You are on page 1of 11

Proceedings of the ASME 2016 Pressure Vessels and Piping Conference

PVP2016
July 17-21, 2016, Vancouver, British Columbia, Canada

PVP2016-63678

SUMMARY AND APPLICATIONS OF THE NEW FATIGUE RULES IN PART 14 OF API


579-1/ASME FFS-1 AND WRC 550

Aaron Stenta Edrissa Gassama Daniel W. Spring


The Equity Engineering Group The Equity Engineering Group The Equity Engineering Group
Shaker Heights, Ohio, USA Shaker Heights, Ohio, USA Shaker Heights, Ohio, USA
astenta@equityeng.com egassama@equityeng.com dspring@equityeng.com

Jeffrey Cochran Charles Panzarella David Osage


The Equity Engineering Group The Equity Engineering Group The Equity Engineering Group
Shaker Heights, Ohio, USA Shaker Heights, Ohio, USA Shaker Heights, Ohio, USA
jcochran@equityeng.com chpanzarella@equityeng.com daosage@equityeng.com

ABSTRACT dλ plastic multiplier


The third edition of API 579-1/ASME FFS-1 Fitness-For- Df cumulative fatigue damage factor
Service includes a new Part 14 covering fatigue assessment D f ,θ ,φ cumulative candidate plane fatigue damage factor
procedures for in-service components. This new part provides δ ij Kronecker delta function
methods for estimating the time to crack initiation using strain- ∆ε strain range
life approaches, which are important for low-cycle fatigue, and ∆ε eq ,k actual equivalent strain range for the k th cycle
is written as a multi-tiered approach covering screening, current ∆ε eqe ,k elastic equivalent strain range for the k th cycle
∆ε eq ,k
p
design code methods, and advanced methods that take into plastic equivalent strain range for the k th cycle
account the latest in technology. Cycle counting methods for ∆ε ij , ∆ε ije increment of total and elastic strain components
both welded joint and smooth-bar fatigue methods are included ∆ε k actual strain range for the k th cycle
for uniaxial and multiaxial loading histories. In addition, the ∆ε ke elastic strain range components for the k th cycle
multiaxial incremental plasticity correction procedure presented ∆ε kp plastic strain range components for the k th cycle
in the new Level 3 Assessment represents a significant update to ∆ε ije,k elastic strain range components for the k th cycle
the plasticity correction used previously. Part 14 of API 579- ∆ε ij ,k
p
plastic strain range components for the k th cycle
1/ASME FFS-1 is a procedural document that is written for ∆ε N , k normal strain range on the k th ∆γ max plane
accurate and practical implementation. WRC Bulletin 550 is the ∆γ max,k maximum shear strain range for the k th cycle
basis for Part 14 and provides the background and supporting ∆Lij actual general loading range components
documentation for its development. This paper summarizes the ∆S P ,k primary plus secondary plus peak equivalent stress
work that went into developing Part 14, and WRC 550, and range for the k th cycle
provides a specific example of its implementation for a practical ∆σ stress range
problem of interest using software developed by The Equity ∆σ eq ,k actual equivalent stress range for the k th cycle
Engineering Group, Inc. In-depth comparisons, key findings,
∆σ ij , ∆σ ije increment of total and elastic stress components
and limitations of the presented methods are included.
∆σ ij ,k actual stress range components for the k th cycle
∆σ ije,k elastic stress range components for the k th cycle
NOMENCLATURE
∆σ N normal stress range on the plane of ∆τ max
aij deviatoric back stress components
∆τ max maximum shear stress range
αij back stress components
e nominal strain for uniaxial plasticity correction
b fatigue strength exponent
ε, ε e, ε p actual, elastic, and plastic strain
c fatigue ductility exponent
ε ij , ε ije , ε ijp actual, elastic, and plastic strain components
Cγ Cγ = 1.5 or 6 for engineering or true strains

