You are on page 1of 118

Stress-Based Fatigue Monitoring

Methodology for Fatigue Monitoring of Class 1


Nuclear Components in a Reactor Water Environment

2011 TECHNICAL REPORT

13307342
13307342
Stress-Based Fatigue
Monitoring
Methodology for Fatigue Monitoring of Class
1 Nuclear Components in a Reactor Water
Environment

EPRI Project Manager


S. Chu

3420 Hillview Avenue


Palo Alto, CA 94304-1338
USA

PO Box 10412
Palo Alto, CA 94303-0813
USA

800.313.3774
650.855.2121

askepri@epri.com 1022876
www.epri.com Final Report, November 2011
13307342
DISCLAIMER OF WARRANTIES AND LIMITATION OF LIABILITIES

THIS DOCUMENT WAS PREPARED BY THE ORGANIZATION(S) NAMED BELOW AS AN ACCOUNT OF


WORK SPONSORED OR COSPONSORED BY THE ELECTRIC POWER RESEARCH INSTITUTE, INC. (EPRI).
NEITHER EPRI, ANY MEMBER OF EPRI, ANY COSPONSOR, THE ORGANIZATION(S) BELOW, NOR ANY
PERSON ACTING ON BEHALF OF ANY OF THEM:

(A) MAKES ANY WARRANTY OR REPRESENTATION WHATSOEVER, EXPRESS OR IMPLIED, (I) WITH
RESPECT TO THE USE OF ANY INFORMATION, APPARATUS, METHOD, PROCESS, OR SIMILAR ITEM
DISCLOSED IN THIS DOCUMENT, INCLUDING MERCHANTABILITY AND FITNESS FOR A PARTICULAR
PURPOSE, OR (II) THAT SUCH USE DOES NOT INFRINGE ON OR INTERFERE WITH PRIVATELY OWNED
RIGHTS, INCLUDING ANY PARTY'S INTELLECTUAL PROPERTY, OR (III) THAT THIS DOCUMENT IS SUITABLE
TO ANY PARTICULAR USER'S CIRCUMSTANCE; OR

(B) ASSUMES RESPONSIBILITY FOR ANY DAMAGES OR OTHER LIABILITY WHATSOEVER (INCLUDING ANY
CONSEQUENTIAL DAMAGES, EVEN IF EPRI OR ANY EPRI REPRESENTATIVE HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES) RESULTING FROM YOUR SELECTION OR USE OF THIS DOCUMENT OR
ANY INFORMATION, APPARATUS, METHOD, PROCESS, OR SIMILAR ITEM DISCLOSED IN THIS
DOCUMENT.

REFERENCE HEREIN TO ANY SPECIFIC COMMERCIAL PRODUCT, PROCESS, OR SERVICE BY ITS TRADE
NAME, TRADEMARK, MANUFACTURER, OR OTHERWISE, DOES NOT NECESSARILY CONSTITUTE OR
IMPLY ITS ENDORSEMENT, RECOMMENDATION, OR FAVORING BY EPRI.

THE FOLLOWING ORGANIZATION, UNDER CONTRACT TO EPRI, PREPARED THIS REPORT:

Structural Integrity Associates, Inc.

NOTE

For further information about EPRI, call the EPRI Customer Assistance Center at 800.313.3774 or
e-mail askepri@epri.com.

Electric Power Research Institute, EPRI, and TOGETHER…SHAPING THE FUTURE OF ELECTRICITY are
registered service marks of the Electric Power Research Institute, Inc.

Copyright © 2011 Electric Power Research Institute, Inc. All rights reserved.

13307342
Acknowledgments
The following organization, under contract to the Electric Power
Research Institute (EPRI), prepared this report:

Structural Integrity Associates, Inc.


5215 Hellyer Avenue, Suite 210
San Jose, CA 95138

Principal Investigators
C. Carney
T. Gilman

This report describes research sponsored by EPRI. The work for this
report was completed with the support and input of industry
reviewers.

E. Berak Exelon
C. Bosch Iberdrola
K. Brune TVA
T. Childress Duke Energy
F. Corchón Nuclenor
J. M. Davis (retired)
M. Dingler WCNOC
C. Faidy Électricité de France S.A.
D. Galbally Iberdrola
C. Oh KEPCO E&C
J. Park KEPCO E&C
J. Rudolph, B. Heinz,
S. Bergholz, D. Wu AREVA NP GmbH
J. Weik PPL Corporation
T. Woodson Entergy

This publication is a corporate


document that should be cited in the
literature in the following manner:

Stress-Based Fatigue Monitoring:


Methodology for Fatigue Monitoring
of Class 1 Nuclear Components in a
Reactor Water Environment.
EPRI, Palo Alto, CA: 2011.
1022876.
g iii h

13307342
13307342
Abstract

The FatiguePro software, developed by the Electric Power Research


Institute (EPRI) and first deployed in 1989, is a fatigue monitoring
program that is widely used around the world to assist with aging
management of nuclear power plants.

The FatiguePro stress-based fatigue (SBF) module has used a single


stress term for calculating fatigue usage factors. This simplified
approach was chosen not only because of computer limitations at the
time, but also because the conventional stress cycle counting
algorithms that are documented in the open literature and that
appropriately handle random, ordered stress histories, as opposed to
idealized transient definitions, are generally limited to a single stress
term. The United States Nuclear Regulatory Commission (NRC), in
Regulatory Issue Summary RIS 2008-30, expressed concerns with
this simplified approach.

EPRI will continue to maintain availability of FatiguePro; however,


it is expected that vendor organizations will develop new fatigue
monitoring software that does not use the simplified approach of
FatiguePro.

This report provides the technical basis for new, multiaxial SBF
technology developed to address the NRC concerns. It serves as a
requirements specification that can be used for the SBF
computations in fatigue monitoring software, and provides detailed
discussions about the rationale for various decisions that were made.

The SBF methodology described includes automatic calculations of


environmentally assisted fatigue, in accordance with applicable
regulatory guidance documents and the latest consensus among
industry experts on calculation methods.

Keywords
Fatigue
Environmental fatigue
FatiguePro
License renewal
Fatigue monitoring software

v

13307342
13307342
Acronyms and
Symbols The following acronyms are used in this report.

ASME American Society of Mechanical Engineers


ASTM American Society for Testing Materials
BWR Boiling Water Reactor
CUF Cumulative Usage Factor
EAF Environmentally-Assisted Fatigue
EPRI Electric Power Research Institute
FSRF Fatigue Strength Reduction Factor
FEA Finite Element Analysis
FEM Finite Element Model
GALL Generic Aging Lessons Learned
HWC Hydrogen Water Chemistry
MRP Materials Reliability Program
NRC Nuclear Regulatory Commission
NWC Normal Water Chemistry
OBE Operating Basis Earthquake
PWR Pressurized Water Reactor
RIS Regulatory Issue Summary
SBF Stress-Based Fatigue
SCF Stress Concentration Factor
SCL Stress Classification Line
SF Scaling Factor
PV Peak/Valley, or the stress extremum region

The following symbols are commonly used in this report.


Δε Strain amplitude

ε Strain rate
α Coefficient of thermal expansion
ν Poisson’s ratio
σ Stress
σm Membrane stress (≡ M)
σb Bending stress (≡ B)
σ1 (or S1) First principal stress
σ2 (or S2) Second principal stress
σ3 (or S3) Third principal stress
σI (or SI) Stress intensity

g vii h

13307342
σx (or SX) Normal stress in the X direction
σy (or SY) Normal stress in the Y direction
σz (or SZ) Normal stress in the Z direction
σij (or Sij) Shear stress in the i plane and j direction
(i=x,y,z and j=x,y,z; i≠j)

B Bending stress
DO Dissolved oxygen
E Modulus of elasticity
Ea Modulus of elasticity used in analysis
Ec Modulus of elasticity shown on applicable
fatigue curve
F Peak stress
Fen Fatigue life environmental correction factor
G(x,t) Unitized Green function for location x at
time t
G0 Steady state value of the unitized Green
function
Gˆ ( x, t ) Normalized Green’s function for location x
at time t
Ke Simplified elastic-plastic penalty factor
M Membrane stress
N Fatigue life (cycles) at alternating stress
S Sulfur content
P Primary stress
Q Secondary stress
Salt Alternating stress intensity
Sm Code allowable stress value
Sn P+Q stress intensity range
Sp Total stress (P+Q+F) intensity range
SX_n SX for primary (P) plus secondary (Q) stress
SX_p SX for primary (P) plus secondary (Q) plus
peak (F) stress

g viii h

13307342
T Temperature
Tref Reference temperature
U Cumulative usage factor
U0 Initial cumulative usage factor
Uen Environmental cumulative usage factor

g ix h

13307342
13307342
Table of Contents

Section 1: Introduction ............................................. 1-1


1.1 Purpose of this Report ................................................. 1-1
1.2 What is Fatigue Damage? .......................................... 1-2
1.3 ASME Code and Environmentally-Assisted Fatigue
Analysis .......................................................................... 1-3
1.4 What is Stress-Based Fatigue (SBF)? ............................. 1-5
1.5 Nuclear Industry and Regulatory Issues ......................... 1-6
1.6 Utility Benefits ............................................................ 1-7
1.6.1 License Renewal ................................................ 1-7
1.6.2 Operations and Maintenance ............................. 1-8
1.7 Outline of Report........................................................ 1-8

Section 2: Calculation of Stresses ............................. 2-1


2.1 Nomenclature for the Three-Dimensional Stress
Tensor............................................................................. 2-1
2.2 Principal Stresses and Stress Intensity............................ 2-2
2.3 Local Stress Calculations ............................................. 2-4
2.3.1 Linear Elastic Methodology ................................. 2-4
2.3.2 Input Objectives for Fatigue Analysis ................... 2-5
2.4 Stress Classification Lines and Linearized Stresses .......... 2-6
2.5 Instantaneous Stresses from Static Loads ....................... 2-8
2.5.1 Material Properties for Use with Static Load
Cases ........................................................................ 2-8
2.5.2 Peak Stresses at Discontinuities ............................ 2-9
2.5.3 Example Calculation ........................................ 2-10
2.6 Time-Dependent Thermal Stresses ............................... 2-11
2.6.1 Use of Green’s Functions .................................. 2-11
2.6.2 Film Coefficients for Use with Green’s
Functions ................................................................. 2-14
2.6.3 Material Properties for Use with Green’s
Functions ................................................................. 2-14
2.6.4 Through-Wall Thermal Stress Distribution ............ 2-15
2.6.5 Normalized Green’s Function ........................... 2-20

g xi h

13307342
2.6.6 Linearization of Thermal Stresses ....................... 2-21
2.6.7 Special Code Rules for Piping Secondary
Stresses ................................................................... 2-23
2.6.8 Peak Stresses at Discontinuities .......................... 2-25
2.7 Benchmark Calculations ............................................ 2-26

Section 3: Stress Cycle Counting and Fatigue


Calculations ............................................. 3-1
3.1 Introduction ............................................................... 3-1
3.1.1 The ASME Stress Cycle ...................................... 3-1
3.1.2 Order Dependence of Fatigue Damage................ 3-3
3.1.3 Order-Dependence and the ASME Code .............. 3-6
3.1.4 How Do We Know What’s Right? ....................... 3-7
3.2 Calculation of Fatigue Usage ...................................... 3-8
3.2.1 Alternating Stress Intensity (Salt) ............................ 3-9
3.2.2 Fatigue Usage Factor ....................................... 3-14
3.3 Peak and Valley Detection ........................................ 3-14
3.3.1 “Rubberband” PV Detection .............................. 3-15
3.4 Stress Cycle Pairing .................................................. 3-17
3.4.1 Background on Uniaxial Rainflow Method .......... 3-17
3.4.2 Adaptation of Rain Flow to Multiaxial Stresses
– “Rainflow-3D”........................................................ 3-20
3.5 Example Calculations ............................................... 3-21
3.5.1 Paperclip Example ........................................... 3-21
3.5.2 EAF Expert Panel Sample Problem ..................... 3-23
3.5.3 Combined Bending and Shear .......................... 3-24
3.5.4 Constant Principal Stresses and Stress Intensity .... 3-25
3.5.5 Evaluation of Real Data .................................... 3-27

Section 4: Environmentally-Assisted Fatigue


Usage Calculations .................................. 4-1
4.1 Background on EAF Calculations ................................. 4-1
4.2 Regulatory Bases........................................................ 4-2
4.2.1 Requirements for License Renewal ........................ 4-2
4.2.2 Requirements for New Plants ............................... 4-3
4.3 Fen Formulations for Ferritic Materials ............................ 4-3
4.3.1 NUREG/CR-6583 (Old Rules) ............................. 4-3
4.3.2 NUREG/CR-6909 (New Rules) ........................... 4-4
4.4 Fen Formulations for Austenitic Stainless Steel
Materials......................................................................... 4-5

g xii h

13307342
4.4.1 NUREG/CR-5704 (Old Rules) ............................. 4-5
4.4.2 NUREG/CR-6909 (New Rules) ........................... 4-5
4.5 Fen Formulations for Nickel Alloy .................................. 4-6
4.5.1 NUREG/CR-6909 (New Rules) ........................... 4-6
4.6 How to Compute Fen for a Stress Cycle.......................... 4-7
4.6.1 Strain Rate Calculations...................................... 4-7
4.6.2 Average versus Maximum Service
Temperature ............................................................... 4-8
4.6.3 Dissolved Oxygen.............................................. 4-9
4.6.4 Calculation of Fen Using the Modified Rate
Approach .................................................................. 4-9
4.6.5 Bounds of the Modified Rate Integration ............. 4-12
4.7 Example Fen Calculations ........................................... 4-13

Section 5: Summary and Conclusions ....................... 5-1

Section 6: References ............................................... 6-1


6.1 References................................................................. 6-1

Appendix A: SBF Example – PWR Charging


Nozzle .....................................................A-1
A.1 Description of the Monitored Location ......................... A-1
A.2 Finite Element Analysis .............................................. A-1
A.2.1 Unit Internal Pressure Analysis ............................ A-3
A.2.2 Branch Piping Interface Loads ............................ A-4
A.2.3 Run Piping Loads .............................................. A-5
A.2.4 Time-Dependent Thermal (Green’s Function)
Analyses ................................................................... A-5
A.3 Development of the SBF Transfer Functions................... A-7
A.4 Example Analyses ..................................................... A-7
A.4.1 Loss of Letdown with Delayed Return................... A-7

g xiii h

13307342
13307342
List of Figures

Figure 1-1 Fatigue Crack Nucleation and Propagation .............. 1-3


Figure 1-2 Illustration of ASME Code, Section III Fatigue
Curve ............................................................................. 1-4
Figure 2-1 Three-Dimensional State of Stress (Stress Tensor) ........ 2-2
Figure 2-2 Stress Classification Lines ........................................ 2-7
Figure 2-3 Example Linearized Membrane and Bending
Stresses ........................................................................... 2-8
Figure 2-4 Lagrange Polynomial ............................................ 2-16
Figure 2-5 Higher Order Lagrange Polynomial........................ 2-17
Figure 2-6 Lagrange Polynomial Fit Example .......................... 2-18
Figure 2-7 Example Green’s Function Configuration File .......... 2-19
Figure 2-8 Benchmark of Stress Linearization Using p=6 .......... 2-23
Figure 2-9 Sample Transfer Function Input .............................. 2-26
Figure 2-10 Sample Transfer Function Loads Input ................... 2-27
Figure 2-11 Benchmark of Green’s Function Calculations ......... 2-28
Figure 2-12 Benchmark of Linearized Thermal Stresses ............ 2-29
Figure 2-13 Benchmark of Metal Temperature Calculations ...... 2-30
Figure 2-14 Benchmark of P+Q Stress Calculations at
Discontinuity with Multiple Loading Types ......................... 2-31
Figure 2-15 Benchmark of P+Q+F Stress Calculations at
Discontinuity with Multiple Loading Types ......................... 2-32
Figure 3-1 Idealized ASME Stress Cycle ................................... 3-2
Figure 3-2 Design versus Real Transient ................................... 3-3
Figure 3-3 Traditional Design Fatigue Analysis Example ............ 3-4
Figure 3-4 Monitoring Fatigue Analysis Example ....................... 3-5
Figure 3-5 Example of Seismic Order Dependence in a
Design Analysis ............................................................... 3-7

g xv h

13307342
Figure 3-6 Illustration of Peak and Valley Time Windows ........... 3-9
Figure 3-7 Example Correlation Between Sn and Ke ................. 3-11
Figure 3-8 PV Time Window Example .................................... 3-15
Figure 3-9 Rainflow Visualization Example ............................. 3-18
Figure 3-10 Example Uniaxial Stress History for Rainflow
Analysis ........................................................................ 3-19
Figure 3-11 “Paperclip” Simulated Stress History..................... 3-22
Figure 3-12 Peak and Valley for EPRI EAF Expert Panel
Sample Problem, SCL1/TRAN1 ....................................... 3-24
Figure 3-13 Combined Bending and Shear Stress History ........ 3-25
Figure 3-14 Constant Principal Stress and Stress Intensity
History .......................................................................... 3-26
Figure 4-1 Parametric Analysis Using Ke in CUF and Strain
Rate................................................................................ 4-8
Figure 4-2 Complex Transient with Multiple Stress Cycles ......... 4-12
Figure 4-3 EAF Expert Panel Sample Problem EAF
Parameters .................................................................... 4-13
Figure A-1 Charging Nozzle Finite Element Quarter Model ....... A-2
Figure A-2 Charging Nozzle Monitored Location (the SCL) ....... A-3
Figure A-3 Charging Nozzle Unit Pressure Analysis.................. A-4
Figure A-4 Charging Nozzle Finite Element 360° Model........... A-5
Figure A-5 Charging Nozzle Convection Surfaces.................... A-6
Figure A-6 Loss of Letdown with Delayed Return to Service ........ A-8
Figure A-7 Stresses and Stress Pairs - Loss of Letdown with
Delayed Return to Service ................................................ A-9

g xvi h

13307342
List of Tables

Table 2-1 Time History Stress and Temperature Calculation


Objectives ....................................................................... 2-5
Table 2-2 Nomenclature for Stresses ........................................ 2-5
Table 2-3 Example Linearized Stresses from Static Unit
Loads............................................................................ 2-10
Table 2-4 Example Combined Stresses for Static Loads ............ 2-11
Table 2-5 Green’s Function Configuration File Format .............. 2-20
Table 3-1 Example Correlation Table Ke vs. Sn ........................ 3-11
Table 3-2 Example Temperature Dependent E and Sm Values .... 3-12
Table 3-3 Example Uniaxial Stress History Spectrum
Summary....................................................................... 3-19
Table 3-4 “Paperclip” Fatigue Analysis Results ........................ 3-23
Table 3-5 Fatigue Analysis Results for EPRI EAF Expert Panel
Sample Problem, SCL1/TRAN1 ....................................... 3-24
Table 3-6 Fatigue Analysis Results for Combined Bending
and Shear Stress ............................................................ 3-25
Table 3-7 Fatigue Analysis Results for Constant Principal
Stress and Stress Intensity ................................................ 3-26
Table 4-1 Example Dissolved Oxygen History ........................... 4-9
Table A-1 Fatigue Usage - Loss of Letdown with Delayed
Return to Service ............................................................. A-9
Table A-2 EAF - Loss of Letdown with Delayed Return to
Service .......................................................................... A-9

g xvii h

13307342
13307342
Section 1: Introduction
1.1 Purpose of this Report

The purpose of this report is to describe the technical basis and requirements of
the Fatigue monitoring software Stress-Based Fatigue (SBF) module, which is
used to monitor fatigue damage of components in nuclear power plants.

