enclosure 2 - maximizing critical infrastructure

NRC – September 27, 2011

Maximizing Critical Infrastructure Performance with Probabilistic MethodsPart I

U.S. Nuclear Regulatory CommissionRockville, MDPresented By:

Ernest Lever and Daniel ErsoyInfrastructure Sector, GTI

© Gas Technology Institute 2011

2NRC – September 27, 2011 © Gas Technology Institute 2011

> Not-for-profit research, with 70 year history

> Facilities ─ 18 acre campus near Chicago─ 200,000 ft2,

28 specialized labs ─ Other sites in

Oklahoma and Alabama

> Staff of 250 > Market opportunities

are creating substantial growth

> 1,200 patents; 750 productsEnergy & Environmental Technology Center

Flex-Fuel Test

FacilityOffices & Labs

Gas Technology Institute (GTI)


Current Landscape in the U.S.


Normative Expert SystemsSound Science + Applied Models = Good Decisions

> Develop NORMATIVE expert systems that:1. Act rationally according to the laws of decision theory,

2. Use the formalism of Bayesian networks to efficiently represent and reason with uncertain knowledge, and

3. Do not attempt to mimic human thought processes.

> These systems will:‒ Accept human subject matter expert knowledge as their start point,

‒ Are capable of learning as live data is added, and

‒ Converge to an accurate representation of actual system behavior.

NRC – September 27, 2011 5© Gas Technology Institute 2011

Practical Infrastructure ConsiderationsPROBLEM

• Quantifying Risk Inherent in the Introduction of New Materials into Safety Related Nuclear Cooling Water Systems

• Aging System/Pipes

• Urban Congestion

• Zero Error Tolerance

• Information Overload

• Data Mining/Tracking

• QA/QC

• Knowledge Retention

• Safely Exploiting Abundant Worldwide Shale Gas Deposits

CONSTRAINTS ($)• Design

• Construction

• Operations

• Maintenance

• Restoration

• Labor

• Permits

• Materials

• Fines

• Liability

• Mandates

• Decommissioning

SOLUTION• Scientific Methods

• Advanced Modeling Approach (probability and multi-physics simulation)

• Sound Engineering Design

• Deployment of Normative Expert Systems

RESULTS• Explicitly Address the

Uncertainty and Unknowns

• Optimization of Policy: Balance Between Risk and Cost

• Improved Knowledge Management (i.e., Data Capture and Interpretation)


Research Strategy

> Research is needed to ensure that the current body of knowledge is leveraged and transferred into a useful set of tools.

> This additional research must focus on:‒ Infrastructure threats‒ Probabilistic models‒ Operations research methods to maximize utility and optimize

policy

> GTI uses advanced modeling and analysis tools to:‒ Enhance the level of infrastructure understanding, and‒ To allow operators to predict asset performance and calculate

system risk.


Coordinated Research and Development


Incorporating Results into Industry Guidelines and Standards

> The full potential of the proposed body of research can only be realized if it finds its way into industry guidelines, standards, and regulations.

> Presenting the results in draft standard form to the appropriate committee along with a supporting data set and GTI’s active technical support will facilitate this objective.


Operator Specific Deliverables Body of Knowledge and Toolboxes


Example - Typical Tasks Contained within anInfrastructure Integrity Management Plan


Example - Understanding Threat Interactions

Pipeline Failure

External Corrosion or Degradation Mechanisms

Internal Corrosion or Degradation Mechanisms

Crack Initiation and

Propagation Mechanisms

3rd Party Damage

Fabrication / Weld Quality / Fusion Quality

Wrinkle Bends / Miter Bend /

Fittings

Residual Stresses

Soil and Other Superimposed

Stresses

> Can individual threats each be at “acceptable” levels but when added together (superimposed) result in a significant threat to the pipeline or even a failure?

> What combinations of threats are most important to understand and control?

> How should threat interactions be calculated and dealt with?

> What process should an operator go through to identify unknown or hidden threats and how should they be worked into the pipeline integrity assessment?

> Can a process or methodology be employed to continuously monitor these threat interactions and flag concerns at trigger points?

Unknown and Hidden Threats

/ Operator Error


What Will It Take?