1 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


ε max
(i )
maximum principal component of the strain tensor The third edition of API 579-1/ASME FFS-1 includes a
ε min
(i )
minimum principal component of the strain tensor new Part 14 for fatigue damage assessment. Four methods are
εN normal strain on the plane of γ max presented that use strain-like concepts based on smooth bar and
ε ′f fatigue ductility coefficient welded joint fatigue curves. The methods include legacy ASME
E modulus of elasticity for material of interest techniques that have been updated to include more modern cycle
G shear modulus for the material of interest counting methods, a new multi-channel critical plane approach
γ max maximum shear strain based on the Brown-Miller strain-life equation, and an
k cycle index incremental multiaxial Neuber’s plasticity correction using a
K material parameter for the hardening behavior combined isotropic/kinematic-hardening Chaboche model. An
Kt concentration factor overview of API 579-1/ASME FFS-1, in much greater detail
K e ,k fatigue penalty factor for the k th cycle than that summarized here, is presented in [2].
Lij actual general loading components One of the difficulties encountered during the development
Leq Von-Mises of the generic loading tensor of Part 14 was distinguishing between the great number of
LReq relative Von-Mises with respect to reference R fatigue methods that have been developed thus far, which are
M total number of counted cycles often slightly modified versions of one another, in order to settle
ncss material parameter for cyclic stress-strain relation on a select few for the purpose of standardization. In addition,
n̂ normal vector of the critical plane many of these methods are not described in sufficient detail in
nk number of repetitions of each cycle the literature to be implemented without any ambiguity. The use
Nf cumulative number of cycles to failure of finite element analysis (FEA) methods in recent years
N f ,k number of cycles to failure for the k th cycle necessitates a more systematic and precise algorithmic
ν Poisson’s ratio description of each method to avoid any confusion and maintain
ν eff effective Poisson’s ratio consistency of the results.
q time index To fill this gap in the literature and take a step towards
qF first time index of the relative equivalent loading standardization, WRC Bulletin 550 was written to clarify some
φ rotation angle about the x-axis of these issues and to provide more detailed step-by-step
S nominal stress for uniaxial plasticity correction procedures for the selected methods in Part 14 from the
Salt ,k alternating equivalent stress for the k th cycle perspective of numerical implementation [3]. The methods
Sij actual components of the deviatoric stress tensor selected are not necessarily based on some universal
σ,σe actual and elastic stress determination about which method is best, but are rather
σ ij , σ iej actual and elastic stress components focused on which methods are adequate for the low-cycle
σ eq , σ eeq actual and elastic equivalent stress (Von-Mises) fatigue applications that are commonly encountered in the
σN normal stress on the plane of τ max refining and petrochemical industry.
σ ′f fatigue strength coefficient WRC 550 serves as a stand-alone repository of all the
σ mean ,k Morrow mean stress correction for the k th cycle knowledge required to perform the selected fatigue damage
σ N .mean ,k mean normal stress on the plane of γ max assessment procedures: (i) generating the elastic or elastic-
σ max,k maximum stress for the k th cycle plastic loading histories, (ii) gathering the necessary material
σ min ,k minimum stress for the k th cycle data, (iii) performing plasticity correction for the effects of local
τ max maximum shear stress plasticity, (iv) identifying the significantly damaging cycles, and
x, y , z global Cartesian coordinate system (v) choosing the correct fatigue damage model for calculating
x′, y ′, z′ transformed coordinate system the cumulative fatigue damage. This paper summarizes the work
q time index that went into developing Part 14 and WRC Bulletin 550 and
X current loading range for uniaxial cycle counting provides specific examples of its implementation for several
Y previous loading range for uniaxial cycle counting problems of practical interest. All the algorithms discussed in
the bulletin (many of which are summarized here) have been
1. INTRODUCTION implemented as stand-alone python modules that can then be
The API 579-1/ASME FFS-1 Standard was developed to incorporated directly into commercial FEA software through
provide guidance for conducting Fitness-For-Service (FFS) user-defined subroutines.
assessments of flaws commonly encountered in the refining and In this paper, we provide a brief overview of the fatigue
petrochemical industry that occur in pressure vessels, piping, assessment procedures in the new Part 14, discuss the most
and tankage [1]. Three levels of assessment are provided for significant updates that are presented in Annex 14C in greater
each flaw and damage type. Level 1 is the most conservative detail (including advanced multiaxial plasticity correction and
and the easiest to use. Practitioners usually proceed sequentially cycle counting procedures), and highlight the most significant
from a Level 1 to a Level 3 assessment (unless otherwise strain-life fatigue damage comparisons from WRC 550. We
directed by the assessment techniques). discuss our findings and demonstrate a FEA application for