This report provides a brief introduction to fatigue damage, including the effects
of reactor water environmental conditions on fatigue initiation, but primarily
serves as a requirements specification for the SBF calculations. Accordingly, it
provides a more complete understanding of how the SBF calculations are
performed in order to monitor against the potential fatigue damage mechanism
and to meet regulatory acceptance criteria.

This report also provides more than just the technical requirements in the form
of various formulas, procedures and algorithms, but offers transparency of design.
Because some of the technology developed here is newly adapted and specific
interpretations of the ASME Code rules and regulatory guidance documents
need to be made, discussions of the technical basis or rationale for how the
various technical decisions were made are also provided.

The primary reasons for development of new and enhanced SBF technology, as
documented in this report, are to provide users with the ability to perform the
following:
1. Multiaxial stress calculations to address Nuclear Regulatory Commission
(NRC) concerns documented in Regulatory Issue Summary (RIS) 2008-30
about the use of a simplified, single stress term in fatigue analyses [1, 2, 3],
and
2. Automatic calculations of environmentally-assisted fatigue (EAF), in
accordance with applicable regulatory guidance documents [13, 17].

For individual users, by understanding the technical basis and the specific
algorithms used to calculate environmentally-assisted fatigue (EAF) usage one
can customize and use SBF calculations more intelligently, making better use of
its capabilities, while also being aware of any limitations. Each end user is not
expected to study and learn all of the information in this document; he or she
needs only to refer to sections of it as required to aid in understanding the
computations performed by the software.

g 1-1 h

13307342
This report does not, and cannot, present a complete description and
interpretation of either the ASME Code, Subarticle NB-3200 fatigue analysis
requirements or the applicable regulatory documents associated with EAF
calculations. To further understand the theory or technical basis behind the
ASME Code rules or the regulatory rules for computing EAF, one should refer
to the various references available on the topics.

1.2 What is Fatigue Damage?

Time-varying thermal, pressure and mechanical loads produce perturbations of


stress (stress cycles) primarily at the surface of a component. Stress cycles of
sufficient magnitude cause fatigue damage, which can ultimately lead to cracking
of the component. When subjected to continued cyclic loading, the crack might
propagate exponentially through the wall of a pressure retaining component and
lead to a leak of the primary system coolant, or, depending on the nature of the
crack and loading distribution, even general structural failure.

Fatigue initiation is predominantly a surface phenomenon and is closely related


to the phenomenon of “slip” due to shear stresses. Slip occurs in ductile metals
within individual grains by dislocations moving along a crystallographic plane.

As fatigue damage continues, cracks may initiate in local slip bands and initially
tend to grow (approximately several grains) in a plane along the maximum shear
stress, as shown in Figure 1-1. This stage (Stage I) is called Fatigue Crack
Nucleation. As cycling continues, fatigue cracks tend to coalesce and grow along
planes of maximum tensile stress range. This stage (Stage II) is called Fatigue
Crack Propagation.

Extensive research has also demonstrated that fatigue damage can be accelerated
in a reactor water environment when certain conditions, related to temperature,
strain rate, dissolved oxygen levels, and material content, are present [5].

g 1-2 h

13307342
Figure 1-1
Fatigue Craack Nucleatio
on and Propag
gation

1.3 ASME Code and Environm


mentally-A
Assisted Fa
atigue Analysis

For analysiss of nuclear power


p plant Class
C 1 vessel and
a piping coomponents in
accordance with the rulees of Section III
I of the ASM ME Boiler an nd Pressure Vessel
V
Code [4], a fatigue analyysis is required if the structture is subjectt to significan
nt
cyclic loadings. The anallysis procedurres are set forrth in ASME Code Subartticle
NB-3200. ForF the purpoose of component initial deesign, the anaalysis is based d on a
conservativve set of design transients.

Fatigue testts performed in support off ASME Cod de fatigue curvve developmeent
used small smooth speciimens cyclicallly loaded in ana air environ nment. The
applied alteernating stresss was related to the numbeer of cycles un
ntil the tensilee
stress decreeased 25% froom the steady--state value. For
F these speccimens this looad
drop corressponded to an n “engineeringg crack” of ap
pproximately 3-mm
3 in deptth.

g 1-3 h

13307342
The mean fatigue curve was then conservatively adjusted for the maximum
possible mean stress effect, and other factors such as scatter, size effect, surface
finish, industrial environment, etc. A more in-depth discussion of the technical
basis for the ASME Code fatigue curve can be found in NUREG/CR-6909 [5].

1000
Alternating Stress Intensity, Sa (ksi)

100

10
10 100 1000 10000 100000 1000000
Allowable Cycles, N

Figure 1-2
Illustration of ASME Code, Section III Fatigue Curve

ASME Section III design qualification does not allow for the presence of planar
flaws. While the ASME Code fatigue curves were based on test specimen flaw
nucleation plus growth to an “engineering crack” of about 3-mm in depth, the
ASME Code cumulative fatigue usage factor (U or CUF) acceptance criteria of
less than or equal to a value of 1.0 is nevertheless intended to provide assurance
that no crack has been formed.

For Class 1 components in the non-isolable pressure boundary in new nuclear


plants and plants in license renewal space, additional penalties (Fen factors) may
need to be applied to the ASME Code CUF calculations to account for the
reduction in life due to the reactor water environment (Uen). Fen factors (see
Chapter 4) are based on parameters such as temperature, strain rate, bulk
dissolved oxygen levels, and material content. A Uen exceeding the acceptance
criteria of 1.0 in the period of extended operation requires a corrective action
(e.g., fatigue reanalysis, inspection and flaw tolerance analysis, repair or
replacement, etc.).

g 1-4 h

13307342
To perform the fatigue analysis, alternating stress intensities based on the
controlling differences of stress components must be determined. Alternating
stress intensity is related to the shear stress range on a critical plane, as illustrated
in Figure 1-1 by the initial orientation of the fatigue crack at the surface.

Guidance for determining stress differences is provided in ASME Code


NB-3216. For any specific location, there are two approaches specified.
1. Per NB-3216.1, guidance is provided if the direction of the principal
stresses remains constant. This is the case if all stress components remain
proportional. This is also the case if shear stresses in the principal stress
coordinate system are zero, such as typically exists for surface stresses in
pressure vessels and piping.
2. Per NB-3216.2, guidance is provided if the directions of the principal
stresses do not remain constant. For this case, the values of the normal and
shear stress components may vary arbitrarily, and the "stress cube" that
determines principal stresses may rotate.

Consideration of all six unique stress components of the stress tensor in


accordance with NB-3216.2 is expected by the NRC of all licensees, as stated in
NRC RIS 2008-30. For this case, derivation of the stress differences, especially
for randomly ordered stresses, as would be computed using stress-based fatigue
monitoring, is not straightforward. This is even recognized in NB-3216.2(b),
which suggests that the analyst should calculate stress differences for multiple
different stress-point pairs to discover which pair produces the largest alternating
stress intensity.

Identifying the extreme pair is the crux of the problem.. The challenge arises
because the difference in computed stress intensity between two points in a cycle
is not equal to the stress intensity difference, since stress intensity is a nonlinear
operator. The stress intensity difference must be determined based on the
difference of the stress components in going from one point in a cycle to another.
This issue and a solution are discussed in detail in the sections to follow.

1.4 What is Stress-Based Fatigue (SBF)?

Stress-Based Fatigue (SBF) is a methodology for computing a CUF for a


component based on stress histories computed using its actual operating
experience.

A traditional ASME Code design analysis computes a CUF for a pressure-


retaining plant component to assure that the component can sustain the cyclic
service loads defined in the component specifications without initiating a fatigue
crack. This is done using “design transients”, which are simplified, idealized
models of expected behavior designed to be more severe than any actual transient.
The principle is that if the component can withstand a certain number of each
design transients without initiating a crack, it clearly must be able to withstand a
similar number of (less damaging) actual transients.

 1-5 

13307342
In stress-based fatigue, the design transients are replaced by computing a more
realistic stress response of components to the plant transients as they have
actually occurred. This computed real-time stress history may be analyzed using
methods similar to the traditional design analyses to develop a stress-intensity
loading spectrum derived from the actual operating history. From that spectrum,
a CUF may be computed that reflects a conservative but much more realistic
cyclic fatigue duty of the component in question.

SBF is a monitoring approach, in that one must use an operating history in


performing the calculations.

Using this additional information, one need not make overly conservative
assumptions about how the plant will operate. Thus, SBF monitoring can be
used with reasonable conservatism to demonstrate lower CUFs than even the
most rigorous design fatigue analyses.

1.5 Nuclear Industry and Regulatory Issues

Nuclear plants in the United States were originally licensed by the NRC for
40 years of operation. Most utilities have or are currently submitting applications
to the NRC in order to renew the license and extend operation to a period of
60 years. Research is also being conducted to assess the feasibility and develop
solutions to safely extend plant lives to 80 years. Plants are required to manage
the aging effects of systems, structures and components in the scope of license
renewal. Part 54 to Title 10 of the U.S. Code of Federal Regulations (10CFR54)
specifies the “Requirements for Renewal of Operating Licenses for Nuclear
Power Plants”.

Many nuclear plants are making commitments to implement fatigue monitoring


systems in support of license renewal. NRC report NUREG-1801, the “Generic
Aging Lessons Learned (GALL) Report” [18], identifies acceptable aging
management programs, including programs for fatigue and cyclic operation.
Although the NRC concluded in a risk study reported in NUREG/CR-6674
that the effect of a reactor water environment is not a safety issue requiring
corrective actions in the original, 40 year licensing basis, the NRC does require
of all license renewal applicants to assess the fatigue effect from a reactor water
environment for the entire period of extended operation. High fatigue locations
studied in NRC report NUREG/CR-6260 [19], at a minimum, must be
evaluated for environmentally-assisted fatigue, in addition to other locations
in the Class 1 pressure boundary that may be more limiting.

New nuclear plants must perform, as part of the design basis, environmentally-
assisted fatigue analyses of Class 1 pressure retaining components in accordance
with Regulatory Guide 1.207 [13]. Monitoring may be used to demonstrate
compliance.

g 1-6 h

13307342
The NRC requires all of these fatigue calculations to be performed in accordance
with Section III of the ASME Code. Subarticle NB-3200 of Section III of the
ASME Code defines a “design by analysis” procedure for qualifying Class 1
nuclear components for fatigue usage. This generalized approach considers all
six unique components of the stress tensor in determining the alternating stress
intensity between extreme stress states for input to the fatigue curve. By contrast,
many nuclear licensees’ previous fatigue calculations have used a simplified
approach, where a scalar (1D) stress value versus time is computed, and then used
to determine peaks and valleys for the fatigue usage computation. This simplified
approach was initially chosen in the 1980’s, due to limitations in desktop
computer capabilities. It has been shown to be conservative for most applications,
which exercised good judgment on the part of the designer of the fatigue
monitoring logic. The intention was that a thorough analysis of the three
dimensional (3D) stress problem would be performed, and the FatiguePro
1D stress would be conservatively constructed to effectively bound the full 3D
stress-intensity range for all significant stress load pairs at that component [7].
Historically, many license renewal applicants have used the simplified approach
to assist in performing environmental fatigue analyses.

In recent years the NRC Staff has increased the scrutiny level on TLAA issues
in general, and environmentally assisted fatigue (EAF) analyses in particular.
Regulatory Issue Summary RIS-2008-30 on metal fatigue in nuclear plants has
been issued, challenging the simplified approach [1, 2, 3]. Specific attention has
been given to locations with asymmetric geometry (e.g., nozzle corners), and/or
locations with asymmetric loading (e.g., thermal stratification), due to the
engineering judgment that is necessary to develop conservative, single stress term
logic for these locations. This has required additional work to justify the use of
these methods, such as comprehensive plant-specific benchmark calculations of
the various component locations using all six stress terms or independent,
confirmatory analyses.

1.6 Utility Benefits

1.6.1 License Renewal

The principal reason utilities have implemented SBF monitoring in nuclear


power plants is that it assists with ongoing environmentally-assisted fatigue
usage calculations, performed for the extension of plant design life.

In order to produce acceptable results when considering the effects of a reactor


water environment, a significant decrease in conservatisms is typically required in
the fatigue calculations. NUREG/CR-6260 concluded that “the best method to
lower the cumulative fatigue usage for the few worst locations appears to be
fatigue monitoring.” As a result, plants often opt for computing fatigue based
on actual plant operating behavior, instead of the more idealized design basis
transient assumptions.

g 1-7 h

13307342
1.6.2 Operations and Maintenance

When transient cycle counts exceed the allowable numbers, as identified in the
plant licensing bases, or when plant components experience events in which the
operating parameters such as assumed maximum temperature ranges are
exceeded, SBF monitoring can justify continued operation.

The use of SBF monitoring can also minimize analyst judgments that are
otherwise inherently necessary to be exercised when using the results of analyzed
design transients and cycle counting alone. Real transients are sometimes
much more complex than idealized transient definitions. The consideration of
environmental effects, where for example EAF is maximized at a particular
temperature ramp rate, can challenge original design assumptions, which often
assumed that an instantaneous step change in temperature is mathematically
bounding.

SBF calculated fatigue usage of a component may exhibit trends, enabling reactor
operators to modify plant operational procedures to mitigate fatigue damage
accumulation. Transients may also be simulated to much more quickly and easily
determine the fatigue damage impacts of proposed new methods of operation,
relative to the current modes of operation.

When operating procedures are modified, power uprates undertaken, etc., the
design transient can change. Operational changes can require complete reanalysis
of components using new design transients or justification that the existing ones
are still bounding. Because SBF performs calculations using plant instrument
data, without regard to the counting or categorization of design transients and
their assumed severity, engineering judgment about the level of conservatism of
the assumed, idealized design transients relative to the actual operation is
reduced. Any operational changes are reflected in the plant data, which is
analyzed using the same SBF system accordingly.

1.7 Outline of Report

The following chapters provide a detailed specification for each of these processes
used to ultimately calculate environmentally-assisted fatigue, based on input
plant data.

Calculation of Stresses. Specifies how to compute a multiaxial Primary plus


Secondary and Total stress history, in accordance with ASME Code Subarticle
NB-3200, using input plant data.

Stress Cycle Counting and Fatigue Calculations. Using the multiaxial stress
history, specifies how to detect stress peaks and valleys, stress cycles, and fatigue
usage in accordance with ASME NB-3200.

Automatic Calculation of Environmentally-Assisted Fatigue. Specifies how to


calculate environmentally-assisted fatigue usage, in accordance with applicable
regulatory requirements, using the SBF results. The methodology is based on the
latest consensus opinions from EPRI’s Environmental Fatigue Expert Panel.

g 1-8 h

13307342
Section 2: Calculation of Stresses
This section discusses how to compute local stress response to time-varying plant
input data.

2.1 Nomenclature for the Three-Dimensional Stress Tensor

The theory of stress discussed here and used by the ASME Code Subarticle NB-
3200 is applicable to any continuum, regardless of the mechanical properties of
the material.

A free body is first established in order to define the stress component


terminology. A three-dimensional state of stress on an infinitesimal body is
illustrated in Figure 2-1. This state of stress, or the stress tensor, has six unique
components composed of three normal stress values (σx, σy, σz.) and three shear
stress values (σxy, σyz , σxz ).

The outward normal of each surface points away from the interior of the free
body. The right-hand face of the stress cube in Figure 2-1 is called the x plane;
its outward normal points in the positive x direction. The left-hand face is the
negative x plane. Positive normal stresses are tension and negative normal stresses
are compression. The normal stress on the x surface is noted by simply σx. Since
a normal stress acts in the direction normal to its plane, only one subscript is
necessary for this component.

A shear stress may act in any direction in its plane. On the x plane shear
components act in the y and z directions. These two shear stresses on the x plane
are denoted σxy and σxz. Shear stresses are positive when the two applicable
positive axes rotate toward each other when subject to the shear stresses.

Note that the nomenclature and sign conventions assumed here are consistent
with those utilized by the ANSYS finite element software [20], an industry
standard in the nuclear industry.

g 2-1 h

13307342
Figure 2-1
Three-Dimensional State of Stress (Stress Tensor)

2.2 Principal Stresses and Stress Intensity

When the coordinate system of the stress cube is rotated into a specific
orientation, there exists an orthogonal set of axes, 1, 2, and 3, called principal
axes. With respect to these principal axes, the stress components are all zero,
except for those in the principal direction. In other words, there always exists a
set of mutually perpendicular planes with zero shear stresses on those planes.
These planes are called principal planes. The directions of their normals are
called principal directions.

The principal stresses (σ1, σ2, σ3) are the roots of the cubic equation, or
eigenvalues, calculated from the stress components by the cubic equation:

σ x −σ0 σ xy σ xz
σ xy σ y −σ0 σ yz = 0 Equation 2-1
σ xz σ yz σ z −σ0

Or,

σ03 - (σx + σy + σz) σ02 + (σx σy + σy σz + σz σx - σxy2 - σxz2 - σyz2 ) σ0


- (σx σy σz + 2 σxy σxz σyz - σz σxy2 - σy σxz2 - σx σyz2 ) = 0

where:

σo = principal stress (3 values)

g 2-2 h

13307342
The three principal stresses are labeled σ1, σ2, and σ3 (fatigue monitoring software
output quantities S1, S2, and S3) and are calculated as follows.

σ1 =
I1 2
+
3 3
( q )cos φ

σ2 =
I1 2
+
3 3
( q )cos⎛⎜φ + 23π ⎞⎟
⎝ ⎠

σ2 =
I1 2
+
3 3
( q )cos⎛⎜φ + 43π ⎞⎟
⎝ ⎠

where:

2
I 1 − 3I 2 if (I12 – 3I2) > 0

q=

0 Otherwise

1 ⎛ 2 I 3 − 9 I 1 I 2 + 27 I 3 ⎞
cos −1 ⎜⎜ 1 3/ 2

⎟ if q > 0
3 ⎝ 2 ( q ) ⎠
φ=

0 Otherwise

I1 = σ x + σ y + σ z

I 2 = σ xσ y + σ y σ z + σ z σ x − σ xy − σ yz − σ xz
2 2 2

I 3 = σ xσ y σ z − σ xσ yz − σ y σ xz − σ z σ xy + 2σ xy σ yz σ xz
2 2 2

Principal stresses are then reordered from highest to lowest such that:

σ1 ≥ σ 2 ≥ σ 3

g 2-3 h

13307342
The stress intensity σI (Fatigue monitoring software output quantity SI) is the
largest of the absolute values of σ1 - σ2, σ2 - σ3, or σ3 - σ1. That is:

σ I = MAX (σ 1 − σ 2 , σ 2 − σ 3 , σ 3 − σ 1 ) Equation 2-2

The planes of maximum shear are sometimes referred to as principal shear stress
planes. It can be shown that maximum shear can be evaluated based on principal
stresses as follows.

σ 1, 2 = (σ 1 − σ 2 ) / 2

σ 2,3 = (σ 2 − σ 3 ) / 2

σ 3,1 = (σ 3 − σ 1 ) / 2

The three potential values of stress intensity are twice the shear stresses on the
principal shear stress planes. As discussed previously, fatigue crack nucleation is
related to slip along planes of maximum shear.

2.3 Local Stress Calculations

2.3.1 Linear Elastic Methodology

SBF monitoring relies on the assumption that stresses are computed linear
elastically. Non-linear plasticity effects are accounted for by computing elastic-
plastic penalty factors (Ke) in accordance with the requirements of ASME Code
Subarticle NB-3200. The linear elastic methodology allows use of the principle
of superposition of stresses. The specifics of this methodology are described in
detail in the sections to follow.