>Alignment with stakeholders to ensure critical needs are met – Regulators are stakeholders, not adversaries

>Research results targeted to agreed upon deliverables - Regulatory concerns must be addressed

>Dedicated team to build, organize, and maintain applications

>Funding and in-kind support


SERVICE WATER PIPING IN NUCLEAR POWER PLANTSNuclear Utility Needs1 U.S. NRC Concerns2

1. Volumetric examination methods for HDPE fused joints

A. Allowable service temperature up to 140°F

2. Standard bend tests for qualification of procedures and operators for making HDPE fused joints

B. Allowable stresses in Class 3 PE piping per CC N-755

3. Electrofusion coupling standardization and qualification

C. Any surface flaw up to 10% of wall thickness is acceptable regardless of diameter and wall thickness: Resistance to slow crack growth (SCG)

4. Standards for fittings

D. Limited set of essential variables for fusion procedure specification

E. Inadequate fusion procedure qualification requirements: Procedure, personnel and equipment

F. No volumetric inspection requirements for butt joints

G. Susceptibility of fusion butt joints to slow crack growth (SCG)

[1] NESCC 10-013 - HDPE Use by Catawba Nuclear Station, Steve Lefler, Duke Energy[2] NRC Issues Regarding the Use of HDPE Piping in Safety-Related Nuclear Applications, Use of HDPE for Power Plant Piping Systems Workshop, June 7-10, 2010, Charlotte, North Carolina. Eric Focht, U.S. NRC


System Risk - Relevant Material, Process and Service Environment Interactions


Cooperative Research to Address Risk Related to HDPE in Service Water Applications – (1)

ResearchArea

ResearchOrganization

Microstructure Characterization NIST

Mechanical Properties and Viscoelasticity NIST, GTI

Fracture NIST, GTI

Fusion Optimization GTI, NIST

Process Optimization GTI


Cooperative Research to Address Risk Related to HDPE in Service Water Applications – (2)


Thank You !

Questions ?

NRC – September 27, 2011

Maximizing Critical

Infrastructure Performance

with Probabilistic Methods

Part II

U.S. Nuclear Regulatory CommissionRockville, MDPresented By:

Ernest Lever and Daniel Ersoy

Infrastructure Sector, GTI

© Gas Technology Institute 2011


Virgin Resin and Pipe Products

> Best possible scenario – new, properly processed

bulk pipe:

‒ Master Curves apply.

> What about?

‒ Butt fusions,

‒ Fittings,

‒ Temperature variations,

‒ Environment variations (e.g., fluid chemistry),

‒ Superimposed stresses,

‒ Defects (e.g., cracks).


Shifting the Master Curve to Account for Variances

> Start with resin

‒ 1st processing of resin during pipe manufacture,

‒ 2nd processing of resin during butt fusion.

> Results in change of material properties

‒ Has been identified in NRC confirmatory research (PENT test of butt fusion joints),

> In order to predict initiation and growth parameters for “processed” material

- one must shift the master curve by:

‒ Processing under carefully controlled conditions,

‒ Conducting a full Rate Process Method (RPM) analysis, and

‒ Correlating the shift in the crack initiation curve to process conditions.

> This will enable the definition of a set of critical variables and their bounds for each processing step that ensures that slow crack growth properties of installed systems meets design requirements.


Master Curve Shifting

Slow Crack Growth in HDPE fittings

Electrofusion Coupler 80°C Performance Comparison

100

600

1100

1600

2100

2600

1 10 100 1000 10000 100000 1000000

Time [h]

Str

ess [

psi]

PE100 "A" ISO 9080

Stress Rupture Curve

Supplied by Resin

Manufacturer

No Failure

PE3408 Stress Rupture Curves Calculated from ISO 9080 Coefficients Supplied by Resin Manufacturer

PE100 "B" Coupler SCG Failure Points

PE100 "A" Coupler SCG Failure Points

PE3408 Coupler SCG Failure Points

The PE100 couplers were produced at three different manufacturing conditions. The spread in the results is due to the influence of

different thermal histories on SCG performance of electrofusion couplers. The data points are not grouped by thermal history in this

plot that is intended to investigate general structure of the data.