2 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


comparison of the multiaxial methods. The two multiaxial • Method D – Screening based on the materials of
methods include: (i) cycle counting the equivalent loading construction (limited applicability), construction details,
tensor using the Wang-Brown algorithm (Level 2 Methods A/B) loading history, and welded joint fatigue curve data.
to account for variations in crack initiation orientation between
loading cycles and identify non-uniform fatigue damage, and 2.2 Level 2 Fatigue Assessment – Overview
(ii) cycle counting using a critical plane approach (Level 3 Method A – In this method, the fatigue damage and
Methods A/B) to project the loading tensor onto candidate remaining life are computed based on an effective equivalent
planes and selecting the plane of maximum unidirectional stress range obtained from a linear elastic stress analysis, and a
fatigue damage using a multi-stage Rainflow approach and smooth bar fatigue curve. This procedure is the ASME fatigue
advanced strain-life models. In-depth discussions are included design method that was developed in the 1960’s and has
for both methods to highlight the inconsistencies in the survived, basically unaltered. The procedure was updated in
literature when determining the damage parameters of interest. 2007 to calculate the alternating stress amplitude for the k th
cycle using the equivalent stress range, Equation (1-2), where
2. PART 14 OF API 579-1, ASME FFS-1 (METHODS) K e ,k is the well-known plasticity correction factor [4].
The fatigue rules in ASME Section VIII, Division 2, Part 5 K ⋅ ∆σ eq ,k
were a starting point for the development of Part 14. In Section Salt ,k = e ,k (1)
2
VIII, Division 2, the fatigue analysis methods are based on
1 ( ∆σ 11,k − ∆σ 22,k ) + ( ∆σ 11,k − ∆σ 33,k ) +
0.5

either smooth bar or welded joint test specimens. This same  2 2



∆σ eq,k =  (2)
approach was followed for Part 14; therefore, the assessment 2  ( ∆ σ − ∆ σ ) 2 + 6 ( ∆ σ 2 + ∆ σ 2 + ∆σ 2 ) 
procedures in Part 14 may be summarized as follows:  22, k 33, k 12, k 13, k 23, k 

• Smooth bar fatigue assessment methods may be used for The Wang-Brown multiaxial cycle counting algorithm was
components with or without welds, while welded joint added to the new Part 14 to identify the k th cycles for variable
methods shall only be used for welded joints. amplitude loading and is described in Section 3.5.
• Smooth bar fatigue assessment methods are applicable up to Method B – In this method, the fatigue damage and
the maximum number of cycles given on the fatigue curves, remaining life are computed based on an effective equivalent
while welded joint methods and curves do not exhibit an strain range obtained from an elastic-plastic stress analysis, and
endurance limit and may be used for any number of cycles. a smooth bar fatigue curve. A cyclic plasticity algorithm with a
• If thermal transients result in a through-thickness stress combined isotropic/kinematic hardening model is to be used.
difference at any time greater than the steady state difference The alternating stress amplitude is calculated using the
for welded joint methods, then the number of design cycles equivalent strain range for the k th cycle from the Wang-Brown
shall be the smaller of the number of cycles for the base multiaxial cycle count, as defined in Equations (3-6).
metal using the smooth bar fatigue method, and the number E ⋅ ∆ε eq ,k
of cycles for the weld using the welded joint method. Salt ,k = (3)
2
WRC 550 extends these concepts by including the most
∆ε e q , k = ∆ε eeq ,k + ∆ε eqp ,k (4)
popular strain-life damage models for comparison. If cyclic and
 ( ∆ε e − ∆ ε e ) 2 + ( ∆ ε e − ∆ ε e ) 2 + 
0.5
fatigue material properties are known, it is recommended to
consider these models after performing the plasticity correction  11,k 22, k 22, k 33, k

 e 
1
( ∆ε 33,k − ∆ε11,k ) +
2
and/or cycle counting procedures in Part 14, Annex 14C. ∆ε eqe ,k
= 
e
 (5)
2(1 + ν )
 