The linear elastic stress state for a monitored location may be computed based on
a linear summation of the stresses caused by various types of loads. Most pressure
vessels and piping system components include stresses due to pressure, thermal
(local temperature distribution), and boundary interface loads, such as forces and
moments caused by thermal expansion, thermal stratification, anchor
displacement, seismic movement, etc. Deadweight and residual stresses are not
needed, because they do not vary with time and therefore do not impact the
computed stress range; they effectively cancel out in linear elastic analyses
(ASME Section III Code rules do not directly allow taking mean stress effects
into account; the worst case situation is inherently built into the fatigue curve).

For a linear elastic stress analysis, stress contributions may be classified as one of
two types:
1. Stresses due to static loads, such as pressure, piping thermal expansion,
seismic, etc. that are directly scalable to pertinent parameters (pressure,
temperature, etc.), and

g 2-4 h

13307342
2. Time-dependent thermal stresses, which vary with time, based on the axial
and radial temperature distributions of the component.

Stress contributions of the second type, which depend on the transient history,
are calculated by a time integration of the product of a predetermined Green’s
function, or influence function, and the transient temperature data. The specifics
of these calculations for each are described below.

2.3.2 Input Objectives for Fatigue Analysis

An objective of the stress calculations is to first determine a time history of the


six individual stress components for the Primary plus Secondary (P+Q) stresses,
and the six for Primary plus Secondary plus Peak (F), or Total (P+Q+F), stresses,
per the stress classification requirements of Table NB-3217-1 and Table
NB-3217-2 of the ASME Code. To improve accuracy and reduce conservatism,
stresses will be evaluated at a 1 second interval (versions up to 3.0 of FatiguePro
used a 10 second interval).

The metal temperature at the surface of the monitored location is also desired, in
addition to the twelve stress values, to enable selection of temperature-dependent
material properties and to support environmentally-assisted fatigue calculations,
etc.

An example of a time history of these values is shown in Table 2-1.

Table 2-1
Time History Stress and Temperature Calculation Objectives

Primary plus Secondary (P+Q) Primary plus Secondary plus Peak (P+Q+F)
Time SX SY SZ SXY SYZ SXZ SX SY SZ SXY SYZ SXZ TEMP
0 0.61106 4.8532 37.055 -0.8166 1.0629 -0.6831 2.866 0.18652 44.606 -0.2931 1.187 -0.802 504.34
1 0.82283 4.9202 37.334 -0.846 1.0727 -0.6605 4.2418 0.23268 46.13 -0.3036 1.2467 -0.7781 500.26
2 1.2317 5.0773 37.905 -0.9157 1.0911 -0.6367 6.8963 0.32585 49.133 -0.3294 1.3658 -0.7541 492.2
3 1.7373 5.2927 38.641 -1.0117 1.1145 -0.6121 9.9906 0.44148 52.715 -0.3664 1.5092 -0.7301 482.54
4 2.3212 5.549 39.513 -1.1266 1.1421 -0.587 13.245 0.57054 56.567 -0.4122 1.6641 -0.7061 472.1
5 2.9786 5.8394 40.511 -1.2574 1.1737 -0.5615 16.576 0.70999 60.594 -0.4657 1.8266 -0.6822 461.14
6 3.7004 6.1553 41.622 -1.4009 1.2091 -0.5358 19.93 0.85665 64.727 -0.5255 1.9937 -0.6584 449.85
7 4.483 6.4935 42.839 -1.5558 1.2482 -0.5098 23.287 1.0094 68.938 -0.591 2.1644 -0.6347 438.3
8 5.3265 6.8539 44.164 -1.7221 1.2908 -0.4836 26.646 1.1681 73.228 -0.6621 2.3385 -0.611 426.49
9 6.2206 7.2324 45.582 -1.898 1.3367 -0.4572 29.995 1.3316 77.576 -0.7379 2.5152 -0.5875 414.48
10 7.1538 7.6239 47.077 -2.0813 1.3854 -0.4306 33.321 1.4982 81.957 -0.8172 2.6933 -0.564 402.36

Table 2-2 shows the stress terminology used in this report versus the terms to be
output in the fatigue monitoring software.

Table 2-2
Nomenclature for Stresses

σx σy σz σ xy σ xz σ yz σ1 σ2 σ3 σI
SX SY SZ SXY SXZ SYZ S1 S2 S3 SI

g 2-5 h

13307342
Stress components are appended with _n to represent Primary plus Secondary
(P+Q) stresses, and _p to represent total Primary plus Secondary plus Peak
(P+Q+F) stresses. For example,

SX_n SX stress component for P+Q stress


SX_p SX stress component for P+Q+F stresses

2.4 Stress Classification Lines and Linearized Stresses

A stress classification line, or SCL, is a location where fatigue usage will be


evaluated and is generally represented by a straight line spanning from the inside
to the outside of the vessel or piping component, as illustrated in Figure 2-2. For
most components an SCL is perpendicular to the inner and/or outer surfaces.
Alternatively, it may be selected based on a line orientation that is generally
perpendicular to the computed stress contours. The non-mandatory Annex 5.A
of ASME Section VIII, Division 2 [11] provides guidance in selecting SCLs.
This guidance is consistent with the requirements of ASME Section III
Subarticle NB-3200.

Multiple SCLs are typically selected for each component, based on consideration
of highest stress regions or geometric, material or other discontinuities, such as
the presence of a notch or weld. In the Figure 2-2 example, SCL1 is located at a
field butt weld, which is a discontinuity requiring a fatigue strength reduction
factor (FSRF). SCL2 is located at a geometric discontinuity, where there may be
additional bending and shear stresses due to axial thermal gradients through the
thickness transition.

As a conservative default, the ASME Code Primary plus Secondary (P+Q)


stresses may be considered to be equivalent to the sum of the linearized
membrane plus bending (M+B) stresses through the thickness of a section
in the pressure vessel or piping component (i.e., the SCL) for each loading
contribution. Allowed exceptions to this conservative rule (e.g., for piping
components) are described later.

The linearization methodology utilized in the SBF calculations is known


as “Cartesian” linearization in the ANSYS finite element software and
“Linearization for three-dimensional structures” in the ABAQUS finite element
software. The non-mandatory Annex 5.A of ASME Section VIII, Division 2
(Section 5.A.4.1.2) [11] also provides guidance for performing stress linearization
calculations using this same approach. These linearization methods are consistent
with the requirements of ASME Code Subarticle NB-3200; they were used and
found to be reasonable by the EAF Expert Panel, as noted in a PVP paper on the
topic [21].

g 2-6 h

13307342
Figure 2-2
Stress Classification Lines

The membrane stress (M) is the average stress across the SCL, and the bending
stress (B) is the through-wall equivalent linear bending stress that produces the
same net bending moment as the actual stress distribution across the SCL.

Peak (F) stresses are the difference between Total stress (after consideration of
SCFs or FSRFs, if applicable) and P+Q. The specifics of these calculations are
provided in sections that follow.

Figure 2-3 is an example that illustrates the linearization of stresses across the
section of a component. TOTAL is the actual, as-computed through-wall stress
distribution for the SY component. MEMBRANE is the average of TOTAL
over the length of the SCL. BEND is the equivalent linear stress distribution
that produces the same net bending moment across the SCL as the actual stress
distribution. MEM+BEND is the sum of MEMBRANE and BEND. All
integrations are through the thickness, assuming a flat plate (infinite radius of
curvature). This linearization is performed for each of the six unique stress
components of Figure 2-1.

g 2-7 h

13307342
Figure 2-3
Example Linearized Membrane and Bending Stresses

2.5 Instantaneous Stresses from Static Loads

For static loads, such as internal pressure and piping interface loads, linearized
stress results for a unit load may be computed using finite element analysis (FEA)
software and scaled directly to the applicable operating parameter, such as
pressure or temperature.

2.5.1 Material Properties for Use with Static Load Cases

Instantaneous (selected at a certain temperature) material properties may


generally be utilized for these analyses, because the linear elastic stresses due to
the unit loads will be scaled. For these loads, the stresses are typically insensitive
to the temperature at which the specific material properties are selected. For
example, calculated pressure stresses are often related to the geometry only
(e.g., Pr/t), as are those due to piping interface loads (F/A, Mc/I, etc.). However,
care must be taken that material properties are selected such that a reasonable
level of accuracy of the computed stresses is attained for the particular situation.

g 2-8 h

13307342
2.5.2 Peak Stresses at Discontinuities

For static loads, the contributions to Primary plus Secondary (P+Q) stresses
consist of the linearized membrane plus bending (M+B) stresses.

When no discontinuity is present at the SCL, the Primary plus Secondary plus
Peak (P+Q+F) stresses are simply the TOTAL stress computed by the FEA.

At SCLs with discontinuities that the FEM does not directly handle, such as a
stress concentration factor (SCF) at a sharp notch or a fatigue strength reduction
factor (FSRF) at a field butt weld, the Total stress is determined by the product
of the linearized (M+B) stresses and the SCF or FSRF. Different SCFs or
FSRFs may be utilized at a particular SCL depending on the static load type. For
example, for the internal pressure static load the ASME NB-3600 stress index of
1.2 (K1) may be applied at a piping field butt weld, while the piping thermal
expansion static load may utilize a stress index of 1.8 (K2) at the same SCL.

A static load’s contribution to the Total (P+Q+F) stresses is as follows (the SX


stress component is shown as an example, and the same relationship is utilized
for each of the remaining 5 Total stress components).

(FSRF/SCF)(SF)(SX_MB) if (FSRF/SCF) > 1

SX_p =
Equation 2-3
(SF)(SX_TOT) Otherwise

where:
FSRF/SCF FSRF or SCF applicable to the SCL and static load type
SF Scaling factor on the unit static load to the current operating
condition
SX_MB SX membrane plus bending stress resulting from the unit static
load (calculated directly by FEA)
SX_TOT SX total stress resulting from the unit static load (calculated
directly by FEA)

g 2-9 h

13307342
2.5.3 Example Calculation

An example calculation is shown below.

For the component shown in Figure 2-2, the linearized stresses due to a unit 1
psig internal pressure and a unit 1 in-kip bending moment in the Z direction
(Mz) have been computed by the FEA software and are shown in Table 2-3.
SCL1 is assumed to be located at a weld with an FSRF of 2.0 (2.0 is applicable
to each static load), and SCL2 has no SCF/FSRF.

Table 2-3
Example Linearized Stresses from Static Unit Loads
*********** M+B STRESSES ************ ********** TOTAL STRESSES **********
Unit Load
SX_MB SY_MB SZ_MB SXY_MB SX_TOT SY_TOT SZ_TOT SXY_TOT
Section = SCL1
P = 1 psig -0.9974 1.426 4.523 -0.05612 -0.9974 1.425 4.578 -0.00195
Mz = 1 in-kip 0.002446 5.104 0.8132 -0.1144 0.002446 5.132 0.8189 -0.00345
Section = SCL2
P = 1 psig -0.9993 1.217 3.713 -0.03126 -0.9993 1.131 3.765 0.006565
Mz = 1 in-kip -0.00699 2.226 0.8519 -0.6705 -0.00699 2.138 0.8592 0.01199
Table note: All stresses in psi. In this example, the model is axisymmetric, and so two of the shear stresses, SYZ and SXZ, are
zero and therefore not shown. The stress component labels are appended with _MB to indicate (M+B) and _TOT to indicate
TOTAL.

The P+Q and P+Q+F stress contributions when the component is subjected at
a moment in time to an internal pressure of 2250 psig and a bending moment
Mz of -3000 in-kip are determined as follows.

The Primary + Secondary (P+Q) stresses at each SCL are determined by


multiplying the linearized (M+B) stresses from the unit loads by the scalable
parameter (pressure=2250 or bending moment = -3000).

The P+Q+F stresses at SCL1 are determined by multiplying the linearized


(M+B) stresses from the unit loads by the scalable parameter and the FSRF of
2.0.

Since SCL2 has no FSRF, the P+Q+F stresses at the section are determined by
multiplying the TOTAL stresses from the unit loads by the scalable parameter.
The result is shown in Table 2-4.

g 2-10 h

13307342
Table 2-4
Example Combined Stresses for Static Loads
Primary + Secondary (P+Q) Primary + Secondary + Peak (P+Q+F)
Scaled Load
SX_n SY_n SZ_n SXY_n SX_p SY_p SZ_p SXY_p
SCL1
P = 2250 psig -2244.2 3208.5 10176.8 -126.3 -4488.3 6417.0 20353.5 -252.5
Mz = -3000 in-kip -7.3 -15312.0 -2439.6 343.2 -14.7 -30624.0 -4879.2 686.4
Combined -2251.5 -12103.5 7737.2 216.9 -4503.0 -24207.0 15474.3 433.9
SCL2
P = 2250 psig -2248.4 2738.3 8354.3 -70.3 -2248.4 2544.8 8471.3 14.8
Mz = -3000 in-kip 21.0 -6678.0 -2555.7 2011.5 21.0 -6414.0 -2577.6 -36.0
Combined -2227.5 -3939.8 5798.6 1941.2 -2227.5 -3869.3 5893.7 -21.2
Table note: In this example, the model is axisymmetric, and so two of the shear stresses, SYZ and SXZ, are zero and
therefore not shown. The stress component labels below are appended with _n to indicate (P+Q) stresses and _p to
indicate (P+Q+F) stresses.

2.6 Time-Dependent Thermal Stresses

Fatigue Monitoring Software uses Green’s functions, but does not generate
them. An engineer must generate the Green’s functions to be used by performing
a thermal stress analysis. Green’s functions are specific to the monitored
component and the operating conditions it experiences and are included in the
plant-specific configuration. Guidance for generating Green’s functions is
provided in this section.

2.6.1 Use of Green’s Functions

Thermal stresses are time-dependent and are computed based on the transient
temperature distribution throughout the component at a particular time. The
thermal stresses are therefore computed with Green’s functions, otherwise known
as influence functions, integrated with the loading temperature. This Green’s
function integration is similar in concept to the well-known Duhamel theory
used in structural dynamics.

The basic principle behind Green’s functions is that the stress response at a
particular component location due to an applied change in temperature on a
surface of the component is directly proportional to the magnitude of the
applied change in temperature. These days Green’s functions are almost always
developed using detailed FEA, which determines the transient stress response at
an SCL to a step change in loading (a thermal shock of an arbitrary magnitude).
The stress response typically consists of the stress peaking at a certain time
following the initial shock and eventually stabilizing to a steady state condition.
This response is the Green’s function.

For an arbitrary loading temperature history, the applied temperature can be


visualized as a series of discrete changes in temperature at each time step (the
forcing function). Each change produces a forward “wave” of stresses. The total
stress versus time is determined using the principle of superposition of the linear

g 2-11 h

13307342
elastic stresses. Metal temperatures at points in the structure may be computed
with Green’s functions the same way.

A detailed derivation of this approach and examples of its application to specific


plant locations are contained in References [6] and [7]. In short, the Green’s
function approach calculates a time-dependent stress component through a
convolution integral. Note that the approach documented below is applicable for
computing all six components of the thermal stress at a given point. Metal
temperatures at points may also be computed using the same Green’s functions
methodology. Calculations for a single stress term are presented for convenience.

t

σ T ( P , t ) = ∫ G ( P, t − γ ) φ (γ ) ∂γ Equation 2-4
0
∂γ

t

= G0 ( P) (φ (t ) − Tref ) + ∫ Gˆ ( P, t − γ ) φ (γ ) ∂γ
t −t d
∂γ

t
= G 0 ( P ) (φ (t ) − Tref ) + ∑ Gˆ ( P, t − γ ) Δφ (γ )
t −t d

where:

G0 ( P ) = lim G ( P, t ) (the Green’s function steady-state value)


t →∞

td = time when G ( P, t ) ≈ G 0 ( P ) (the decay period of the Green’s


function)

Gˆ ( P, t ) = G ( P, t ) − G0 ( P) (the normalized Green’s function)

Tref = Stress-free temperature of the component (zero if computing


a temperature).

Explaining the rest of the terms:

σT(P,t) is the thermal stress at location P at time t,


G(P,t) is the unitized Green’s function for location P, elapsed time t
φ(t) is the local temperature (Tlocal) at time t,
Δφ(t) is the change in Tlocal at time t,
t the time at which the stress is computed,
γ the parameter of integration.

The approach above assumes that the Green’s function G(P,t) has been unitized.
This means that the initial stress is zero, because the component is initially at
steady state conditions at the stress free temperature, and a step change of one
degree in temperature is imposed at time 0+. The process for converting a stress
analysis with an arbitrary step change in temperature to the normalized Green’s
function format is discussed in Section 2.6.5.
g 2-12 h

13307342
It is desirable to optimize the integration process for speed and accuracy. With
the new, multiaxial stress methodology described in this report, considerably
more data needs to be evaluated; dozens of Green’s function integrations at a 1
second interval may be evaluated.

The integration can be optimized if we take advantage of the fact that local
temperature is recorded as a sequential series of values, and we treat the
temperature history φ(t) as a continuous, piece-wise linear function connecting
those values. The optimal calculation would then consist of one term for each
linear portion of the temperature history. This approach is derived below:

t

σ T ( P , t ) = ∫ G ( P, t − γ ) φ (γ ) ∂γ
0
∂γ

t ∂
= G0 ( P ) ( φ ( t ) − Tref ) + ∫ Ĝ( P , t − γ ) φ( γ ) ∂γ
t −t d ∂γ


Since φ(t) is piece-wise linear, φ (γ ) is a constant for each linear segment and
∂γ
can be factored out of the integral. The equation above becomes:

τ

n −1 i

σ T ( P, t ) = G0 ( P) (φ (t ) − Tref ) + ∑ φ (τ i ) ∫ Gˆ ( P, t − γ ) ∂γ Equation 2-5


i = 0 ∂γ τ i +1

where:

τi = the collection of n times, over the last td seconds, when the Tlocal
values are defined, with τ0 = t, τ1 is the most recent prior point, τ2
is the previous point, etc. (i.e., i increases moving back in time).
∂ φ (τ i ) − φ (τ i +1 )
φ (τ i ) = = the slope of the segment between τi+1 and τi,
∂γ (τ i − τ i +1 )
and

τi

∫τ Gˆ ( P, t − γ ) ∂γ = the area under the curve Ĝ(P,t) between t-τi+1 and t-τi.
i +1

The sum of the areas under the curve at each time step in the Green’s function
can be predetermined, so that the area between t-τi+1 and t-τi can be computed
based on a simple difference in the two values.

The optimization process described here has been shown to dramatically increase
speed relative to performing the same number of integrations using the previous
algorithms. The optimized process does not reduce accuracy. Accuracy is
enhanced due to the smaller time steps used.

g 2-13 h

13307342
2.6.2 Film Coefficients for Use with Green’s Functions

One feature of the Green’s function method is that it assumes a constant heat
transfer coefficient between the fluid and the metal. Typically, a conservative heat
transfer coefficient is computed for a bounding range of the service temperature
and flow rates.

We can approximate the effect of varying heat transfer coefficients by defining


multiple Green’s functions that cover a range of heat transfer coefficient values.
By switching between theses Green’s functions as the conditions change, we get
an approximate solution for varying heat transfer coefficient.

2.6.3 Material Properties for Use with Green’s Functions

The Green’s function methodology relies on the principle of a linear stress


response to the input forcing function. As such, constant (selected at a certain
temperature) material properties are typically used for these analyses. In general,
there are a number of options for selecting the material properties for use in
generating the Green’s function. Three are discussed below.
1. Selection of constant material properties at an approximate mean value of the
service temperature.

For example, if a component has a service temperature range from approximately


70°F to 650°F, then material properties at 350°F, the approximate mean
temperature, may be utilized. In this case, the instantaneous, instead of the mean,
coefficient of thermal expansion is chosen at the given temperature. Selection of
instantaneous properties at the mean service temperature has been demonstrated
to be reasonably accurate for most applications.
2. Use of temperature dependent material properties with an applied
temperature shock that is typical of a dominant transient.

