on real data)

• Limitations Generally not very pipeline-specific

• Data sets do not support subdivision by key pipeline attributes

• No link between failure rate and maintenance actions

Ignores systematic changes in pipeline condition

• Cannot account for time-dependent deterioration



www.cfertech.com

Reliability-based Methods

• Approach

Develop failure prediction models that define the sets of conditions that can lead to failure

Use structural reliability methods where appropriate to combine deterministic models with input uncertainties to estimate probability (or frequency) of failure for individual threats


Kaohsiung, Taiwan

Load or Resistance Mean

Resistance

Mean Load

Probability Distribution of the Resistance (R)

Probability Distribution of the Load (L)

Small region of overlap proportional to probability of failure

(POF)

POF = P(R < L)

Requires formal characterization of the uncertainties in the applied load and the available resistance for each damage or deterioration mechanism

12

www.cfertech.com

Uncertainties Inherent in the Probability Estimation Process

• Random variations Loads imposed on the line Internal pressure

Third party impact force

• Measurement uncertainty Pipe properties and line condition Joint-by-joint yield strength & fracture toughness

Number and size of defects

Defect growth rates

• Model uncertainty Pipe behavior under loads Burst capacity at corrosion defect

Dent-gouge failure susceptibility

Tensile capacity of girth welds



www.cfertech.com

Reliability Model Chosen Depends on the Type of Threat

• Consider management of progressive (time dependent) damage Determine existing damage severity

• Find and size existing defects

Assess anticipated behavior over time

• Estimate rate of defect growth and assess time for defects to become failure critical

Manage integrity through periodic inspection and remediation or proof-testing

• Examples include corrosion and cracking Consider metal loss corrosion



www.cfertech.com


• Corrosion – example of a time dependent damage mechanism

Failure rate (per km-yr) = No. defects (per km) x Probability of failure per defect (per yr)



• Considerations in developing failure rate estimate

– Identifying and characterizing defect population • Assumed number of features and feature sizes should reflect the

probability of detection and sizing accuracy of inspection method

– Estimating probability of failure over time structural reliability model • Failure predictions should reflect uncertainties in: defect sizes and growth rates,

variability in pipe properties, and accuracy of the failure prediction model

– Considering impact of maintenance • Effects of defect remediation, re-inspection interval and modified

operating conditions should be reflected in probability estimates

www.cfertech.com

Probability Estimation Process

• Select deterministic failure prediction models (consider leak and burst separately)

• Characterize parameter uncertainties using probability distributions

• Calculate failure probability using standard techniques (e.g. simulation)


Kaohsiung, Taiwan

Inspection

data

Material property

& dimension data

Operating

pressure

Inspection-related

uncertainties

Inspection

and/or operating

history data

Failure probability

as function of time

Model Results

Test

res

ult

s

x x x

x x

x x

x

x x

x

Model

uncertainties

Feature Dimension

Defect

characteristics

Pro

bab

ility

Den

sity

Growth Rate

Defect growth

parameters

Yield stress & Toughness

Freq

uen

cy

Pipe

properties

Pressure

Load

characteristics

Pro

bab

ility

Den

sity

Failure model

& test data

Growth model

uncertainties

Freq

uen

cy

Example for corrosion

16

www.cfertech.com

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

0 5 10 15 20 25

Time (years)

Pro

bab

ilit

y o

f F

ailu

re

(per

km

yr)

Probability Estimation Results

• Obtain pipeline failure rate versus time


Kaohsiung, Taiwan

Leak

Burst

Allowable

POF leak*

Allowable

POF burst*

Not OK

Repair or re-inspection at or before

8

*based on risk considerations that reflect failure consequences

17

www.cfertech.com

Impact of Maintenance

• Mitigation philosophy Find and eliminate significant defects before they reach critical size

• Maintenance options, e.g. In-line Inspection (ILI) and selective repair

Hydrostatic testing

• Maintenance impact Eliminate contribution to failure probability from defects removed



www.cfertech.com

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

0 5 10 15 20 25

Time (years)