Cγ ⋅ ( ( ∆ε12,k ) + ( ∆ε 23,k ) + ( ∆ε13,k ) ) 
e 2 e 2 e 2
2.1 Level 1 Fatigue Assessment – Screening
The Level 1 assessment procedure is a fatigue screening
 ( ∆ε p − ∆ ε p ) 2 + ( ∆ ε p − ∆ ε p ) 2 + 
0.5
criterion to identify the need for a more detailed fatigue
assessment method. If any one of the screening methods is  11,k 22, k 22, k 33, k

2 
( ∆ε 33,k − ∆ε11,k ) +
2
satisfied, then a Level 2 or 3 fatigue assessment is not required. ∆ε= p p p
(6)
3  
eq , k
• Method A – Experience with comparable equipment  
Cγ ⋅ ( ( ∆ε12,k ) + ( ∆ε 23,k ) + ( ∆ε13,k ) ) 
p 2 p 2 p 2
operating under similar conditions.
• Method B – Screening based on materials of construction
Though computationally expensive, this method has the
(limited applicability), construction details, loading history,
advantage of evaluating plastic strains accurately even with
and smooth bar fatigue curve data.
significant net-section plasticity and is effective for many low-
• Method C – Screening based on the materials of
cycle fatigue problems.
construction (unlimited applicability), construction details,
Method C – In this method, the fatigue damage and
loading history, and smooth bar fatigue curve data.
remaining life are computed based on an equivalent structural
stress range obtained from a linear elastic stress analysis, and a

3 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


welded joint fatigue curve. The equivalent structural stress is a product of the elastic-plastic stress and strain concentration
function of the membrane and bending stresses normal to a factors. In doing so, he showed that the product of the elastic
hypothetical crack plane on which a fatigue crack would occur. stresses and strains that is computed from elastic FEA is
Neuber’s plasticity correction procedure is included after approximately equal to the product of the elastic-plastic stresses
cycle counting to compute the alternating stress range. This is and strains. Neuber stipulates that, for this approximation to
only applicable for minimum yielding with local plasticity at the hold, the region of plasticity must be small and be completely
notch root or weld zone location only [5]. This method is surrounded by a zone of elastic material.
recommended for evaluation of welded joints that have not been
machined to a smooth profile. Welded joints with controlled
smooth profiles may also be evaluated using Method A or
Method B with a suitable fatigue strength reduction factor.
The addition of the welded joint fatigue technology to the
2016 Edition of API 579-1/ASME FFS-1 represents a
significant departure and upgrade from the legacy methods for
welded joints that are more in line with other international
codes and standards. Plans to upgrade the welded joint method
to the Structural Strain approach (low-cycle fatigue
applications) are in progress but not included in the 2016
Edition of API 579-1/ASME FFS-1 [6].