If a component is dominated by a single transient, for example a downward


shock from 550°F to 100°F, then it may be appropriate to utilize temperature
dependent material properties for generating the Green’s function. Instead of an
arbitrary (unit) temperature shock, the actual shock from 550°F to 100°F would
be applied, and the FEA program would automatically select temperature
dependent properties, based on the calculated temperatures. The Green’s
function input stresses would need to be normalized to account for the fact that
they were based on a downward shock of 450°F, however the mathematics of the
integration remain the same.
3. Selection of bounding material properties over the range of temperature.

Bounding material properties over the range of temperature may be selected for
the Green’s function thermal stress analysis. In general, the magnitude of the
stress response is dependent on two material properties: (1) the product of the
elastic modulus (E) times the coefficient of thermal expansion (α), or Eα, and (2)
the thermal diffusivity, TD, which is the thermal conductivity (k) divided by the
density (ρ) times specific heat (Cp); that is, TD = k/(ρCp).

g 2-14 h

13307342
For plate or cylinder subjected to a thermal shock, the maximum thermal stress
intensity response will generally be maximized if Eα is maximized and TD is
minimized [8, Figure 3.15-1]. An analyst may ensure conservatism by choosing
E and α at the temperature where the product of the two is maximized, and k
and Cp at the temperature where TD is minimized (ρ of steel is generally
constant over typical service temperature ranges).

This process may yield over-conservatism for some applications, and may not
necessarily be bounding for components with multiple material types with
competing factors. For example, choosing the minimum conductivity for a
component’s thermal sleeve may yield unconservatism, as it would result in
increasing the shielding effect on the component.

Ultimately, an inexperienced analyst may need to perform a benchmark


calculation if complex or competing factors are involved in order to select
material properties for the Green’s function analysis that will result in
conservative or accurate results.

2.6.4 Through-Wall Thermal Stress Distribution

An accurate history of the linearized membrane plus bending thermal stresses


is desired. Unlike for static loads, thermal stresses are time-dependent. The
linearized thermal stresses are therefore not directly scalable relative to any single
parameter. For this reason, a detailed knowledge of the through-wall thermal
stress distribution is needed, in order to continually perform the linearization as
the distribution varies with time.

Modern finite element software typically performs linearization of stresses at an


SCL via numerical integration using a large number of points. For example, the
ANSYS finite element software determines the stresses at 49 points through-wall
to perform the linearization, regardless of how many nodes and elements across
the thickness of the section are used in the model. The method used to fit,
smooth and/or extrapolate the stresses to these many locations may be complex
and varies from one finite element program to another.

Instead of replicating this many calculations or performing a more crude


numerical integration with a much smaller number of points, it is convenient to
represent the through-wall thermal stress distribution using an equation. For a
given set of distinct points (x, σ(x)) that are equally spaced along the x axis, as
shown in Figure 2-4, a Lagrange Polynomial of the following form can be
determined.

σ ( x) = H 0 + H 1 x + H 2 x 2 + ... + H p x p Equation 2-6

where:

σ = stress component (e.g., σx, or σy, etc.)


x = Distance along the SCL, where x = 0 at the inside and x = t
(the thickness) at the outside.

g 2-15 h

13307342
‐86
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

‐88

‐90

‐92
σ(x)

‐94

‐96

‐98

‐100

‐102
x

Figure 2-4
Lagrange Polynomial

The Lagrange Polynomial evaluates stresses at the known points precisely, where
they are computed using Green’s Functions, and provides a convenient equation
for representing the stress distribution between those known points. It is
important to note that using a higher degree of polynomial is not always
preferable, as higher order polynomials can generate unusual shapes, as shown in
Figure 2-5. While using a higher degree polynomial does not impact the total
stress calculations, it can impact the linearized stress calculations.

g 2-16 h

13307342
Figure 2-5
Higher Order Lagrange Polynomial

Figure 2-6 is an example of an actual thermal stress distribution, as computed by


ANSYS, for a typical pressure vessel SCL and a distribution calculated using the
Lagrange Polynomial fit, showing excellent agreement, using a polynomial
degree (p) of only 3.

Fatigue monitoring software can also be configured to use a Piece-Wise Linear


fit, instead of the Lagrange Polynomial fit. An example linear fit is shown in
Figure 2-6. It is apparent that when comparing the two types, Lagrange
Polynomial can be more accurate than the Piece-Wise Linear fit, when a fewer
number of divisions is selected. With an increased number of divisions, the
Piece-Wise Linear fit is not subject to the errors associated with a higher order
Lagrange polynomial, as shown in Figure 2-5.

g 2-17 h

13307342
50000

0
0 0.5 1 1.5 2 2.5

‐50000
Hoop Stress (psi)

‐100000

SZ (Lagrange Fit)
SZ (ANSYS)

‐150000 Piece‐Wise Linear

‐200000

‐250000

‐300000
Distance x (inches)

Figure 2-6
Lagrange Polynomial Fit Example

In order to calculate the through-wall thermal stresses, a significant number of


Green’s function calculations need to be performed. For example, if there are
eight points through-wall, then 56 Green’s function integrations need to
performed (eight points times six components of stress plus temperature). As
such, much more information is needed to perform the calculations, compared to
versions up to 3.0 of FatiguePro that used a simplified single stress term.

An ANSYS macro was developed to support multiaxial, through-wall stress and


temperature calculations. This macro generates a file that contains information
about the SCL, the applied load (temperature step and flow rate or film
coefficient) and blocks of data over time including the stresses and temperatures.
Figure 2-7 shows the format of the Green’s function file (while this example was
generated using the ANSYS finite element software, the information could be
developed independent of ANSYS, assuming the same format is used).

g 2-18 h

13307342
Figure 2-7
Example Green’s Function Configuration File

The format of the Green’s function file is described in Table 2-5. Different
values on a single line may be separated by either spaces or tab stops.

g 2-19 h

13307342
Table 2-5
Green’s Function Configuration File Format

Line Format Description


1 N/A This line is used as a comment field.
2 Node1 Inside Node number of the SCL.*
3 Node2 Outside Node number of the SCL.*
4 p The degree of the polynomial, which is also
the number of through-wall divisions at which
stress points were extracted. For example,
if p=8, there are 8 divisions through-wall
composed of 9 points at which stresses are
calculated.
5 nTimeBlocks Number of time blocks in the Green’s function
file.
6 RSYS Coordinate system of the output stresses.*
7 t Length of the SCL path (the linear distance
between Node1 and Node2)
8 XYZ The X, Y, Z global coordinates of Node 1.*
9 XYZ The X, Y, Z global coordinates of Node 2.*
10 TREF Stress-free reference temperature used in the
stress analysis.
11 T1 T2 The temperature range of the arbitrary step
change in temperature. T1 is the initial
temperature, which is stepped to T2.
12 HCOEF The film coefficient at the inside node or flow
rate used to generate the Green’s function.*
13** Time Time value for data block
14 – SX SY SZ SXY SYZ SXZ TEMP The six components of stress plus the metal
(14+p)** temperature. One line of input is included for
each of the points through-wall.
Table Notes:
* This does not affect the calculations and is included for information and verification purposes only.
** These lines (which compose the “time block”) are repeated nTimeBlocks times. For example, the
“Time” line of the second time block will start at line number 13+p+2.

2.6.5 Normalized Green’s Function

The following process is used to convert the results of the stress analysis,
described in the previous section, to a Normalized Green’s Function.

g 2-20 h

13307342
First, we determine the unit steady-state Green’s function response. By
definition, the steady-state thermal stress in the component at T = TREF is zero
(0). The Green’s function methodology assumes that stress response is linear
with respect to temperature. Since the stress analysis run ends at time = td with
the component at steady-state for a surface fluid temperature of T2, we have:

σT,ss = G0∙(T2 – TREF) = Rslt(td)

Therefore, the unit steady-state stress response may be calculated as:

G0 = Rslt(td) / (T2 – TREF) Equation 2-7

The normalized Green’s function, Ĝ(τ), can now be determined by subtracting


the total steady-state response from all data points, and then dividing by the
magnitude of the temperature step:

Ĝ(τ) = (Rslt(τ) – Rslt(td)) / (T2 – T1) Equation 2-8

where:

Rslt(t) = Stress history from the finite element analysis.


T1 = The initial temperature of the FEA, from Line 11 of
Table 2-5.
T2 = The final temperature of the FEA, from Line 11 of Table 2-5.
TREF = The stress-free temperature used by the FEA, from Line 10 of
Table 2-5.

2.6.6 Linearization of Thermal Stresses

Using the Lagrange Polynomial fit, linearization of stresses may be computed


using integrals. Linearization computations are performed using the Cartesian
method, as discussed earlier. This stress linearization procedure is described in
Section 5.A.4.1.2 of ANNEX 5.A of ASME Section VIII, Division 2.

The membrane stress, σm, is the average stress across the section and is computed
as follows. A membrane stress at an SCL is calculated for each of the six stress
components.

t t
1 1
σm =
t0∫ t0
( )
σ ( x)dx = ∫ H 0 + H 1 x + H 2 x 2 + ... + H p x p dx Equation 2-9

The bending stress, σb, is the equivalent linear bending stress that produces the
same net bending moment across the section as the actual stress distribution. A
bending stress at the inside of an SCL is calculated for each of the six stress
components and is computed as follows.

 2-21 

13307342
t t

∫ (H )
6 ⎛t ⎞ 6 ⎛t ⎞
2 ∫
σb = σ ( x)⎜ − x ⎟dx = 2 0 + H 1 x + H 2 x 2 + ... + H p x p ⎜ − x ⎟dx Equation 2-10
t 0 ⎝2 ⎠ t 0 ⎝2 ⎠

For outside locations, the bending stress is the same value but with an opposite
sign (-σb).

For a Piece-Wise Linear fit, the membrane stresses are calculated as follows.

1 ⎡ σ 0 σ p p −1 ⎤
σm = ⎢ + + ∑σ i ⎥ Equation 2-11
p⎣ 2 2 i =1 ⎦

Where σ0 is the stress at the inside surface (0th point in a zero-based array), and
σp is the stress on the outside surface (pth point in a zero-based array).

Bending stresses are determined by using the integral formulation above, i.e.:

t
6 ⎛t ⎞
2 ∫
σb = σ ( x)⎜ − x ⎟dx Equation 2-12
t 0 ⎝2 ⎠

t
⎛t ⎞
However, the integral σ ( x)⎜∫
0 ⎝2
− x ⎟dx is evaluated as the sum of the area

under each piece-wise linear segment [σ(x)dx] times the distance to the centroid
of that area (t/2-xc).

Figure 2-8 is a comparison of ANSYS linearized stresses to those computed with


a Lagrange Polynomial fit and Piece-Wise Linear fit and the linearization
integrations shown above. Seven stress points through wall were extracted
(polynomial degree p=6). Excellent agreement is attained between the
independent linearization methods.

g 2-22 h

13307342
120000

100000

80000

60000

TOT (Lagrange)
Stress (psi)

40000
M+B (Lagrange)
M (Lagrange)
TOT (Piece‐Wise Linear)
20000
M+B (Piece‐Wise Linear)
M (Piece‐Wise Linear)

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

‐20000

Method M+B (psi) % diff


‐40000 ANSYS 50410 --
Lagrange 50585 0.3%
Linear 49200 -2.4%
‐60000
Distance x (inches)

Figure 2-8
Benchmark of Stress Linearization Using p=6

2.6.7 Special Code Rules for Piping Secondary Stresses

In the 1977 edition, Summer 1979 addenda of the ASME Code a subtle but
very significant change was made to the Section III piping rules. Specifically,
the stress due to linear, radial temperature distribution, which was previously
classified as a secondary stress, was re-classified as a peak stress. As a result,
the ΔT1 stress term in Equation 10 of the NB-3600 rules was removed. The
NB-3200 rules for piping were also changed accordingly. Table NB-3217-2 was
reworded to state that stresses due to radial temperature distribution (both the
linear and the nonlinear portion), are peak stresses. Stresses due to axial
temperature distribution must still be considered secondary.

For most components, thermal stresses are composed of contributions from both
axial and radial temperature distribution. However, when calculating thermal
stresses using FEA for an NB-3200 fatigue analysis, there is no direct way in the
software itself to separate the resulting values into axial and radial contributions.
That is, the linearized membrane plus bending thermal stress will contain stresses
due to both axial and radial temperature distribution.

g 2-23 h

13307342
A PIPING flag is therefore included in the SBF module to adjust the secondary
stresses. If PIPING is set to TRUE, the location will classify the stress due to
radial temperature distribution as peak, and adjust the secondary stresses
accordingly. In order to use this option, all of the following conditions must be
met:
1. The component is part of a piping system (eligible for NB-3600 rules).
2. The SCL is located at a region of piping where axial and radial have clear
meanings. For example, the butt weld adjoining a nozzle to a section of
straight pipe would meet the criteria. The nozzle corner of a pipe to pipe tee
intersection would not.
3. The stresses are output in a cylindrical coordinate system, such that
SX=radial, SY=hoop, and SZ=axial.

The axial and hoop stress in an infinite cylinder due to linear, radial temperature
distribution may be determined using the ΔT1 stress term in Equation 11 of NB-
3600.

1
σ ΔT = sgn Eα ⋅ ΔT1 Equation 2-13
1
2(1 − ν )

where (use the same values used in the FEA for the Green’s function, unless
justified otherwise):

E = Elastic modulus
α = Coefficient of thermal expansion
ν = Poisson’s ratio
sgn = 1 for the inside surface and -1 for the outside surface

t
6 ⎛t ⎞
2 ∫
ΔT1 =− T ( x )⎜ − x ⎟dx
t 0 ⎝2 ⎠

T(x) = Lagrange polynomial representing temperature distribution


at the SCL (computed using Green’s functions).

The radial and shear stress terms are negligible for pure, radial temperature
distribution.

The linear σ ΔT1 stress is then subtracted from the previously calculated hoop and
axial membrane plus bending stress components. The secondary thermal stresses
are then:

SX_n = SX_MB
SY_n = SY_MB - σ ΔT1
SZ_n = SZ_MB - σ ΔT1

g 2-24 h

13307342
SXY_n= SXY_MB
SYZ_n = SYZ_MB
SXZ_n = SXZ_MB

Using this option does not impact the calculation of Total (Primary + Secondary
+ Peak) stresses. This option impacts the Secondary (Primary + Secondary)
stresses, and therefore potentially the Ke value and alternating stress intensity.

2.6.8 Peak Stresses at Discontinuities

When accounting for the effects of discontinuities in the calculation of total


thermal stresses, the process is the same as that for static loads. That is, the
linearized membrane plus bending thermal stress is multiplied by the SCF or
FSRF, except that the thermal peak is also superimposed. A thermal load’s
contribution to the Total (P+Q+F) stresses is as follows (the SX stress
component is shown as an example, and the same relationship is utilized for each
of the remaining five Total stress components).

(FSRF/SCF)(SX_MB) + SX_PEAK if (FSRF/SCF) > 1

SX_p =
Equation 2-14
SX_TOT Otherwise

where:
FSRF/SCF FSRF or SCF applicable to the SCL and thermal load
SX_MB SX linearized M+B thermal stress computed using
Green’s functions
SX_PEAK SX PEAK thermal stress computed using Green’s
functions
SX_TOT SX TOTAL thermal stress computed using Green’s
functions

Adding the thermal peak back in to the Total stress is analogous to the ΔT2 stress
term in NB-3600. Section 6.7 of the Companion Guide to the ASME Boiler &
Pressure Vessel Code, Volume 1 [9] provides guidance on handling of thermal
stresses. This guidance is consistent with the requirements of ASME Section III
rules which indicate that peak thermal stresses result from a local temperature
constraint rather than stress flow and are not concentrated by notches, and that
membrane-plus-bending stress must be developed before applying the SCF while
the thermal peak stresses must be superimposed after applying the SCF.

g 2-25 h

13307342
A counterargument to this guidance is that the thermal peak will contain peak
stresses due to both the through-thickness thermal gradient and whatever notch
or discontinuity is modeled in the FEM itself. However, the process, by default,
is consistently conservative.

2.7 Benchmark Calculations

This report serves as a requirements specification for development of fatigue


monitoring software. Detailed benchmarking of the fatigue monitoring software
will be performed and documented in a separate report using the final software.

However, during the process of developing this new technology, technical


prototype software was developed to benchmark and validate the stress
calculation methodologies as outlined in this report. The software calculates
thermal stresses using the optimized Green’s functions methodology previously
described, and has the ability to superimpose other stresses due to static loads,
while including the effects of SCFs/FSRFs, etc. for simulated transients.

Transfer function inputs are shown in Figure 2-9. These include unit linearized
and total stresses due to static loads, SCFs/FSRFs (K1 applies to pressure, K2 to
piping moments, and K3 to thermal), the Green’s function configuration files,
and the stress-free temperatures.

Figure 2-9
Sample Transfer Function Input

g 2-26 h

13307342
Figure 2-10 shows input for the simulated transient, including a time history of
the local temperatures, pressures, scaling factoring for the piping interface
moments, etc.

Figure 2-10
Sample Transfer Function Loads Input

For SCL1 shown on the FEM model inFigure 2-2, Green’s functions were
developed using constant material properties at the average service temperature.
The simulated transient consists of a simple injection of cold water (producing
a tensile peak) followed by an injection of hot water (producing a compressive
valley). The transient stresses were computed using Green’s function integrations
of the simulated transient. The same transient was also analyzed directly in the
ANSYS finite element software using both the same, constant material
properties, as well as temperature dependent material properties.

Figure 2-11 is a benchmark of the total axial thermal stresses at the inside
surface. Note that these results show thermal stresses only without the impact
of an SCF/FSRF to assess the accuracy of the Green’s function methodology
specifically. The results show very close agreement between the Green’s function
methodology and the independent calculations using the FEA software, which
tested both constant and temperature-dependent material properties.

g 2-27 h

13307342
SY_ANSYS (constant MPs) SY_FatiguePro SY_ANSYS (variable MPs)

100

80

60

40

20
Stress at SCL_1 (ksi)

0
0 500 1000 1500 2000 2500 3000 3500

‐20

‐40

‐60

‐80

‐100
Time (seconds)

Figure 2-11
Benchmark of Green’s Function Calculations

Figure 2-12 is a benchmark of the linearized membrane plus bending thermal


stresses using the multi-dimensional Green’s function and Lagrange Polynomial
methodology defined earlier. The methodology was quite accurate quantitatively
and qualitatively in nature, and in this case was conservative compared to
independent ANSYS thermal stress and linearization calculations that considered
both constant and temperature-dependent material properties.

g 2-28 h

13307342
SY_MB_ANSYS (constant MPs) SY_MB_FatiguePro SY_MB_ANSYS variable MPs)

60

Stress Range (ksi)


40 106.8 FatiguePro
104.5 ANSYS (constant MPs) ‐2.2%
Lineareizd Membrane plus Bending Stress at SCL_1 (ksi)

20
101.4 ANSYS (variable MPs) ‐5.4%

0
0 500 1000 1500 2000 2500 3000 3500

‐20

‐40

‐60

‐80
Time (seconds)

Figure 2-12
Benchmark of Linearized Thermal Stresses

Figure 2-13 demonstrates the suitability of calculating metal temperature at the


surface using the Green’s function methodology.

g 2-29 h

13307342
Tm_ANSYS (constant MPs) Tm_FatiguePro Tm_ANSYS (variable MPs)

700

600

500

400

300

200

100

0
0 500 1000 1500 2000 2500 3000 3500

Figure 2-13
Benchmark of Metal Temperature Calculations

The next set of plots includes the effects of the other loading contributions
(pressure and bending moment) plus an FSRF of 2.0, to account for the presence
of a field weld. Figure 2-14 shows the Primary + Secondary (P+Q) stresses, and
Figure 2-15 shows the total Primary + Secondary + Peak (P+Q+F) stresses. Note
that this benchmark is a completely independent validation: The ANSYS results
use internal calculations of thermal stress, temperature dependent material
properties, and the ANSYS method of linearization of stresses; the fatigue
monitoring software results use Green’s function calculations of thermal stress,
constant material properties at the average service temperature, and the Lagrange
Polynomial method of linearization of stresses.

g 2-30 h

13307342
80

60

40

20 SX_n
SY_n
SZ_n
SXY_n
0
0 500 1000 1500 2000 2500 3000 3500 SX_n_FP
SY_n_FP
SZ_n_FP
‐20 SXY_n_FP

‐40

‐60

‐80

Figure 2-14
Benchmark of P+Q Stress Calculations at Discontinuity with Multiple Loading Types

g 2-31 h

13307342
200

150

100

50 SX_p
SY_p
SZ_p
SXY_p
0
0 500 1000 1500 2000 2500 3000 3500 SX_p_FP
SY_p_FP
SZ_p_FP
‐50 SXY_p_FP

‐100

‐150

‐200

Figure 2-15
Benchmark of P+Q+F Stress Calculations at Discontinuity with Multiple Loading
Types

g 2-32 h

13307342
Section 3: Stress Cycle Counting and
Fatigue Calculations
3.1 Introduction

ASME Section III, Subarticle NB-3200 establishes criteria to demonstrate


the “suitability of a component for specified service loadings involving cyclic
application of loads and thermal conditions.” This process often involves a fatigue
analysis in accordance with NB-3222.4(e). Step one of this analysis is as follows:
“For each condition of normal service, determine the stress differences and the
alternating stress intensity Sa in accordance with NB-3216.” The general case of
this procedure, defined in NB-3216.2, involves calculating stress vs. time for each
“complete stress cycle” and choosing a point in time “when the conditions are one
of the extremes for the cycle”.