Onset of SCG

Regression Line Through

3408Coupler SCG Failure Points


Detailed Slow Crack Growth Study


Study Triaxial, Bi-Axial, and Uni-Axial Stress

States (Possible in Multiple Fluid Media)


Adding Cracks/Flaws Into the Equation –

Outline of Method (in following slides)

1. Use fundamental crack equations

2. Introduce uncertainty found in input parameters

3. Construct a designed experiment with low probability of error

4. Use an analysis technique with high statistical power

5. Track all uncertainties and errors through the analysis to calculate solution’s standard error.

6. Arrive at solution with confidence intervals.


Example of Design of Experiment and Synthetic Probability Analysis

Applied to Leak Rupture Boundary of Steel Gas Transmission Lines -

Two Factor Regression LCL Surfaces with Conclusive Rupture Data Set

> 95% two tailed CI LCL surfaces (97.5% probability above LCL) rupture failure stress as a function of yield stress [30 to 80 ksi] and CVN [15 to 60 ft-lb] for C = 1 to 7 [top to bottom surfaces] with 638 conclusive incident and full size test rupture points overlaid.


Predictive Capability of Regression Model

> Predictive Capability of the four parameter full quadratic (15 coefficient) regression model solution in Fundamental Instability Stress Model for C = 6.


Standard Error and

Solution Surface Example


Perturbation Plot Example


Desirability Plot Example

> C = 6, Pipe Stress = 30%SMYS, Yield Strength = 40ksi, CVN = 20ft-lb. Red

is failure by leak, blue is failure by rupture.


Probabilistic framework for estimating likelihood of failure due

to flaw (crack) tip stress – “a closed form predictive equation”.

> Using the Synthetic Design of Experiments (DoE) for HDPE Pipe Flaws:

1. Fundamental crack equations – FFS-1/API 579-12. Take these and introduce a complete set of uncertainty (from input parameter

variance/variability) via probability distributions.3. Explicitly include correction factors for temperature and crack tip radius in the

DoE4. Construct a robust DoE around power plant design/operational space; set up

to ensure a low Probability of Error (PoE) in the design space.5. Use synthetic Monte Carlo analysis with adequate replicates and spacing to

provide a high statistical power.6. Arrive at a 97.5% upper confidence level for crack tip stress as a function of

temperature, operating pressure, pipe and flaw dimensions.


Constructing Probability Distributions from

Actual Field Defects to Model Crack Tip Stress

> Conduct survey to obtain data and develop a comprehensive set of distributions of actual flaw geometries and sizes.

> Take these distributions and run Finite Element Analysis (FEA) to estimate crack tip stress levels found in practice.


Overlay Theoretical and Empirical Results and

Relating Results to

> Overlay calculated field crack tip stresses on the closed form predictive equation response surfaces (ucl most important).

> This provides a quantitative measure of the conservatism inherent in the closed form, pure fracture mechanics approach which is based on a “mathematically perfect” flaw. Actual defects do not have an atomically

sharp crack tip radius.

> Finally, relate the calculated crack tip stress (i.e., the conservative ucl) back to the shifted master curves for predictive likelihood of failure at operating temperatures and pressures.


Relate Crack Tip Stresses Back to Shifted Master Curves

> Finally, relate the calculated crack tip stress (i.e., the conservative ucl) back to the shifted master curves for predictive likelihood of failure at operating temperatures and pressures.

Electrofusion Coupler 80°C Performance Comparison

100

600

1100

1600

2100

2600

1 10 100 1000 10000 100000 1000000

Time [h]

Str

es

s [

ps

i]

PE100 "A" ISO 9080

Stress Rupture Curve

Supplied by Resin

Manufacturer

No Failure

PE3408 Stress Rupture Curves Calculated from ISO 9080 Coefficients Supplied by Resin Manufacturer

PE100 "B" Coupler SCG Failure Points

PE100 "A" Coupler SCG Failure Points

PE3408 Coupler SCG Failure Points

The PE100 couplers were produced at three different manufacturing conditions. The spread in the results is due to the influence of

different thermal histories on SCG performance of electrofusion couplers. The data points are not grouped by thermal history in this

plot that is intended to investigate general structure of the data.

Onset of SCG

Regression Line Through

3408Coupler SCG Failure Points


Identifying Essential Variables and Their

Bounds for Butt Fusion

> Currently available data logging systems for butt-fusion machines reflect an operating window that captures common practice and, on average, ensures a reliable butt-fusion joint.