Pro

bab

ilit

y o

f F

ailu

re

(per

km

yr)

Impact of Maintenance

• Obtain pipeline failure rate versus time – accounting for maintenance


Kaohsiung, Taiwan

Leak

Burst

Allowable

POF leak*

Allowable

POF burst*

Time to re-inspection

12

Repair #2

Repair #1

*based on risk considerations that reflect consequences

19

www.cfertech.com

Reliability Model Depends on the Type of Threat

• Consider management of random (time independent) damage Determine likelihood of damaging event occurrence

• Characterize event frequency

Assess anticipated pipeline response to event

• Characterize the damage tolerance of pipeline

Manage integrity through control of event likelihood or potential for failure given an event

• Examples include mechanical damage and sudden ground movement Consider mechanical damage



www.cfertech.com


• Mechanical damage – example of time independent damage

Failure rate (per mi-yr) = Hit frequency (per km-yr) x Probability of failure given hit



• Considerations in developing failure rate estimate

– Estimating hit frequency • Likelihood of having excavation activity and effectiveness of damage prevention

measures should be reflected in hit frequency estimate

– Estimating failure probability given a hit • Failure probability given line hit should reflect: uncertainty in damage caused by

contact, variability in pipe properties, and accuracy of failure prediction model

– Considering impact of maintenance • Effect of changes in damage prevention measures or modified operating conditions

should be reflected in probability estimates

www.cfertech.com

Assessment Approach

• Calculate equipment impact failure probability

Slide 22

Pipelinehit by

excavator

Excavationto pipelinelocation

Operator ROWpatrols fail todetect activity

Contractorfails to avoid

pipeline

Constructionactivity

Inadequatecover

Patrolfails todetectactivity

Intervalbetweenpatrolstoo long

Contractorunaware of

pipelineContractor

ignoreswarnings

Inadequatemarkings

Inadequateone-callsystem

Yield stress (MPa)

Frequency

Data on impact

force and

dent-gouge

geometry

Model results

Test results

xxx

x

xx

x

xx

x

x

Data on steel

properties and

pipe dimensions

Failure models

and

test results

Failure probability

given hit

model

uncertainties

pipe

properties

Load (kN)

Frequency

outside

force

Hit Frequency (inductive logic model – fault tree)

Failure given Hit (structural reliability models for puncture or dent-gouge)

Inadequate

cover or

protection

UofA CME 694 – Guest Lecture – Stephens 2018 QRA of Underground Pipelines, Aug 9, 2019,

Kaohsiung, Taiwan

www.cfertech.com

Hit Frequency


Kaohsiung, Taiwan

P i p e l i n e h i t b y

e x c a v a t o r

E x c a v a t i o n t o p i p e l i n e l o c a t i o n

O p e r a t o r R O W p a t r o l s f a i l t o d e t e c t a c t i v i t y

C o n t r a c t o r f a i l s t o a v o i d

p i p e l i n e

C o n s t r u c t i o n a c t i v i t y

P a t r o l f a i l s t o d e t e c t a c t i v i t y

I n t e r v a l b e t w e e n p a t r o l s t o o l o n g

C o n t r a c t o r u n a w a r e o f

p i p e l i n e C o n t r a c t o r

i g n o r e s w a r n i n g s

I n a d e q u a t e m a r k i n g s

I n a d e q u a t e o n e - c a l l s y s t e m

Basic event probabilities (depend on line attributes)

Calibrated using historical data and engineering models

Simplified fault tree model - to illustrate concept

Inadequate

cover or

protection

23

www.cfertech.com

Hit Frequency Estimation


Kaohsiung, Taiwan

- land use and presence of crossings - depth of burial - one call system type - dig notification requirement - dig notification response - public awareness level - right-of-way indication - alignment markers - explicit signage - alignment markers - above ground - alignment markers - buried - surveillance method / interval - mechanical protection