Properly enumerating stress cycles and identifying the extreme conditions of


actual, ordered multiaxial stress histories, while meeting the requirements of the
ASME Code, requires a non-trivial solution, as we will explain.

3.1.1 The ASME Stress Cycle

The ASME Code NB-3213.16 defines a stress cycle in NB-3213.16; it represents


the situation where the stress in a component varies from an initial state, to a
maximum extent, to a minimum extent, then back to the initial state.

Figure 3-1 illustrates an idealized stress cycle using a uniaxial stress history. In
this example, the stress begins at an initial state, experiences a local maximum
(the “peak”), followed by a local minimum (the “valley”), and then returns to the
initial state. The alternating stress intensity, Salt, which is used to determine the
allowable number of cycles on the fatigue curve, is one half of the range between
the peak and the valley (Salt is further adjusted for any effects of plasticity and
temperature). An operational cycle, or a plant transient, can contain multiple
stress cycles. Stress peaks and valleys in pressure vessels and piping design
analyses are typically associated with sudden or rapid upward or downward
changes in temperature or pressure or abrupt changes in temperature slopes.

g 3-1 h

13307342
Stress

2· σalt

Time

Figure 3-1
Idealized ASME Stress Cycle

In plant monitoring real-time stress histories derived from measured parameters


rarely resemble the idealized design transients, and often contain many smaller
internal stress cycles that hide the larger overall stress cycles. In addition, the
identification of extreme stress states is complicated by the multiaxial state of
stresses (different stress components reach their maximum extents at different
times), and the time phasing between peak and secondary stresses and the
temperature of the component.

Figure 3-2 illustrates how real transients can be much more complex than the
idealized ones used for design purposes. It is apparent from this example that
considering every reversal of a stress component as “peaks” and “valleys” and
pairing them appropriately for fatigue evaluation will not result in a correct
CUF factor.

g 3-2 h

13307342
60000 600

50000
500

40000

400 SX
SY
SZ

Temperature (F)
30000
Stress (psi)

SXY
SYZ
300
SXZ
TCHG
20000
Tm
TCOLD
200 DESIGN_TCHG

10000

100
0

‐10000 0

Figure 3-2
Design versus Real Transient

3.1.2 Order Dependence of Fatigue Damage

In an ASME Code design fatigue analysis, the order of future transients is


unknown (because they have not occurred yet). They are therefore assumed to
occur in the worst possible order. In addition, transients that consist of multiple
peaks and valleys are typically split so that each significant, extreme condition is
treated like a separate event.

For example, potential peaks and valleys may be identified based on the change
in sign of the slope of any of the three principal stresses. Peaks and valleys are
then assumed to occur in the worst possible order, to mathematically bound the
stress range and fatigue usage calculations for conservatism, often regardless of
whether that assumption of order is physically possible or not.

To demonstrate the importance of order, consider the following example.


Transient 1 occurs around the zero load state, say when the plant is first
beginning to startup, and consists of a local valley. Transient 2 occurs around a
higher mean stress state, say when the plant is operating at 100% power, and
consists of a local peak. Ten occurrences of each transient are assumed for design
purposes.

g 3-3 h

13307342
Figure 3-3 illustrates the fatigue usage calculation, using a traditional design
analysis. In this case, assuming that Transients 1 and 2 alternate in sequence
for a total of ten times produces the largest stress range and fatigue usage. The
maximum stress range is 200 ksi (150 ksi – (-50 ksi)), and the alternating stress
intensity is further multiplied by an assumed plastic penalty factor, Ke, of 3.333,
because of the large stress range. This calculation assumes that the component in
question will experience the maximum possible stress range between the two
conditions ten times.

TRANSIENT 1 TRANSIENT 2
n = 10 cycles n = 10 cycles
Time Stress Time Stress
0 0 0 100
1 ‐50 1 150
2 0 2 100
0 150

‐50 100

Ke = 3.333
Salt = [150 ‐ (‐50)]/2*Ke = 333.3 ksi
Nallow = 54.47
CUF = n/Nallow = 0.18

Figure 3-3
Traditional Design Fatigue Analysis Example

Now suppose we have the same transient set with the same severity of stresses,
but in this case the transients have already occurred, and we know the order in
which they happened. In this case, Transient 1 occurred ten times followed by
Transient 2 occurring ten times, as shown in Figure 3-4.

Since the order is known, it can be seen that the stress history consists of one
large stress cycle with a stress range of 200 ksi, highlighted in red, followed by 18
smaller stress cycles with a stress range of 50 ksi, highlighted in yellow. The large
stress range has the same plastic penalty factor of 3.333, but the smaller ranges
have no penalty factors (Ke=1.0).

Comparing the two analyses shows that the design fatigue analysis calculated a
fatigue usage factor (0.18) ten times larger than the monitoring fatigue usage
factor (0.018) which evaluated the stress cycles in the order that they actually
occurred.

g 3-4 h

13307342
One knows intuitively, and sound science and engineering demonstrates, that
with the monitoring example above, the order dependent fatigue calculation is
more appropriate than the design one, when that order is known.

Stress History
200

150

100

Stress (ksi)

50

‐50

‐100

Sp Ke Salt n Nallow U
200 3.333 333.3 1 54.47 0.01835873
50 1 25 18 1481072 1.21534E‐05
CUF = 0.018

Figure 3-4
Monitoring Fatigue Analysis Example

Fatigue’s order dependent principle can be validated simplistically by applying a


similar loading to a paperclip. For example, start by bending it back and forth
with a large range that produces plasticity (permanent deformation) until the
paperclip cracks. Next, wiggle another paperclip back and forth the same number
of times but with a small stress range (elastic response), followed by one large
bend that introduces plastic deformation, followed by the same number of small
wiggles back and forth, before returning it to its original position. It will be
apparent that the damage introduced in the second test is much less severe than
the first.

Aside from simply reducing conservatism, it is important to explicitly consider


order in monitoring for the following reasons.
1. Real data tends to produce stresses that routinely fluctuate in ways not
assumed with idealized transient definitions. Isolating each fluctuation
(the peak or valley) and pairing them in the worst-possible order can produce
untrue conclusions and an over-conservative fatigue calculation, even higher
than the design CUF calculations, despite the fact that the actual operation is
much less severe.

g 3-5 h

13307342
2. A CUF computed using worst order assumptions is mathematically
conservative, but may not be indicating anything real either quantitatively or
qualitatively. An over-conservative stress cycle counting methodology can
lead to incorrect conclusions about which component may actually be the
most limiting in a system and the first to potentially require corrective
actions, such as augmented inspections, repair or replacement.
3. Over-conservative methodologies may inherently lead to users needing to
manually adjust unreasonable results, based on engineering judgment. A
robust stress cycle counting algorithm obviates the need for user judgments,
thereby providing reproducibility of the calculations and enhancing quality.

3.1.3 Order-Dependence and the ASME Code

It is important to note that the ASME Code does not prohibit consideration of
order. ASME Section III is a design Code intended to provide assurance of
structural integrity for a set of transients with an unknown order of future
occurrences. However, to the extent that the order of stress cycles can be
known the Code does allow consideration of this.

An example of order dependence in design space is the handling of operating


basis earthquake (OBE) events in fatigue evaluations. A design fatigue evaluation
may consider, for example, 10 events over the life of the plant, each with 20
internal cycles. The design specification may further postulate that an OBE event
can happen at any time, superimposing the earthquake during the peak or valley
of the transient pair with the highest stress range.

Figure 3-5 illustrates a traditional method of accounting for the OBE event’s
impact on the stress range. Because the seismic cycling occurs quickly, the stress
range is essentially stretched once by one half of the OBE stress range, and
the remainder of the OBE cycling self-pairs. This logic is consistent with the
order-dependent stress cycle counting, which is specified later in this report.

g 3-6 h

13307342
Internal OBE cycles

New transient stress


range increased by OBE
stress amplitude

Transient stress
range without OBE

Figure 3-5
Example of Seismic Order Dependence in a Design Analysis

3.1.4 How Do We Know What’s Right?

Since the Code does not provide specific procedures or guidance for performing
stress cycle counting for random, ordered data, we must develop algorithms that
meet the requirements of the ASME Code. Indeed, this is a primary goal of this
research project.

The following criteria were defined to guide evaluation of the suitability of any
algorithm:
1. Accuracy. When stress cycles are evaluated in the same order, benchmark
calculations should reproduce precisely known ASME NB-3200 design
calculation problems.
2. Validation. We should confirm that the results make physical sense
and, while meeting Code requirements, validate that the results are
consistent with sound science and engineering principles related to the
order-dependency of fatigue damage. For example, assuming a uniaxial stress
with random, ordered loading, we should with our algorithms identify
virtually the same stress cycles as that from heavily vetted algorithms such
as Rainflow.
3. Repeatability. Analyst judgments and manual adjustments to the stress cycle
counting should not be necessary to produce consistently meaningful results.

g 3-7 h

13307342
Peak and valley detection and stress cycle counting algorithms to meet these
challenges are provided later in this chapter.

3.2 Calculation of Fatigue Usage

Before describing the peak and valley detection and stress cycle counting logic, it
is necessary to define how the fatigue usage is computed for a stress cycle, which
consists of a peak and valley combination. Fatigue usage for a stress cycle is a
function of the stress intensity range, which itself is used to identify and count
the stress cycles.

Figure 3-6 illustrates a stress history with a potential peak and valley (this
example is for an axisymmetric structure, and thus only one shear stress value is
shown). Let us assume here that a peak and valley detection algorithm has
identified a potential peak to be located somewhere in the highlighted time
window, and a potential valley in the other highlighted time window. The
details for how these time windows were identified will be explained later in the
discussion of the peak and valley detection algorithm. The precise time of the
peak in the window and time of the valley in the other, that when paired will
produce the largest alternating stress intensity, is dependent on multiple
parameters. For example, the Total stress intensity range, the Primary plus
Secondary stress intensity range, and the temperature can all impact the selection.
Each of the twelve stress values and the temperature can vary out of phase
relative to the others, so it is necessary to test all combinations of time points to
determine the most conservative pair. The need for this step is recognized in
ASME Code NB-3216.2 which suggests that different points in time should be
tried in order to find the one which results in the largest value of alternating
stress intensity.

g 3-8 h

13307342
150 Peak 700

100 600

50 500

SX_n
SY_n

0 400 SZ_n

Temperature (°F)
0 500 1000 1500 2000 2500 3000 3500 SXY_n
Stress (ksi)

SX_p
SY_p
‐50 300 SZ_p
SXY_p
Temp, °F

‐100 200

‐150 100

Valley

‐200 0

Figure 3-6
Illustration of Peak and Valley Time Windows

The calculation of alternating stress intensity for this peak and valley pair, or the
potential stress cycle, is provided below.

3.2.1 Alternating Stress Intensity (Salt)

The stress differences are determined for all of the potential time pairs between
the peak and valley. If a peak window has n points and the valley window has m,
then n∙m different combinations will be evaluated. The pair producing the largest
alternating stress intensity (Salt), including the effects of Ke and the elastic
modulus ratio, is used for input to the fatigue curve. Salt is calculated as:

S p Ec
S alt = K e Equation 3-1
2 Ea

The terms for this equation are briefly described as follows, and the details of
how to compute each one are provided after that.

g 3-9 h

13307342
Sp = P+Q+F (total) stress intensity range
Ke = Simplified elastic-plastic penalty factor
Ec = Modulus of elasticity shown on applicable fatigue curve
Ea = Modulus of elasticity used in the analysis

Sp

Sp is calculated as follows. First, the stress differences between time point “i” in a
peak and time point “j” in a valley for each individual stress component are
determined. The stresses used here are the P+Q+F total stresses.

σ x,i σ y,i σ z,i σ xy,i σ yz,i σ xz,i

– σ x, j σ y, j σ z, j σ xy, j σ yz, j σ xz, j Equation 3-2

= σx' σy' σz' σ xy ' σ yz ' σ xz '

From the six differences, σ x ' , σ y ' , σ z ' , σ xy ' , σ yz ' , and σ xz ' , the principal
total stress ranges, σ 1 ' , σ 2 ' , and σ 3 ' are calculated, using the process described
in Section 2 for calculating principal stresses. Sp is the total stress intensity range
( σ SI ' ), based on those three principal stress ranges.

Ke

The Ke value is the simplified elastic-plastic penalty factor, which accounts for
the effects of plasticity, based on the results of the linear elastic stress analysis.

Unless specified otherwise, Ke is developed based on the primary plus secondary


stress intensity range (Sn) along with several other factors. This formulation from
Paragraph NB-3228.5 is as follows:

1 for Sn ≤ 3Sm

Equation 3-3
1 − n ⎡ Sn ⎤
1+ ⎢ − 1⎥ for 3Sm < Sn < 3mSm
Ke = ( n)(m − 1) ⎣ 3S m ⎦

1
for Sn ≥ 3mSm
n

g 3-10 h

13307342
As an alternative to using the simplified elastic-plastic penalty factor above,
ASME Section III NB-3228.4(a) through (c) includes a provision for performing
a full plastic (non-linear) fatigue analysis (“shakedown” analysis) using the actual
stress-strain curve of the material. Shakedown analyses are typically performed on
individual load pairs to determine the actual, alternating strain for input to
refined fatigue calculations. This has the effect of reducing the effective Ke value
of the load pair. If several load pairs are run with various primary plus secondary
stress ranges (Sn), then a correlation between the elastically-computed Sn and the
plastically-computed effective Ke may be developed. A conceptual example of this
correlation is shown in Table 3-1 and Figure 3-7.

Table 3-1
Example Correlation Table Ke vs. Sn

Sn/3Sm Ke
1.00 1.00
1.23 1.00
1.45 1.00
1.71 1.25
1.92 1.40
2.17 1.50

3.50

3.00

2.50
Ke

Full E-P Ke
2.00
Simplified E-P Ke

1.50

1.00
1.00 1.20 1.40 1.60 1.80 2.00 2.20 2.40

Sn/3Sm

Figure 3-7
Example Correlation Between Sn and Ke

g 3-11 h

13307342
If a Ke vs. Sn/3Sm correlation table is provided in the fatigue monitoring software
configuration for the location, Ke for use in the fatigue calculations is calculated
by this alternative methodology. Linear interpolation is used between Ke vs.
Sn/3Sm points.

m and n

The m and n values are configurable and shall be selected from Table NB-
3228.5(b)-1 based on the material type of the SCL.

Sn

Sn is the Primary plus Secondary (P+Q) stress intensity range and is calculated
the same way as Sp, except that P+Q stresses are used instead of Total (P+Q+F).

Sm

Sm is the Code allowable stress value, which is a function of the material type
and temperature. Sm values are configurable, with an example configuration table
shown in Table 3-2. Sm values between given metal temperatures are calculated
with linear interpolation.

Table 3-2
Example Temperature Dependent E and Sm Values

T, °F Ea, ksi Sm, ksi


-100 30,000 23.3
70 29,300 23.3
100 29,100 23.3
200 28,600 21.9
300 28,100 21.3
400 27,500 20.6
500 27,100 19.4
600 26,500 17.8

It is conservative to choose the lower Sm value at the maximum temperature


of the pair, however note 4 of ASME Code Figure NB-3222-1 does allow the
average Sm value at the high and low temperatures during the transient to be used
if no mechanically induced (or seismic) loads contribute to the secondary stress.
Mechanically induced or seismic loads are rare in the analysis of most real plant
data. Both options are supported as follows.

g 3-12 h

13307342
1
2
[
S m (Ti ) + S m (T j ) ]
if USE_AVG_SM = TRUE
Sm =

S m (max[Ti , T j ]) Otherwise Equation 3-4

where:

Ti = Metal temperature of the monitored location at peak time i


Tj = Metal temperature of the monitored location at valley time j

Ea

Ea is the “elastic modulus of the analysis”. The ratio of the elastic modulus from
the fatigue curve (Ec) divided by Ea corrects the calculated alternating stress
intensity in accordance with ASME Section III NB-3222.4(e)(4) to account for
the fact that the y axis of the ASME fatigue curve (Sa) is actually the alternating
strain multiplied by a constant (the arbitrary Ec value). In other words, dividing
the calculated Salt by Ea converts the alternating stress intensity to alternating
strain. Multiplying the resulting alternating strain value by Ec converts it back to
an alternating pseudo-stress intensity range, consistent with the fatigue curve.

Given that many design fatigue analyses utilize temperature-dependent material


properties, the ASME Code is unclear about what “E of the analysis” is intended
to mean. Given the non-prescriptive nature of the Code, it is conventional for
analyses to use either the E at the maximum temperature of the pair, which is
mathematically conservative, or E at the average temperature of the pair, which is
more reasonable in approximating the effective alternating strain.

In SBF monitoring, stress analyses supporting Green’s functions and unit loads
may compute stresses using a single elastic modulus at an instantaneous
temperature with the objective of computing a conservative or accurate
alternating stress intensity value, relative to a more refined analysis that utilizes
temperature-dependent properties. While one interpretation of the rule would be
to perform the E ratio correction with whatever “E of the analysis” was used for
the stress analysis, we will instead utilize a temperature-dependent lookup of the
Code E values to perform the correction. This is more consistent with the intent
of the Code rule, given that alternating strain will depend on the elastic modulus
at the actual temperature.

Ea values are configurable, with an example configuration table shown in


Table 3-2. Ea values between given metal temperatures are calculated with linear
interpolation.