> Inspection protocols rely mainly on visual cues such as bead size and shape that reflect a large body of experience with historic materials.

> It is well known that it is possible to produce sub-standard joints while following the existing procedure and complying with all visual inspection criteria.

> These uncommon circumstances arise when the polymer state at the interface does not meet the minimum requirements to ensure successful co-crystallizaton across the interface.

> Continually improving the safety of fused systems obligates us to reduce the likelihood of rogue joints that will constitute a “weakest link”

to a quantifiable minimum.


Identifying Essential Variables and Their

Bounds for Butt Fusion - Continued

Design Summary

Study Type Response Surface Runs 187Design Type Historical Data Blocks No BlocksDesign Model Quadratic

Factor Name Units Type Subtype Minimum MaximumA OD in Numeric Continuous 24.00 48.00B DR Numeric Continuous 11.00 31.00C Plate Temp F Numeric Continuous 400.00 500.00D Soak Time s Numeric Continuous 240.00 960.00E Ambient Temp F Numeric Continuous 16.00 120.00F Wind Speed mph Numeric Continuous 1.00 13.00G Squeeze Temp C Numeric Continuous 155.00 185.00

> Heater Plate Flux [W/in^2]

> Heater Plate Temperatures [°F]

> Typical heat soak times [s]

> Dwell (Open) time

> Other parameters that need to be taken into account to correctly simulate the process


Statistical Analysis (ANOVA) to

Identify Significant Interactions

> Average Interface Cool Time to 80°C:

‒ 6 First Order Terms significant 15 Second Order Terms significant

‒ In this case A, B, C, D, E, F, AD, AE, AG, BD, BE, CD, CE, CG, DE, DF, EF, C2, E2, F2, G2 are significant model terms.

> Melt Depth After Squeeze:

‒ 5 First Order Terms significant 8 Second Order Terms significant

‒ In this case B, C, D, E, G, AD, BD, CD, CE, CG, DE, DG, EG are significant model terms.

Design-Expert® Software

Time to 80 C

Actual FactorsA: OD = 48.00B: DR = 11.00C: Plate Temp = 450.0D: Soak Time = 960.0E: Ambient Temp = 68.00F: Wind Speed = 1.000G: Squeeze Temp = 155.0

Perturbation

Deviation from Reference Point (Coded Units)

Tim

e t

o 8

0 C

-2.000 -1.000 0.0000 1.000 2.000

0.0000

2750.

5500.

8250.

1.100E+004

AA B BC

C

D

D

E

E

F

F

G G


Butt Fusion Temperature Profiles

12 IPS SDR 9

Datalog vs FEA

25

75

125

175

225

0 100 200 300 400 500 600 700 800 900

Seconds

Deg

rees C

0 dl

1/16 dl

1/8 dl

1/4 dl

1/2 dl

1 dl

0

1/16

1/8

1/4

1/2

1

450F

266F

158F

A look at thermocouple probe temperatures vs. FEA probe temperatures, over time.


Rate Process Method (RPM) Analysis on

Fusion Interfaces

> It is essential to subject butt-fusion interfaces fused under carefully controlled and documented fusion conditions to a full RPM analysis in order to:

‒ Establish a Lower Confidence Limit on the predicted lifetime of the joint as a function of operating stress and temperature,

‒ Develop a valid model for SCG on a fused interface to resolve the current over emphasis on baseline PENT performance of the raw material that does not reflect the impact of multiple processing histories that exist in a butt-fused joint, and

‒ Provide the basis for a probabilistic causal model that can be used to establish system capability under widely varying operating conditions.


In Summary - Normative Expert SystemsSound Science + Applied Models = Good Decisions

> Develop NORMATIVE expert systems

that:1. Act rationally according to the laws of decision theory, 2. Use the formalism of Bayesian networks to efficiently

represent and reason with uncertain knowledge, and3. Do not attempt to mimic human thought processes.

> These systems will:‒ Accept human subject matter expert knowledge as their

start point,‒ Are capable of learning as live data is added, and‒ Converge to an accurate representation of actual system

behavior.


Thank You !

Questions ?

enclosure 2 - maximizing critical infrastructure

Documents