Excavation

depth

exceeds

cover depth

Failure of

mechanical

protection

Failure of

protective

measures

Buried

markers fail

to convey

location

Accidental

activity on

located

alignment

Parties

fail to call

before

moving onto

ROW

Operator

fails to

ensure correct

location of

alignment

Pipeline hit by third-

party mechanical

equipment

Failure of

preventive

measures

Pipeline operator

not notified of

pending activity

No patrol

during

period of

activity

Patrol

personnel

fail to detect

activity

One-call

system fails

to notify

relevant

operator

One-call system

fails to notify

pipeline operator

Activity on

pipeline

alignment

ROW

indicators

not

recognized

Parties

ignore ROW

indicators

Permenant

above-

ground

markers fail

to convey

location

Explicit

signage not

seen

Parties

ignore explicit

signage

Temporary measures

fail to correctly

locate alignment

Alignment not

correctly located

ROW patrols

fail to detect

activity

Parties fail to use

one-call system

Parties fail to use

one-call system

when on ROW

Parties fail to notify

operator directly

E1

B1

E2

E4

E5

E6

E9 E7

E10

E11

B3 B4

B2

B5 B6B8 B9

B10

B11B12

B13 B14

Model reflects effect on hit frequency of a range of system attributes and damage prevention measures, including:

Actual fault tree model required - more complex

24

www.cfertech.com

Effect of Maintenance

• Mitigation philosophy Reduce potential for line hits

• Maintenance option examples ROW condition enhancement

Increased signage/markers

Public awareness improvements

Increase burial depth

Introduce mechanical protection

• Maintenance impact Reduce hit frequency proportionate reduction in failure probability



www.cfertech.com

Effect of Damage Management


Kaohsiung, Taiwan

Excavation

depth

exceeds

cover depth

Failure of

mechanical

protection

Failure of

protective

measures

Buried

markers fail

to convey

location

Accidental

activity on

located

alignment

Parties

fail to call

before

moving onto

ROW

Operator

fails to

ensure correct

location of

alignment

Pipeline hit by third-

party mechanical

equipment

Failure of

preventive

measures

Pipeline operator

not notified of

pending activity

No patrol

during

period of

activity

Patrol

personnel

fail to detect

activity

One-call

system fails

to notify

relevant

operator

One-call system

fails to notify

pipeline operator

Activity on

pipeline

alignment

ROW

indicators

not

recognized

Parties

ignore ROW

indicators

Permenant

above-

ground

markers fail

to convey

location

Explicit

signage not

seen

Parties

ignore explicit

signage

Temporary measures

fail to correctly

locate alignment

Alignment not

correctly located

ROW patrols

fail to detect

activity

Parties fail to use

one-call system

Parties fail to use

one-call system

when on ROW

Parties fail to notify

operator directly

E1

B1

E2

E4

E5

E6

E9 E7

E10

E11

B3 B4

B2

B5 B6B8 B9

B10

B11B12

B13 B14

Example application

26

- land use and presence of crossings - depth of burial - one call system type - dig notification requirement - dig notification response - public awareness level - right-of-way indication - alignment markers - explicit signage - alignment markers - above ground - alignment markers - buried - surveillance method / interval - mechanical protection

Model reflects effect on hit frequency of a range of system attributes and damage prevention measures, including:

Improve awareness Enhance indication

www.cfertech.com


• Attribute changes affect basic event probabilities change hit frequency change failure rate


Kaohsiung, Taiwan

Probability that ROW indicators are not recognized

Public Awareness Level Right - of - way Indication Below average Average Above average

None

1.0 1.0 1.0

Intermittent and/or

very limited indication

0.87 0.65 0.43

Continuous but

limited indication

0.47 0.35 0.23

Continuous and highly indicative

0.21 0.16 0.11

27

www.cfertech.com


• Example: from C-FER fault tree model



1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00

Probability of Line Hit (per mi-yr)

Painted Slab

Plain Slab

Awareness & ROW Improvement

Increase Public Awareness

ROW Condition Improvement

Base Case

Probability of Line Hit (per km-yr) ROW=right of way

www.cfertech.com

Comments on Consequence Estimation

•Considerations Safety, Environment and Cost



Population effects dominated by short term impact of flammable or toxic gas or liquid product releases