An option to use either E at the maximum or average temperature is supported.

g 3-13 h

13307342
E (average[Ti , T j ])
if USE_AVG_EA = TRUE

Ea =
E (max[Ti , T j ]) Equation 3-5
Otherwise

where:

Ti = Metal temperature of the monitored location at peak time i


Tj = Metal temperature of the monitored location at valley time j

3.2.2 Fatigue Usage Factor

The cumulative effect of stress cycles is calculated using a linear damage


relationship (Miner’s Rule). That is:

k
1
U = U0 + ∑ ≤ 1.0 Equation 3-6
i =1 Ni

where:
U0 = An initial fatigue usage factor for the component
k = Number of stress cycles in the loading spectrum
(i.e., the number of peak/valley pairs)
Ni = Fatigue life at Salt,i, computed using applicable S-N
fatigue curve and logarithmic interpolation of the values
per Note 2 of ASME Code Table I-9.1
Ui = 1/ Ni = Fatigue usage for the ith stress cycle
Salt,i = Alternating stress intensity of stress cycle i
U = Cumulative usage factor

3.3 Peak and Valley Detection

Periods where the conditions are known to be extreme (i.e., local peaks and
valleys) must be identified before determining the stress cycles.

Peak and valley detection using a uniaxial stress is simply based on the change in
sign of the slope of the stress. With a multiaxial stress state a different strategy
for detecting peaks and valleys must be developed. We are not only interested in
stress inflections per se, but regions where stress intensity range maximizes
relative to other extreme conditions in time. In addition, we must first identify
stress peak and valley regions not limited to a single instant in time, in order to
maximize Salt by accounting for potential non-proportional loading and the time
phasing between peak and secondary stresses and the temperature.

g 3-14 h

13307342
In a monitoring system, any given stress peak could pair with many different
stress valleys over the operating life of the plant. Therefore, a stress extremum (or
PV, for peak/valley) is defined as a period of time over which the total stress is at
or near a relative maximum (or minimum) value. Examples of these time
windows are illustrated in Figure 3-6. The extent of the PV will be defined to
assure that, for any two points in the PV, Salt is less than the endurance limit of
the fatigue curve. For example, for the range of time defined by the yellow
window (~190 < time < ~220) in Figure 3-8, Salt between any two time points
within that range is less than the endurance limit.

Sx Sy Sz Sxy Syz Sxz


001i

125

100

75
Stress [ksi]

50

25

-25

125 150 175 200 225 250 275 300


Time [sec]

Figure 3-8
PV Time Window Example

3.3.1 “Rubberband” PV Detection

Extensive research and testing was conducted to select the optimum PV


detection methodology. The following algorithm, referred to as Rubberband, was
selected for its ability to correctly identify the significant PVs, while filtering out
most of the minor stress reversals that do not contribute to the fatigue.

g 3-15 h

13307342
The algorithm uses a “range” function between the current stress state σ(t) and
the stress state at the previously identified PV, σlast , or range[σ(t), σlast], to
identify stress points that maximize the extent of any stress cycles and to filter
out stress reversals that do not result in significant stress cycles. For the best
agreement with traditional design fatigue analyses, the stress intensity range
(as defined in ASME Section III NB-3216.2) was selected as the optimum range
function to use.

When a new PV is found, the newly identified point σ(t) becomes the new σlast
for the next iteration. The initial time point is selected as the first σlast.

The algorithm is defined, in pseudo-code, as follows:

Let:

S(t) = The (six-term) stress history at the location being monitored,


with t = the timestep in the stress history ∈ {0, tmax},
P0, P1, … Pi, … PN = the points of extreme stress,
S0, S1, … Si, … SN = the (six-term) stress values at P0, P1, … PN,
range(S(t), Si) = the stress intensity range between S(t) and Si,
δmin = the filter threshold to eliminate insignificant reversals
(2 × PCT × Se)
Se = The endurance limit of the fatigue curve (the lowest stress value for
the largest number of allowable cycles in the configured
fatigue curve)
PCT = Configurable filter percentage of the endurance limit
(0 < PCT < 1)

Initially, set P0 = S0 = S(0) = the first recorded stress value in the stress history,
and S1 = S(1) = the second recorded stress value in the stress history.

Let i = 1, and let δ = range(S0, S1)

Loop (t = 2, 3 … tmax-1, tmax) {


If ( range(S(t), Si-1) > δ and range(S(t), Si-1) ≥ range(S(t), Si)) Then
// S(t) replaces Si as the new ‘current’ extreme stress
Si = S(t)
δ = range(S(t), Si-1)
Else If ( range(S(t), Si) ≥ δmin ) Then
// S(t) is a significant stress reversal, establishing Si as a local PV
Pi = Si
expand(Pi)
i=i+1 // start looking for the next PV
Si = S(t)
δ = range(S(t), Si-1)
End If
} Next t

g 3-16 h

13307342
When this algorithm has run to completion over the full stress history, all of the
significant stress reversals in the history are identified as Pi points. The ending
value of i gives the (zero-based) number of such points. Each Pi point has been
expanded to a PV as discussed below.

The procedure expand( Pi ) identifies the range of time around Pi that will be
included in the corresponding PV. This essentially defines the time window
around the PV, which will be used for the subsequent stress cycle counting
process. This is done by scanning both backwards and forwards from Pi until
range(S(t), Si) > (δmin / 2). The PV is defined as all of the points inside of that
time range (i.e., the largest contiguous collection of points containing Pi such
that range(S(t), Si) ≤ (δmin / 2)).

3.4 Stress Cycle Pairing

In this step, the previously identified stress extrema time windows, or PVs, will be
paired to create stress cycles for the fatigue computation. This will be done in such
a way to conservatively superimpose stress cycles of various origins to create the
maximum alternating stress intensity, while taking into consideration the time
sequence of the underlying actual stress extrema.

Cycle-pairing is performed using an algorithm derived from the Rainflow Cycle


Counting Method, which is the industry standard for computing fatigue usage in
cases of uniaxial or proportional loading histories.

3.4.1 Background on Uniaxial Rainflow Method

The Simplified Rainflow Cycle Counting Method, documented in Section 5.4.5 of


ASTM E1049-85 [10] and used in ASME Section VIII Division 2 Annex 5.B
(non-mandatory guidance) [11], is a classical approach for computing stress
cycles using a uniaxial stress history.

The stress peaks are identified by imagining water that "drips" down a pagoda
roof, as shown in Figure 3-9. Here the stress history is rotated so that the time
axis is vertical, and the ordered stress history represents a number of roofs.

g 3-17 h

13307342
Figure 3-9
Rainflow Visualization Example

Figure 3-10 shows an example uniaxial stress history. When this stress history is
processed with the Rainflow logic, it is shown to be composed of one large stress
cycle, and hundreds of much smaller stress cycles. Table 3-3 shows a spectrum of
the results. If the endurance limit of the fatigue curve were above 10 ksi, there
would only be one stress cycle (only one stress cycle with Salt above 10 ksi). The
peak and valley of this one stress cycle is highlighted on the figure.

The nature of this loading demonstrates the importance of order-dependence.


If each peak were paired with a respective valley, without consideration of order,
there would be many stress cycles on the order of the maximum range of stress.

g 3-18 h

13307342
10

0
0 500 1000 1500 2000 2500

‐5

‐10
Stress (ksi)

‐15

‐20

‐25

‐30

‐35

‐40

Figure 3-10
Example Uniaxial Stress History for Rainflow Analysis

Table 3-3
Example Uniaxial Stress History Spectrum Summary

Screen Salt_upper n
99 -100% 22.0159 1
97 - 99% 21.7957 0
94 - 97% 21.3554 0
90 - 94% 20.6949 0
85 - 90% 19.8143 0
79 - 85% 18.7135 0
72 - 79% 17.3926 0
64 - 72% 15.8514 0
55 - 64% 14.0902 0
45 - 55% 12.1087 0
34 - 45% 9.90716 5
22 - 34% 7.48541 10
9 - 22% 4.8435 33
4 - 9% 1.98143 57
0 - 4% 0.880636 312

g 3-19 h

13307342
3.4.2 Adaptation of Rain Flow to Multiaxial Stresses –
“Rainflow-3D”

The uniaxial rain flow methodology has been adapted to use a multiaxial stress
history, called Rainflow-3D, based on consideration of the ASME Code
Subarticle NB-3200 alternating stress intensity value, Salt, between potential
peaks and valley time windows (PVs).

As discussed previously, PVs are detected using the Rubberband methodology


developed herein. Each PV consists of a time window of at least one but typically
more than one time point of data, to allow determination of the maximum Salt
between PVs, based on consideration of pertinent parameters, such as Sp, Ke, and
the elastic modulus ratio, Ec/Ea.

The Rainflow-3D algorithm is fundamentally similar to Rainflow in determining


stress cycles, except that PVs consist of time windows instead of single points,
and Salt is used in determining ranges, instead of the simple difference in single
stress values between two points. The actual algorithm is as follows.

Let:
P0, P1, … Pi, … PN = the regions of extreme stress (i.e., stress extrema,
or PVs),
range(Pi, Pj) = the maximum stress difference between any two stress
values Si
and

Sj, with Si in Pi and Sj in Pj.


Pair(k) = an array of PV pairs { Pi, Pj } to record the cycle pairing
results, k = 0, kmax.
1. Start with the Pair() array empty (i.e., k = 0). Find the maximum value of
range(Pm, Pn), and the associated m and n values (m < n).
2. Re-sequence the Pi points so that Pm is first in the sequence; note that care
must be taken with the P0 and PN values to make sure the new sequence
alternates correctly, i.e.:

If (P0 is between PN and P1), then delete P0, and


If (PN is between PN-1 and P1), then delete PN
Else if (PN is between PN-1 and P0), then delete PN, and
If (P0 is between PN-1 and P1), then delete PN.
3. Read the next point from Pi. If there are no Pi points left to read, go to
Step 7.
4. If there are < 3 points to consider, go to Step 3; otherwise, call the last 3
points p0, p1 and p2. (e.g., the first three points will be p0=Pm, p1=Pm+1,
p2=Pm+2).

g 3-20 h

13307342
5. Calculate the ranges X = range(p1, p2), Y = range(p0, p1).
a. If (X < Y), go to Step 3
b. If (X ≥ Y), go to Step 6
6. Count Y as a cycle: Pair(k) = { p0, p1 }; k = k + 1,
Discard points p0 and p1 from the collection,
Go to Step 4.
7. If there are an odd number of points left, discard the last point from the
collection.
8. If there are < 2 points left, go to Step 9.
Otherwise, repeatedly count and remove the last two points as a cycle:
Pair(k) = { p1, p2 }; k = k + 1,
Discard points p1 and p2 from the collection
Go to Step 8
9. DONE; Number of cycle pairs (kmax) = k.

3.5 Example Calculations

Technical prototype software was developed to benchmark and validate the


algorithms. The technical prototype calculates fatigue usage using the following
inputs:
Primary plus Secondary (P+Q) and Total (P+Q+F) stresses (all six
components) and metal surface temperature as a function of time.
Material properties, as required in this report, including fatigue parameters
m and n, E of the fatigue curve Ec, the S-N fatigue curve, and temperature-
dependent values of the material strength (Sm) and elastic modulus of the
analysis (Ea).

The software starts by reading the input above, performs PV detection using the
Rubberband method, performs stress cycle counting using the Rainflow-3D
algorithm, and then computes the fatigue usage factor, outputting pertinent
parameters in a summary report.

This next sections attempt to validate the chosen methodology by applying it to


several theoretical and empirical cases. These examples demonstrate that the
proposed methodology produces excellent results for a wide variety of typical
(and unusual) cases.

3.5.1 Paperclip Example

The “paperclip” example described previously was analyzed using the prototype
software. A single stress component was used to simulate the stress (all other
components were held constant at 0), and the Primary plus Secondary stress was
assumed to be the same as the Total. The ASME Code stainless steel fatigue
curve and associated material parameters were used to perform the fatigue
evaluation.

g 3-21 h

13307342
Figure 3-11 shows the simulated stress input, and Table 3-4 shows the
fatigue analysis results. The results are in agreement with the expected results.
Specifically, there is one large stress cycle with significant plasticity (Ke = 3.333)
and 18 smaller stress cycles in the elastic regime. The CUF of 0.018 matches the
hand calculation determined previously to 3 decimal places.

200

150

100
Stress (ksi)

50

0
0 5 10 15 20 25 30 35 40 45

-50

-100

Figure 3-11
“Paperclip” Simulated Stress History

 3-22 

13307342
Table 3-4
“Paperclip” Fatigue Analysis Results

*** Fatigue Output Report


method = Rainflow-3D

row ___i__ ___j__ _nCyc ____Sp___ ____Sn___ __Ke__ ___Salt__ ____Na___ ____Ui___ LD:stp@time LD:stp@time
1 2: 0 3: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 1 @ 1.00, 1: 2 @ 2.00)
2 4: 0 5: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 3 @ 3.00, 1: 4 @ 4.00)
3 6: 0 7: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 5 @ 5.00, 1: 6 @ 6.00)
4 8: 0 9: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 7 @ 7.00, 1: 8 @ 8.00)
5 10: 0 11: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 9 @ 9.00, 1: 10 @ 10.00)
6 12: 0 13: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 11 @ 11.00, 1: 12 @ 12.00)
7 14: 0 15: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 13 @ 13.00, 1: 14 @ 14.00)
8 16: 0 17: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 15 @ 15.00, 1: 16 @ 16.00)
9 18: 0 19: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 17 @ 17.00, 1: 18 @ 18.00)
10 21: 0 22: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 22 @ 22.00, 1: 23 @ 23.00)
11 23: 0 24: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 24 @ 24.00, 1: 25 @ 25.00)
12 25: 0 26: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 26 @ 26.00, 1: 27 @ 27.00)
13 27: 0 28: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 28 @ 28.00, 1: 29 @ 29.00)
14 29: 0 30: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 30 @ 30.00, 1: 31 @ 31.00)
15 31: 0 32: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 32 @ 32.00, 1: 33 @ 33.00)
16 33: 0 34: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 34 @ 34.00, 1: 35 @ 35.00)
17 35: 0 36: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 36 @ 36.00, 1: 37 @ 37.00)
18 37: 0 38: 0 1 50.00000 50.00000 1.000 25.00000 1481072. 6.752e-07 ( 1: 38 @ 38.00, 1: 39 @ 39.00)
19 20: 0 39: 0 1 200.0000 200.0000 3.333 333.3333 54.46088 0.018362 ( 1: 19 @ 19.00, 1: 40 @ 40.00)
--- ------ ------ ----- --------- --------- ------ --------- --------- --------- ----------- -----------
19 CUF = 0.018374

3.5.2 EAF Expert Panel Sample Problem

A sample problem devised and previously evaluated by EPRI’s Environmental


Fatigue Expert Panel was analyzed using the prototype software. The official
stresses calculated by the Expert Panel were input directly into the test program
for consistency. These represent SCL1 using the “fast transient”, TRAN1.

Figure 3-12 shows the peak and valley regions identified for this transient (the
peak is also shown in detail in Figure 3-8). Table 3-5 shows the fatigue analysis
results. The results are identical to those evaluated using SI’s independent and
nuclear QA verified software [12] that computes fatigue usage factors in
accordance with the ASME Section III Subarticle NB-3200 design
requirements.

g 3-23 h

13307342
Sx Sy Sz Sxy Syz Sxz
001i

100

50
Stress [ksi]

-50

-100

001j
-150

0 500 1000 1500 2000 2500 3000


Time [sec]

Figure 3-12
Peak and Valley for EPRI EAF Expert Panel Sample Problem, SCL1/TRAN1

Table 3-5
Fatigue Analysis Results for EPRI EAF Expert Panel Sample Problem, SCL1/TRAN1

*** Fatigue Output Report


method = Rainflow-3D

row ___i__ ___j__ _nCyc ____Sp___ ____Sn___ __Ke__ ___Salt__ ____Na___ ____Ui___ LD:stp@time LD:stp@time
1 2: 3 3: 3 1 280.6228 118.2572 3.333 518.9791 19.42923 0.051469 ( 1: 84 @ 200.50, 1:177 @ 1195.50)
--- ------ ------ ----- --------- --------- ------ --------- --------- --------- ----------- -----------
1 CUF = 0.051469

3.5.3 Combined Bending and Shear

Combined bending and shear were analyzed. The stress history is shown in
Figure 3-13. The fatigue analysis results are shown in Table 3-6. The stress
cycles were correctly identified, along with correct calculation of alternating stress
intensity between PVs.

g 3-24 h

13307342
Sx Sy Sz Sxy Syz Sxz
001i 002i 003i 004i
30

25

20

15

10

5
Stress [ksi]

-5

-10

-15

-20

-25
001j 002j 003j 004j
-30

0 250 500 750 1000 1250 1500


Time [sec]

Figure 3-13
Combined Bending and Shear Stress History

Table 3-6
Fatigue Analysis Results for Combined Bending and Shear Stress

*** Fatigue Output Report


method = Rainflow-3D
row ___i__ ___j__ _nCyc ____Sp___ ____Sn___ __Ke__ ___Salt__ ____Na___ ____Ui___ LD:stp@time LD:stp@time
1 2: 37 3: 37 1 70.72614 0.000000 1.000 35.36307 43603.59 2.293e-05 ( 1: 90 @ 90.00, 1:270 @ 270.00)
2 4: 37 5: 37 1 70.72614 0.000000 1.000 35.36307 43603.59 2.293e-05 ( 1:450 @ 450.00, 1:630 @ 630.00)
3 6: 37 7: 37 1 70.72614 0.000000 1.000 35.36307 43603.59 2.293e-05 ( 1:810 @ 810.00, 1:990 @ 990.00)
4 8: 37 9: 37 1 70.72614 0.000000 1.000 35.36307 43603.59 2.293e-05 ( 1:1170 @ 1170.00, 1:1350 @
1350.00)
--- ------ ------ ----- --------- --------- ------ --------- --------- --------- ----------- -----------
4 CUF = 9.174e-05

3.5.4 Constant Principal Stresses and Stress Intensity

A stress history was developed where two normal stress components were equal
and opposite, varying sinusoidally over time. A shear stress was also varied
sinusoidally, but was 180° out of phase. Figure 3-14 shows stress history. The
odd characteristic of this transient is that it produces a stress history with
constant stress intensity and constant principal stresses. Essentially, while the
principal stresses are constant in magnitude, the principal directions are
constantly rotating. However, over time the stress intensity range does increase,
and a component if loaded in this manner would accumulate fatigue damage.

g 3-25 h

13307342
One peak and valley detection algorithm that was considered during this
development was to detect the change in sign of the slope of any of the principal
stresses and stress intensity as a way of determining extreme stress states
(we called this PV detection method InflectPS). While InflectPS worked
reasonably well in some situations, we discovered multiple disadvantages to the
methodology, and clearly in this case it does not work.

The Rubberband and Rainflow-3D methodology did detect one stress cycle and
computed the stress range correctly, as shown in Table 3-7. While it would be
very unusual if not impossible for a traditional pressure vessel or piping system
to experience loading like this, the test demonstrates the robustness of the
algorithm.

Sx Sy Sz Sxy Syz Sxz

100

75

50

25
Stress [ksi]

001j
0

-25

-50

-75

-100

0 50 100 150 200 250 300 350 400 450


Time [sec]

Figure 3-14
Constant Principal Stress and Stress Intensity History

Table 3-7
Fatigue Analysis Results for Constant Principal Stress and Stress Intensity

*** Fatigue Output Report


method = Rainflow-3D

row ___i__ ___j__ _nCyc ____Sp___ ____Sn___ __Ke__ ___Salt__ ____Na___ ____Ui___ LD:stp@time LD:stp@time
1 1: 0 2: 22 1 400.0000 0.000000 1.000 200.0000 221.5823 0.004513 ( 1: 0 @ 0.00, 1:180 @ 180.00)
--- ------ ------ ----- --------- --------- ------ --------- --------- --------- ----------- -----------
1 CUF = 0.004513

g 3-26 h

13307342
3.5.5 Evaluation of Real Data

Appendix A documents the entire process of applying this methodology


to an actual component in an operating nuclear plant. This example includes
developing unit stresses and Green’s functions, performing stress calculations
using actual plant operating data, processing those stresses through the
Rubberband peak and valley detection algorithm, performing Rainflow-3D
stress cycle counting, and finally computing a fatigue usage factor using the
methodology in this report.

g 3-27 h

13307342
13307342
Section 4: Environmentally-Assisted
Fatigue Usage Calculations
4.1 Background on EAF Calculations

Fatigue monitoring software SBF implements the Fen approach for computing
EAF. While Fen formulations vary among different materials and various
regulatory guidance documents, the basic methodology is essentially the same
for each case.

NUREG/CR-6909 defines the Fen fatigue correction factor as the ratio of fatigue
life in air at room temperature (Nair,RT = allowable number of stress cycles in air)
to that in water at the service temperature (Nwater = allowable number of stress
cycles in the reactor water environment). Since fatigue usage for one stress
cycle is the inverse of its allowable number of cycles (1/N), the Fen factor is an
adjustment to the ASME Code calculated usage using an air fatigue curve,
as follows.

Fen = N air , RT / N water

Based on a large body of fatigue tests performed under various environmental


conditions, statistical modeling has demonstrated that Fen factors are a function
of a few specific material and environmental parameters. These parameters are:
Material type,
Sulfur content,
Strain rate (on the increasing tensile side of the stress cycle),
Bulk dissolved oxygen content, and
Water temperature.

Some guidance documents also define a strain amplitude threshold; if the strain
amplitude for a stress cycle is below the threshold, then Fen = 1.0 (i.e., no
correction factor applies).

Environmental effects are incorporated into NB-3200 fatigue analyses by


multiplying the partial ASME usage factors for each stress cycle by an
appropriate Fen correction factor. For example, given n different stress
cycles, the cumulative environmental fatigue usage is:

g 4-1 h

13307342
U en = U 1 ⋅ Fen ,1 + U 2 ⋅ Fen , 2 + U 3 ⋅ Fen ,3 + U i ⋅ Fen ,i ... + U n ⋅ Fen , n [5, Eq. A.20] Equation 4-1

where:

Ui = computed fatigue usage using the air fatigue curve for the “i’th”
stress cycle
Fen,i = computed Fen for the “i’th” stress cycle

In the most rigorous analysis, the usage contribution from each stress cycle is
multiplied by the computed Fen value to account for the reduction in life due to
the reactor water environment. In SBF monitoring, a unique Fen value can be
applied to each individual stress cycle identified.

4.2 Regulatory Bases

For existing U.S. nuclear power plants operating within their original, 40 year
licensing basis, there is no back fit NRC requirement to evaluate and monitor
EAF. Plants need only meet their design basis requirements.

U.S. nuclear plants seeking license renewal to 60 or 80 years do require EAF


evaluations of Class 1 pressure retaining components, as do new nuclear power
plants subject to Reg. Guide 1.207.

Guidance for computing environmentally-assisted fatigue are promulgated in


different sets of documents. The applicable guidance document depends on the
licensing basis. Depending on the situation, multiple options are available. In
addition, regulatory documents on the topic of EAF are expected to revise some
of the existing Fen formulations in the future.

For these reasons, it is important for the fatigue monitoring software SBF
module to incorporate and continue to support Fen formulations from multiple
guidance documents. These various formulations are described in the sections to
follow.

4.2.1 Requirements for License Renewal

Part 54 to Title 10 of the U.S. Code of Federal Regulations (10CFR54) specifies


the “Requirements for Renewal of Operating Licenses for Nuclear Power
Plants”.

NRC report NUREG-1801, Revision 2, the “Generic Aging Lessons Learned


(GALL) Report”, identifies acceptable aging management programs for fatigue
and cyclic operation.

Plants have the options of computing EAF in accordance with guidance from
either an older set of NUREGs (NUREG/CR-5704 for austenitic stainless steel
materials and NUREG/CR-6583 for carbon and low alloy steels), or the newer
NUREG/CR-6909, which is applicable to all materials.

g 4-2 h

13307342
For nickel alloy materials, the formulations from NUREG/CR-6909 must be
used.

In the case of the older NUREGs, 5704 and 6583, the ASME Code fatigue
curve is used, and the Fen factors are applied on the ASME Code fatigue usage
values.

For the newer NUREG/CR-6909 formulations, new fatigue curves were


developed, in part, to back out any existing factors in the ASME curves that
account for industrial environment, but adjustments were also made to reflect
newer fatigue test data. When using NUREG/CR-6909 rules, the new fatigue
curve must be used for stainless steel materials, as the high cycle portion
eliminated a potential unconservatism in the ASME Code fatigue curve. For
ferritic materials, it is optional to use the NUREG/CR-6909 fatigue curve
(the ASME curve is more conservative).

4.2.2 Requirements for New Plants

For new plants, EAF requirements and Fen formulations are contained in
Regulatory Guide 1.207 [13]. The technical basis for this Regulatory Guide is
NUREG/CR-6909. Thus, the NUREG/CR-6909 rules implemented here can
meet the regulatory EAF requirements of both license renewal and new plants.

4.3 Fen Formulations for Ferritic Materials

4.3.1 NUREG/CR-6583 (Old Rules)

The following are the appropriate Fen relationships from NUREG/CR-6583 [14]
for carbon and low alloy steels. These expressions are:

For Carbon Steel [14, p. 69]:

Fen = exp(0.585 − 0.00124T '−0.101S *T * O *ε& * ) = exp(0.554 − 0.101S *T *O *ε& * ) Equation 4-2

For Low Alloy Steel [14, p. 69]:

Fen = exp(0.929 − 0.00124T '−0.101S *T *O *ε& * ) = exp(0.898 − 0.101S *T *O *ε& * ) Equation 4-3

Note that the above expressions have been corrected as summarized in Reference [16].

where:

Fen = fatigue life correction factor


T' = 25°C (NUREG/CR-6583, Section 6, Fen relative to
room temperature air)
S* = S for 0 < S ≤ 0.015 wt. %

g 4-3 h

13307342
= 0.015 for S > 0.015 wt. %
S = weight percent sulfur of steel
T* = 0 for T < 150°C
= (T - 150) for 150 ≤ T ≤ 350°C
T = service temperature (°C)
O* = 0 for DO < 0.05 parts per million (ppm)
= ln(DO/0.04) for 0.05 ppm ≤ DO ≤ 0.5 ppm
= ln(12.5) for DO > 0.5 ppm
DO = dissolved oxygen
ε& * = 0 for ε& > 1%/sec
= ln( ε& ) for 0.001 ≤ ε& ≤ 1%/sec
= ln(0.001) for ε& < 0.001%/sec
ε& = strain rate, %/sec

4.3.2 NUREG/CR-6909 (New Rules)

For Carbon Steel (CS) [5, p. A.1]:

Fen = exp(0.632 − 0.101S *T *O *ε& * ) Equation 4-4

For Low Alloy Steel (LAS) [5, p. A.1]:

Fen = exp(0.702 − 0.101S *T *O *ε& * ) Equation 4-5

Where S , T * , O , and ε& are the transformed sulfur content, service


* * *

temperature, dissolved oxygen (DO), and strain rate, respectively, which are
defined as follows [5, p. A.1 and A.2]:

Fen = fatigue life correction factor


S = weight percent sulfur of steel
S* = 0.001 for S ≤ 0.001 wt.%
= S for 0.001 < S ≤ 0.015 wt. %
= 0.015 for S > 0.015 wt. %
T* = 0 for T < 150°C
= (T - 150) for 150 ≤ T ≤ 350°C
T = service temperature (°C)
O* = 0 for dissolved oxygen, DO ≤ 0.04 parts per
million (ppm)
= ln(DO/0.04) for 0.04 ppm < DO ≤ 0.5 ppm
= ln(12.5) for DO > 0.5 ppm
ε& * = 0 for strain rate, ε& > 1%/sec

g 4-4 h

13307342
= ln( ε& ) for 0.001 ≤ ε& ≤ 1%/sec
= ln(0.001) for ε& < 0.001%/sec

For both carbon and low-alloy steels, a threshold value of 0.07% for strain
amplitude (one-half the strain range for the cycle) is defined, below which
environmental effects on the fatigue life of these steels do not occur. This strain
threshold corresponds to 21 ksi (145 MPa) alternating stress intensity from the
fatigue analysis. That is, if σalt ≤ 21 ksi then Fen = 1.0.

Fen = 1 for strain amplitude, ε a ≤ 0.07% or Salt ≤ (Ec)(0.07%)/(100%) = 21 ksi


(145 MPa)

4.4 Fen Formulations for Austenitic Stainless Steel Materials

4.4.1 NUREG/CR-5704 (Old Rules)

For Wrought and Cast Austenitic Stainless Steel [17]:

Fen = exp( 0.935 − T *ε& *O * ) Equation 4-6

where:

Fen = fatigue life correction factor


T = service temperature of transient, °C
T* = 0 for T < 200°C
= 1 for T ≥ 200°C
ε& * = 0 for strain rate, ε& > 0.4%/sec
= ln( ε& /0.4) for 0.0004 ≤ ε& ≤ 0.4%/sec
= ln(0.0004/0.4) for ε& < 0.0004%/sec
O* = 0.260 for dissolved oxygen, DO < 0.05 parts per
million (ppm)
= 0.172 for DO ≥ 0.05 ppm

4.4.2 NUREG/CR-6909 (New Rules)

For wrought and cast austenitic stainless steels (SS) [5, p. A.2]:

Fen = exp(0.734 − T ' O' ε& ' ) Equation 4-7

where:

Fen = fatigue life correction factor


T' = 0 for T < 150°C
= (T - 150)/175 for 150 ≤ T < 325°C
= 1 for T ≥ 325°C
T = service temperature (°C)
ε& ' = 0 for strain rate, ε& > 0.4%/sec

g 4-5 h

13307342
= ln( ε& /0.4) for 0.0004 ≤ ε& ≤ 0.4%/sec
= ln(0.0004/0.4) for ε& < 0.0004%/sec
O' = 0.281 for all dissolved oxygen levels

For wrought and cast austenitic stainless steels, a threshold value of 0.10% for
strain amplitude (one-half the strain range for the cycle) is defined, below which
environmental effects on the fatigue life of these steels do not occur. This strain
threshold corresponds to 28.3 ksi (195 MPa) alternating stress intensity from the
fatigue analysis. That is, if Salt ≤ 28.3 ksi then Fen = 1.0. Thus,

Fen = 1 for strain amplitude, ε a ≤ 0.10% or Salt ≤ (Ec)(0.10%)/(100%) = 28.3 ksi


(195 MPa)

4.5 Fen Formulations for Nickel Alloy

4.5.1 NUREG/CR-6909 (New Rules)

New rules are required for Nickel Alloy materials per both the GALL report
(license renewal) and Reg. Guide 1.207 (new plants).

For Ni-Cr-Fe alloys [5, p. A.2]:

Fen = exp(−T ' ε& ' O' ) Equation 4-8

where:

Fen = fatigue life correction factor


T' = T/325 for T < 325°C
= 1 for T ≥ 325°C
T = service temperature (°C)
ε& ' = 0 for strain rate, ε& > 5.0%/sec
= ln( ε& /5.0) for 0.0004 ≤ ε& ≤ 5.0%/sec
= ln(0.0004/5.0) for ε& < 0.0004%/sec
O' = 0.09 for NWC BWR water
= 0.16 for PWR or HWC BWR water

For Ni-Cr-Fe alloys, a threshold value of 0.10% for strain amplitude (one-half
the strain range for the cycle) is defined, below which environmental effects on
the fatigue life of these steels do not occur. This strain threshold corresponds to
28.3 ksi (195 MPa) alternating stress intensity from the fatigue analysis. That is,
if Salt ≤ 28.3 ksi then Fen = 1.0. Thus,

Fen = 1 for strain amplitude, ε a ≤ 0.10% or Salt ≤ (Ec)(0.10%)/(100%) = 28.3 ksi


(195 MPa)

g 4-6 h

13307342
4.6 How to Compute Fen for a Stress Cycle

In each of the Fen formulations documented previously, the basic set of inputs is
the same. It is therefore desirable to develop a common process to evaluate the
Fen for each cycle, while using the equation that is applicable to the specific
material and regulatory basis.

Discussion about some of the common parameters and how they are calculated is
provided below.

4.6.1 Strain Rate Calculations

Fen factors only apply when the stress is increasingly tensile in nature. In each of
the Fen formulations, strain rate, ε& , is defined as a single, scalar value. Guidance
from MRP-47, Rev. 1 [15] advises using differences in the scalar, instantaneous
stress intensity values of the current and previous time steps to calculate strain
rate, and using the sign of a single, dominant stress component, determined a
priori, to evaluate whether it is increasingly tensile or compressive in nature.

In reality, the strain increment is not always based on the difference in stress
intensity values, specifically with a multiaxial stress state. Both increasing
compression and increasing tension can also simultaneously occur along different
axes.

Based on recent discussions in EPRI’s Environmental Fatigue Expert Panel,


there is general consensus on an improved solution to the calculation of strain
rate that accounts for the multiaxial stress state and automatically determines
whether the strain increment is increasingly tensile or compressive in nature,
based on the sign of the dominant principal of the stress differences. This
methodology is outlined in detail in the sections that follow.

A second issue is that MRP-47 currently advises to not include the simplified
elastic-plastic penalty factor Ke in the strain rate calculations. The reason for this
is that the ASME Ke formula is conservative, but when used in strain rate
calculations may yield an unconservative Fen factor (a slower strain rate produces
a higher Fen).

EPRI’s EAF Expert Panel has discussed this issue in detail as well. There is
general agreement that inclusion of Ke in the strain rate is conservative, so long as
it is not greater than the Ke used in the CUF calculations. The reason for this is
that due to the nature of the mathematics, Ke in the CUF calculations always
overwhelm the effect of Ke in the strain rate and Fen calculations. Figure 4-1
demonstrates this trend with a parametric analysis of Uen (CUFen). In general, it is
always conservative to amplify the stress amplitude in the CUF calculations,
even when that stress amplitude increases the strain rate accordingly in the Fen
calculations. The inclusion of Ke in the strain rate calculation is shown in detail in
the sections that follow.

g 4-7 h

13307342
Figure 4-1
Parametric Analysis Usinng Ke in CUF and Strain Rat
ate

4.6.2 Av
verage vers
sus Maxim
mum Servicce Tempera
rature

NUREG/C CR-6583 and d NUREG/CR R-5704 requiire use of the maximum service
temperature of the load set pair in com
mputing Fen.

NUREG/C CR-6909 in several section


ns of the reporrt specificallyy allows for usse of
average temmperature in the ns. However, the intent off this allowancce is
t calculation
clarified on
n page A.5:

In general, the “aaverage” temperature that shhould be used in i the calculatiions


shoould produce reesults that are consistent witth the results thhat would be
obttained using thhe modified raate approach… The maximuum temperaturre can
be used to performm the most connservative evaluation.

EPRI’s EA AF Expert Pan nel has discusssed this issuee as well and verified
v that
with a relattively simple sample
s probleem, use of aveerage temperaature can be leess
conservativve than use off the modifiedd rate method dology. Since modified ratee
(see below) is used in thee SBF calculaations this issu ue is of little importance,
i b
but
for conservatism and to maintain con nsistency of methodology
m b
between the

g 4-8 h

13307342
various Fen formulations available, the maximum service temperature will be
used. For the SBF calculations, the “service temperature” is taken as the metal
temperature of the wetted surface of monitored location. This surface metal
temperature is computed using Green’s functions, as described in earlier sections.

4.6.3 Dissolved Oxygen

Time-based dissolved oxygen (DO) values will be used in Fen modified rate
calculations. DO values may be provided by either configured, bounding DO
values, or direct instrumentation, or a combination of both.

Typically, these time histories provide bounding DO values based on time


intervals with different water chemistry programs. For example, in BWRs plants
switched between modes of normal water chemistry, hydrogen water chemistry,
noble metal addition, etc, all of which can significantly impact the DO levels. An
example bounding DO history is shown in Table 4-1.

Table 4-1
Example Dissolved Oxygen History
Begin Value Based On DO (ppb)
1/1/1970 00:00:00 VALUE 120
3/1/1985 12:00:00 VALUE 80
12/1/2011 03:00:00 INSTRUMENT DO_RPV_UPPER

For the modified rate Fen integrations described below, DO values will be
selected based on this configuration information provided for each SBF location
that computes EAF.

4.6.4 Calculation of Fen Using the Modified Rate Approach

The calculation of Fen here is consistent with the modified rate approach,
described in NUREG/CR-6909, and general consensus among EPRI’s EAF
Expert Panel recommendations.

The Fen for a stress cycle is computed as:


n

∑F en ,i Δε i
Fen = i =2
n
Equation 4-9
∑ Δε
i=2
i

where:

Fen computed at Point i = fn ( ε&i , T ' i or T i , O' i or


*
Fen ,i =
O * i , S ' i or S * i ). This function is specific to the
applicable regulatory document and the material of
interest.
g 4-9 h

13307342
n = Number of time points in the stress cycle.
i = The collection of points (n) for the stress cycle,
where the time increases with increasing i.
T ' i or T * i = fn(T), where T = maximum metal temperature
of the current (i) and previous (i-1) time steps.
This function is specific to the applicable
regulatory document and the material of
interest.
O'i or O * i = fn(DO), where DO is the dissolved oxygen level.
This function and the DO value to use are
specific to the applicable regulatory document
material of interest.
S ' i or S * i = fn(S), where S is the sulfur content of the
material. This function and the S value to use
are specific to the applicable regulatory
document and the material of interest.
ε&i = Strain rate at Point i, %/sec
= Δε i / Δt ∙100%
Δt = Change in time at Point i, sec
= t i − t i −1
Δε i = Change in strain at Point i, in/in.

Δε i , based on all six components of the stress tensor, is computed by performing


the following procedure.

First, the principal stress ranges and stress intensity range from Point i-1 to
Point i are computed per NB-3216.2, which uses all six stress components and
considers the possibility of varying principal stress directions. Total P+Q+F
stresses are used in the calculation of strain rate.
σ x ,i σ y ,i σ z ,i σ xy ,i σ yz ,i σ xz ,i
– σ x ,i −1 σ y ,i −1 σ z ,i −1 σ xy ,i −1 σ yz ,i −1 σ xz ,i −1 Equation 4-10
= σx' σy' σz' σ xy ' σ yz ' σ xz '

From σ x ' , σ y ' , etc. the principal stress ranges ( σ 1 ' , σ 2 ' , σ 3 ' ) and stress
intensity range σ I ' are computed. Per convention, σ 1 ' , σ 2 ' , and σ 3 ' are
reordered so that σ 1 ' ≥ σ 2 ' ≥ σ 3 ' .

The sign of the principal total stress range from Point i-1 to i with the largest
absolute value determines whether the strain increment is primarily increasingly
tensile or increasingly compressive in nature.

The increasingly tensile change in strain, or the strain increment, Δε i , is


computed as follows.

g 4-10 h

13307342
σI'
Ke if Sgn = 1 and range(Si, Sv) > max_range
E (Tmin )

Δε i = Equation 4-11

0 Otherwise

where:

sgn(σ 1 ) if σ1' ≥ σ3'


Sgn =
sgn(σ 3 ) Otherwise

sgn(σ ) = The sign (-1 or +1) of σ.

Ke = Plastic penalty factor for the load pair, based on


previously performed NB-3228.5 Simplified
Elastic-Plastic Analysis. It is conservative to assume that
Ke = 1.0. Ke used in the strain rate calculation shall not be
greater than Ke used in the fatigue analysis for the stress
cycle under consideration.
E (Tmin ) = Elastic modulus at the minimum temperature, Tmin, of
the current (i) and previous (i-1) time points. As a
conservative alternative, the elastic modulus of the
fatigue curve may be used.
Si = Stress at (time) Point i.