Product Release

Rate

Pool fire

Vapour cloud explosion

Vapour cloud fire

Toxic or asphyxiating vapour cloud

Fireball Jet fire

Chance of casualty

Number of casualties

Measures of Safety Impact

A measure of Environmental and

Socioeconomic Impact

Environmental impact dominated by longer term impact of persistent liquid product releases

Volume Released

Environment Sensitivity

or Importance

Product Toxicity

Effective Release Volume

www.cfertech.com

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

1.E-01

0 5 10 15 20 25

Time (years)

Pro

bab

ilit

y o

f F

ailu

re

(per

km

yr)

Risk Management

• Obtain pipeline risk versus time Determine required action


Kaohsiung, Taiwan

Risk Threshold

Option 2 – corrosion mitigation

Option 1 – mechanical damage prevention

Base Case S

eg

me

nt

Ris

k (

pe

r k

m-y

r)

30

Risk Thresholds Safety: - Individual risk of fatality - F-N Curve (expected no. fatalities) Environment: - Expected release volume*

*volume adjusted for spill impact

Criteria currently under review for possible inclusion in CSA Z662 - 2023

calculate over sliding evaluation length

www.cfertech.com

Risk Evaluation of Linear Systems

• Key considerations

Some integrity threats are concentrated at explicit locations

• Locations are know (e.g. corrosion defects found during inspection)

• Best evaluated as discrete, location-specific probability units: per year

Some integrity threats are distributed along pipeline length

• Locations not known (e.g. future mechanical damage, corrosion defects not found)

• Best evaluated as failure rate or distributed probability units: per km-yr



Pipeline

Location specific probability pfi

Distributed probability pd

Evaluation Length, Le

What is required is the aggregated probability or average failure rate over a relevant length

n

i

fi

e

df pL

pP1

1

www.cfertech.com

Evaluation Length Considerations

• Example: safety implications of natural gas pipeline



Interaction Length is segment length with potential to affect dwelling occupants - level of safety risk exposure depends on aggregated failure probability from all active threats within IL

Rupture hazard

zone

Length Swept Out by Hazard Circle

Interaction Length, IL

Dwelling unit

Pipeline

www.cfertech.com


• Example: environmental implications of liquid product pipeline



Interaction Length is segment length with potential to impact river - level of environmental risk exposure depends on aggregated failure probability from all active threats within IL

Length that Can Draining into River

Interaction Length, IL

Rupture spill

path

Pipeline

www.cfertech.com


• Pipeline risk should be evaluated over a sliding evaluation length

• In many situations the appropriate evaluation length = hazard interaction length

• The appropriate evaluation length depends on The risk measure being evaluated

Various pipeline and right-of-way attributes



www.cfertech.com

Quantitative Risk Assessment - Summary

• Benefits of Quantitative Risk Assessment (QRA)

An objective and consistent basis for assessing pipeline operating risk

• Added benefits of QRA employing structural reliability methods

Better able to reflect line-specific factors including impact of maintenance actions

Framework for consideration of all significant sources of uncertainty

Sound basis for decision making

• What actions are required

• When action is required



www.cfertech.com

Comments on Quantitative Risk Assessment Using Structural Reliability Methods

• Feasibility Structural reliability methods and models for specific pipeline integrity threats have been under development for more than 20 years

• Various Joint Industry Research Programs

• PRCI Research C-FER’s Reliability Based Design and Assessment (RBDA) process

• Reliability models directly applicable in risk assessment

Many models in public domain, e.g. Annex O of CSA Z662

• Validity Model development activities have included extensive calibration and validation exercises

• Models have been used to hindcast historical failure rates for the existing North American pipeline network – agreement shown to be good



Quantitative Risk Assessment of Underground Pipelines with a … · 2019-08-09 · Quantitative...

Documents

Transcript of Quantitative Risk Assessment of Underground Pipelines with a … · 2019-08-09 · Quantitative...