Sv = Stress at the closest valley preceding Point i.

range(Si, Sv) = range function from Section 3.3.1 above.

max_range = the largest range range(Sj, Sv) for any stress value Sj
between Sv and Si-1.

The check against max_range is used because stresses computed using real data
can have a large number of small reversals that are not peaks and valleys (for
instance, see Figure 3-2). This condition prevents the method from integrating
over many small increments that do not increase the overall tensile strain range of
the stress cycle.

g 4-11 h

13307342
Note that if the stress cycle under consideration is composed of two extreme
stress states that are not chronologically continuous (e.g. the peak of one
transient pairs with the valley of a different one), then the strain increment
and strain rate within the discontinuity should not be considered in the Fen
formulation (i.e., Δε i =0).

4.6.5 Bounds of the Modified Rate Integration

For a stress cycle where the valley precedes the peak and where the paired loads
are chronologically contiguous with no intermediate (internal) peaks or valleys
between them, calculation of the Fen using the Modified Rate Approach is
straightforward. The integration is simply performed from the stress valley
to the stress peak.

When the peak and valley combination is not continuous, as shown in Figure
4-2, the integration is performed in two parts: from the valley in the stress cycle
to the next peak in the sequence, and (for the peak in the stress cycle) from the
valley preceding the peak up to that peak. This is illustrated in Figure 4-2 for the
largest (004i – 004j) stress range pairing.

Sx Sy Sz Sxy Syz Sxz


004j

75

50 002i
001i

25

0
001j
Stress [ksi]

-25 003j
002j

-50

-75

-100

-125
004i

0 500 1000 1500 2000 2500 3000 3500


Time [sec]

Figure 4-2
Complex Transient with Multiple Stress Cycles

g 4-12 h

13307342
4.7 Example Fen Calculations

The sample problem previously evaluated and shown in Figure 3-12 was
evaluated for EAF using the prototype software and the NUREG/CR-6909 Fen
formulations. In this case the peak precedes the valley. The integration is
performed discontinuously: from the beginning to the peak, and from the valley
to the end.

Figure 4-3 shows a time history of the Fen,i and the Fen,i times the strain
increment versus time. The resulting Fen for the stress cycle was calculated to be
5.14.

14 0.025

12

0.02

10

0.015
8

Fen,i∙Δε i
Fen,i

Fen

6 Fen∙de
0.01

0.005

0 0
0 500 1000 1500 2000 2500 3000 3500
Time (seconds)

Figure 4-3
EAF Expert Panel Sample Problem EAF Parameters

g 4-13 h

13307342
13307342
Section 5: Summary and Conclusions
This report provides the technical basis for a stress-based fatigue (SBF)
monitoring system. It can serve as a basis for software to perform SBF
computations, but also provides detailed discussions about the rationale for
various decisions that were made.

A primary reason for developing this new approach was to address the NRC’s
concerns, as documented in RIS 2008-30, about the use of a simplified, single
stress term in fatigue analyses. All six components of the stress tensor are used in
the fatigue calculations presented herein, and all of the fatigue calculation
requirements of the ASME Code, Subarticle NB-3200 are met.

A second important reason for this document was to update the SBF
methodology to include automatic calculation of environmentally-assisted
fatigue, in accordance with applicable regulatory guidance documents.

As this document serves as a requirements specification for the software itself,


detailed verification and benchmarking of the final fatigue monitoring software
software will be performed and documented separately. However, during this
research, technical prototype software was developed in order to benchmark and
validate the various methodologies, as described in this report.

Guiding principles for the development of this methodology included:


1. Accuracy. Benchmarking calculations reproduce known problems and meet
design basis (i.e., ASME Code) and regulatory requirements.
2. Validation. Results are demonstrated to make physical sense, and are
consistent with sound science and engineering principles.
3. Repeatability. While being flexible enough to meet different and unique
needs, a special attempt was made throughout the development process to
minimize or eliminate the need for engineering and user judgments, where
practical.

The methodology presented here brings together many proven practices.


However, it also includes three basic areas of new technology developed for this
project:

g 5-1 h

13307342
1. Optimized Multiaxial Stress Calculations. A multiaxial methodology is
supported that considers all six components of both Primary plus Secondary
and Total stresses, in accordance with ASME Code Subarticle NB-3200
fatigue analysis requirements. The through-wall thermal stress distribution is
determined using either a Lagrange Polynomial or Piece-Wise Linear
formulation, allowing detailed and automatic linearization of time-dependent
stresses. Various options are supported to adjust the secondary stresses in
accordance with Code rules (such as for piping components), and thereby
reduce conservatism. The Green’s function methodology for computing
thermal stresses was also optimized for speed and accuracy.
2. Smart Peak and Valley Detection and Stress Cycle Counting Algorithms. A
stress peak and valley detection algorithm called Rubberband was developed
to detect “time windows” containing the local stress extrema (the peaks and
valleys), considering the possibility of non-proportional loading and the time
phasing between the secondary and peak stresses and the temperature. A
stress cycle counting algorithm called Rainflow-3D was developed that takes
the order of stress perturbations into account. Rainflow-3D takes guidance
from the well known and vetted uniaxial Rainflow algorithm, but computes
ranges between stress extrema using the ASME Code requirements for
computing alternating stress intensity, using all six components of the stress
tensor.
3. Automatic Calculation of Environmentally-Assisted Fatigue. A
methodology was developed to automatically calculate environmentally-
assisted fatigue usage, using the SBF results. The methodology supports
multiple regulatory guidance documents, including NUREG/CR-5704,
NUREG/CR-6583, and NUREG/CR-6909 (Reg. Guide 1.207). The
methodology is based on the latest consensus opinions from EPRI’s
Environmental Fatigue Expert Panel.

Appendix A provides an example of using this methodology. It illustrates the


process used to configure a typical SBF location. This location is then analyzed
for a real transient, and the results are presented.

g 5-2 h

13307342
Section 6: References
6.1 References
1. Draft NRC RIS-2008-XX, Fatigue Analysis of Nuclear Power Plant
Components, May 2008.
2. Staff Responses to Public Comments on Proposed NRC Regulatory Issue
Summary 2008-XX “Fatigue Analysis of Nuclear Power Plant Components”
May 1, 2008 (73 FR 24094).
3. NRC RIS-2008-30, Fatigue Analysis of Nuclear Power Plant Components,
December 2008.
4. ASME Boiler and Pressure Vessel Code, Section III, Division 1, 2004
Edition.
5. NUREG/CR-6909 (ANL-06/08), Effects of LWR Coolant Environments on
the Fatigue Life of Reactor Materials Final Report, February 2007.
6. FatiguePro On-Line Fatigue Monitoring System: Demonstration at the Quad
Cities BWR, January 1989. EPRI Report No. NP-6170-M.
7. Kuo, A. Y., Tang, S. S., and Riccardella, P. C., An On-Line Fatigue
Monitoring System for Power Plants, Part I – Direct Calculation of Transient
Peak Stress Through Transfer Matrices and Green’s Functions, ASME PVP
Conference, Chicago, 1986.
8. Burgreen, D. (1971), Elements of Thermal Stress Analysis, First Edition,
Cherry Hill, NJ: Arcturus Publishers.
9. Rao, K. R., et al, Companion Guide to the ASME Boiler and Pressure Vessel
Code – Criteria and Commentary on Select Aspects of the Boiler & Pressure Vessel
and Piping Codes, Volume 1, the American Society of Mechanical Engineers,
2002.
10. ASTM E 1049-85, (Reapproved 2005), Standard practices for cycle counting in
fatigue analysis, ASTM International.
11. ASME Boiler and Pressure Vessel Code, Section VIII, Division 2, Annex
5.B (non-mandatory guidance), 2010 Edition.
12. VESLFAT, Version 1.42, 02/06/07, Structural Integrity Associates.

g 6-1 h

13307342
13. U.S. Nuclear Regulatory Commission, Regulatory Guide 1.207,
GUIDELINES FOR EVALUATING FATIGUE ANALYSES
INCORPORATING THE LIFE REDUCTION OF METAL
COMPONENTS DUE TO THE EFFECTS OF THE LIGHT-WATER
REACTOR ENVIRONMENT FOR NEW REACTORS, 2007.
14. NUREG/CR-6583 (ANL-97/18), Effects of LWR Coolant Environments on
Fatigue Design Curves of Carbon and Low-Alloy Steels, 1998.
15. Guidelines for Addressing Fatigue Environmental Effects in a License Renewal
Application (MRP-47 Revision 1). Palo Alto, CA: April 2005. EPRI
Technical Report TR-1012017.
16. EPRI/BWRVIP Memo. No. 2005-271, Potential Error in Existing Fatigue
Reactor Water Environmental Effects Analyses, July 1, 2005.
17. NUREG/CR-5704 (ANL-98/31), Effects of LWR Coolant Environments on
Fatigue Design Curves of Austenitic Stainless Steels, 1999.
18. NUREG-1801, Revision 2, Generic Aging Lessons Learned (GALL) Report,
U. S. Nuclear Regulatory Commission, December 2010.
19. NUREG/CR-6260 (INEL-95/0045), Application of NUREG/CR-5999
Interim Fatigue Curves to Selected Nuclear Power Plant Components, March
1995.
20. ANSYS Mechanical and PrepPost, Release 11.0 (w/Service Pack 1),
ANSYS, Inc., August 2007.
21. Stevens, G.L. (NRC), Rathbun, H.J. (NRC), Gilman, T.D. (SI),
Investigation Of Differences In The Finite Element Solution Of A Sample
Fatigue Cumulative Usage Factor Calculation Problem, Proceedings of the
ASME 2011 PVP Conference, July 17-21, 2011, Baltimore, Maryland.
PVP2011-57651.

g 6-2 h

13307342
Appendix A: SBF Example – PWR
Charging Nozzle
This appendix includes a high level description of the overall process of
configuring an SBF monitoring location and performing analysis of plant data.

A.1 Description of the Monitored Location

The example used here is a pressurized water reactor (PWR) charging


branch nozzle, located on the reactor coolant system main loop (RCL). This
NUREG/CR-6260 [19] location was determined to be limiting (the location
with highest fatigue usage) in the chemical and volume control system.

The approximately 30 inch diameter RCL, 3 inch nozzle forging and attached 3
inch charging piping are all composed of various types of austenitic stainless steel
material. A thermal sleeve is not attached to the nozzle, because of concerns
about flow-induced vibration.

The fatigue usage will be computed using the ASME Code NB-3200 rules,
except that the fatigue curve from NUREG/CR-6909 will be utilized. The EAF
will be computed using NUREG/CR-6909 Fen formulations. During normal
operation, the PWR primary system operates in a low oxygenated environment.

A.2 Finite Element Analysis

A finite element model (FEM) of the nozzle and attached piping is first
constructed. 3D elements are utilized, because of the non-axisymmetric nature of
the cylinder-to-cylinder intersection. A symmetric quarter model is shown in
Figure A-1.

Instantaneous material properties at 350°F, the approximate average temperature


of the most limiting transient, were utilized.

The SCL shown in Figure A-2 was shown to be a limiting location in a previous
fatigue analysis. It is selected for this example.

g A-1 h

13307342
Figure A-1
Charging Nozzle Finite Element Quarter Model

g A-2 h

13307342
Figure A-2
Charging Nozzle Monitored Location (the SCL)

Static loads needing to be considered are:


Internal pressure
Branch (charging) piping interface loads
Run (RCL) piping interface loads

A.2.1 Unit Internal Pressure Analysis

A unit internal pressure analysis is performed. Results may be scaled to any


internal pressure conditions, based on the linear relationship for an elastic
analysis.

In addition to application of internal pressure loading, cap loads are placed at the
longitudinal boundaries on the charging line branch piping and the cold leg run
piping to account for the axial stresses produced from the internal pressure. Cap
loads are computed by multiplying the unit pressure by the ratio of the cross
sectional area of the water (inside diameter) by the cross sectional area of the
metal.
g A-3 h

13307342
Symmetric displacement boundary conditions are applied, and the axial cross
sections are coupled as well, in order to simulate the presence of the adjacent
piping and prevent gross distortion of the cross section. The analysis boundary
conditions are shown in Figure A-3.

Figure A-3
Charging Nozzle Unit Pressure Analysis

A.2.2 Branch Piping Interface Loads

The branch piping interface loads are analyzed to account for thermal expansion
of the adjacent charging piping at the boundary of the model. A full 360 degree
(instead of a quarter) model is used in order to account for the asymmetric nature
of the loading, as shown in Figure A-4.

g A-4 h

13307342
Figure A-4
Charging Nozzle Finite Element 360° Model

Piping forces and moments due to thermal expansion are input from a previously
calculated piping analysis. The forces and moments are applied on the branch
piping using appropriate boundary conditions and finite element analysis
techniques. The calculated stresses are then scalable to the indicated temperature.

A.2.3 Run Piping Loads

The run piping interface loads are analyzed to account for thermal expansion of
the adjacent RCL piping at the boundary of the model. The loads are applied on
the run piping using the 360° model and appropriate boundary conditions and
finite element analysis techniques. The calculated stresses are then scalable to the
indicated temperature.

A.2.4 Time-Dependent Thermal (Green’s Function) Analyses

There are two separate convection surfaces on the model relevant to our analysis
of this nozzle. Each has flows and temperatures that can vary independently of
the other.

g A-5 h

13307342
One surface, labeled H1 on Figure A-5, is the inside of the charging piping and
nozzle. The other surface, labeled H2, is the inside of the cold leg (RCL) piping.

Figure A-5
Charging Nozzle Convection Surfaces

For this demonstration example, bounding heat transfer coefficients are


computed for each of the two surfaces, and two separate Green’s function
analyses are performed. Constant material properties are used, and the stress free
temperature is assumed, for analysis purposes, to be 0°F.

The analysis begins with the component at a uniform, steady state, and stress free
temperature of 0°F. For Load Case 1, the H1 surface is subjected to a 1000°F
step change in temperature, while H2 is maintained at 0°F, and the conditions
are run to steady state.

Load Case 2 is performed the same way, except that the H2 surface has an
applied 1000°F step change in temperature, while H1 is maintained at 0°F. The
run is continued until steady state conditions are achieved.

g A-6 h

13307342
Using the process described in Chapter 2, Green’s function files are generated
using an ANSYS macro, and a polynomial degree (p) of 5 (6 points through wall)
is specified.

The Green’s functions will both be unitized internally to the software, and
integrated with their respective, independently-varying temperatures. The total
thermal stresses and temperatures at the location of interest are determined using
superposition.

A.3 Development of the SBF Transfer Functions

Typically, instrumentation is not directly available at the monitored location for


direct input of all relevant temperatures, flow rates, etc. Transfer functions relate
the available plant instrumentation to the calculation of stresses and other
pertinent parameters at the monitored location and must be determined and
documented plant-specifically.

For this example, the following parameters are needed to compute stresses:
Local charging flow rate (H1)
Local charging temperature (H1)
Local cold leg flow rate (H2)
Local cold leg temperature (H2)
Internal pressure

In this case, the local cold leg flow rate, the cold leg temperature, and the internal
pressure may be determined adequately from direct instrumentation.

The local charging flow rate can be calculated, based on examination of the flow
diagrams and available instruments. Here, the flow rate through the nozzle is
ultimately calculated by using a number of different plant instruments, such as
charging header flow rate, reactor coolant pump seal flow rate, and various valve
positions.

The local charging temperature may be computed with the use of a thermal
hydraulic model, that uses an upstream temperature sensor and the calculated
flow rate to predict the local temperature at the nozzle. The thermal hydraulic
model would account for the fluid transit time and convective heat transfer
between the fluid, piping and atmosphere, as the fluid travels down the piping
system. When the charging flow stops, the effects of swirl penetration from the
RCL coolant into the nozzle (reflood) must be accounted for as well.

A.4 Example Analyses

A.4.1 Loss of Letdown with Delayed Return

Real plant data for a transient from a sample plant was used for this
demonstration. The calculated local temperatures are shown in Figure A-6.

g A-7 h

13307342
Stresses were computed using the unit loads, Green’s functions, and transfer
function logic. The stress history was then evaluated for PV detection, stress cycle
counting, and fatigue analysis. The computed stresses and stress cycle pairing are
shown in Figure A-7.

The fatigue usage results are shown in Table A-1.

For stress cycle #2 (the largest alternating stress intensity range), the Fen is
integrated from the valley 002i to peak 002j, because the stress history between
them is continuous with no intermediate peaks or valleys. For stress cycle #1, the
Fen is integrated from the beginning of the stress history to the peak 001i, and
from the valley 001j to the end of the stress history. The environmental fatigue
results are shown in Table A-2. The overall Fen is approximately 2.5.

600

500

400
Temperature (°F)

300
TCHG
TCOLD

200

100

0
27000 29000 31000 33000 35000 37000 39000 41000 43000 45000
Time (seconds)

Figure A-6
Loss of Letdown with Delayed Return to Service

g A-8 h

13307342
Sx Sy Sz Sxy Syz Sxz
150
002j

125

100

75
Stress [ksi]

001i

50

25 001j

002i

28000 29000 30000 31000 32000 33000 34000 35000 36000 37000 38000 39000
Time [sec]

Figure A-7
Stresses and Stress Pairs - Loss of Letdown with Delayed Return to Service

Table A-1
Fatigue Usage - Loss of Letdown with Delayed Return to Service

*** Fatigue Output Report


method = Rainflow-3D

row ___i__ ___j__ _nCyc ____SR___ ____Sn___ __Ke__ ____Sa___ ____Na___ ____Ui___ LD:stp@time LD:stp@time
1 2: 13 5: 16 1 36.93673 20.64353 1.000 20.36806 534270.1 1.872e-06 ( 1: 60 @ 28390.00, 1:485 @ 37774.00)
2 3: 31 4: 24 1 123.2131 75.05846 1.874 127.9009 644.2506 0.001552 ( 1:116 @ 28653.00, 1:238 @ 36739.00)
--- ------ ------ ----- --------- --------- ------ --------- --------- --------- ----------- -----------
2 CUF = 0.001554

Table A-2
EAF - Loss of Letdown with Delayed Return to Service

Pair Ui Fen,i Uen,i


1 1.87E-06 6.04 1.13E-05
2 0.001552 2.54 3.94E-03
Uen = 0.003952
Note: Stress cycle #2 has an alternating stress intensity Salt below the threshold limit. Therefore,
Fen=1.0 for this pair is justified. The detailed calculation is shown above for illustration.

g A-9 h

13307342
13307342
13307342
The Electric Power Research Institute Inc., (EPRI, www.epri.com)
conducts research and development relating to the generation, delivery
and use of electricity for the benefit of the public. An independent,
nonprofit organization, EPRI brings together its scientists and engineers
as well as experts from academia and industry to help address challenges
in electricity, including reliability, efficiency, health, safety and the
environment. EPRI also provides technology, policy and economic
analyses to drive long-range research and development planning, and
supports research in emerging technologies. EPRI’s members represent
more than 90 percent of the electricity generated and delivered in the
United States, and international participation extends to 40 countries.
EPRI’s principal offices and laboratories are located in Palo Alto, Calif.;
Charlotte, N.C.; Knoxville, Tenn.; and Lenox, Mass.

Together...Shaping the Future of Electricity

Program:
Nuclear Power

© 2011 Electric Power Research Institute (EPRI), Inc. All rights reserved. Electric Power
Research Institute, EPRI, and TOGETHER...SHAPING THE FUTURE OF ELECTRICITY are
registered service marks of the Electric Power Research Institute, Inc.

1022876

Electric Power Research Institute


3420 Hillview Avenue, Palo Alto, California 94304-1338 • PO Box 10412, Palo Alto, California 94303-0813 USA
800.313.3774 • 650.855.2121 • askepri@epri.com • www.epri.com
13307342

You might also like