2.3 Level 3 Fatigue Assessment – Overview


The Level 3 fatigue assessment uses a critical plane
approach with a strain-life damage model to calculate the plane
of maximum fatigue damage (the critical plane). Level 3
Method A uses the multiaxial Neuber’s plasticity correction
procedure with the Chaboche isotropic/kinematic-hardening
model to determine the elastic-plastic loading history, prior to
cycle counting, and Method B uses the known elastic-plastic
loading history directly.
The Brown-Miller strain-life equation, adjusted for mean
stress effects, is the Level 3 damage model proposed in Part 14.
∆γ max,k ∆ε N ,k σ ′ − 2σ N .mean ,k
( 2 N f ,k ) + 1.75ε ′f ( 2 N f ,k ) (7)
b c
= + 1.65 f
2 2 E
A flow diagram summarizing the fatigue damage assessment Figure 1 – General procedure for the fatigue assessment methods
methods in Part 14 (Levels 2 and 3) is shown in Figure 1. in API 579-1, Part 14 and WRC 550 that incorporates the cycle
counting algorithms and fatigue damage methods.
3. ANNEX 14C AND WRC 550 (METHODS)
WRC 550 provides a self-contained review of several For the elastic loading history at the weld toe, notch tip, or
currently available approaches for plasticity correction, cycle critical location, Neuber’s rule, as defined in Equation (8), is
counting, and fatigue damage that are based on the methods used, where σ e and ε e are the elastic stress/strain components
presented in Part 14 of API 579-1/ASME FFS-1. The intention and σ and ε are the plastically corrected (actual) stress/strain
is to gather all required documentation into a single report, so components. However, if the nominally elastic stress/strain
that the extensive literature search and compilation of histories and stress concentrations some distance away from the
information that was performed to arrive at Part 14 and WRC critical location are known, then Neuber’s rule, as defined in
550, does not have to be repeated by others with similar goals. Equation (9), is used, where K t = σ / S = ε / e is the
Here, we summarize the most important components and key concentration factor, S is the nominal stress, and e is the
findings of each method to familiarize the community in a nominal strain. For both situations, a constitutive relation is
concise and precise manner. included to solve for σ and ε . The Ramberg-Osgood relation
(Equation (10)) is used to obtain the cyclic stress-strain curve
3.1 Uniaxial Plasticity Correction up to the first turning point, and the Masing’s hysteresis relation
The most well-known method for plasticity correction was (Equation (11)) is used for the remaining loading history [8].
proposed by Neuber [7]. Neuber developed a general theory The most recent turning point is used to capture the effects of
relating the theoretical elastic stress concentration factor to the cycle closure and material memory, as shown in Figure 2 [9].

4 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


After closure of cycle 4-5 the turning point is reinitialized to damage model. It is always suggested to cycle count the actual
point 3, and then after closure of cycle 2-3 the turning point is elastic-plastic loading history, if possible. It is also sometimes
reinitialized to point 1 to account for material memory. required to calculate the corresponding elastic-plastic maximum
σ eε e = σ ε (8) or mean stress within each cycle.
2 The DS2 algorithm can count cycles as the data is received
2 S
( K t S )(=
K t e) K=
t σε (9) and does not require reordering. The only additional variation
E from the DS1 algorithm to the DS2 algorithm is that an
σ σ 
1/ ncss
additional condition is included to track the starting point S of
ε =ε e + ε p = +   (10)
E K each loading block. A cycle is identified when X ≥ Y and Y
does not contain S . When the end of the loading history is
∆σ  ∆σ 
1/ ncss

∆ε
= +2 ⋅   (11) reached, the data points from the beginning of the history, prior
E  2⋅ K  to S , are appended to the end of the loading history. The
counting procedure is concluded when S has been reread or
less than three data points remain. Both DS1 and DS2 are
shown to produce the same cycle counting results. Generating
the hysteresis curve with material memory and cycle closure
effects requires the data to be reordered, thus the DS1 algorithm
is the most practical choice for uniaxial cycle counting.

3.3 Uniaxial Fatigue Damage


Uniaxial fatigue damage models for low-cycle fatigue
applications (strain-life methods) are based on local strains as
measured from a strain gauge or calculated using FEA. The true
stress/strain is used to develop the underlying relationships for
Figure 2 – Elastic-plastic hysteresis curve calculated from both elastic and plastic strain versus fatigue endurance. These
Neuber’s rule and Ramberg-Osgood for Example 6.2 in Lee [9]. relationships are derived from smooth bar fatigue experiments,
Closed loops identify cycles as labeled in the figure.
and correlate well with known data [8].
3.2 Uniaxial Cycle Counting The total strain is written as the sum of the elastic and
The Rainflow method was developed by Endo and plastic strains and is the basis for the strain-life model (shown
Matsuishi and counts closed hysteresis loops in the here with Morrow mean stress correction), see Equation (14).
uniaxial/proportional stress/strain response [10]. While many Mean stress effects and the log-linear elastic and plastic strain
variations of the Rainflow method have been successfully ranges are illustrated in Figure 3.
∆ε k ∆ε ke ∆ε kp σ ′ f − σ mean ,k
proposed, all typically yield the same cycle count. Thus, two of
( 2 N f ,k ) + ε ′ f ( 2 N f ,k )
b c
= + =
the most notable approaches, Downing-Socie I (DS1) and 2 2 2 E (14)
Downing-Socie II (DS2), are presented [11]. wher=e σ mean ,k 0.5 (σ max ,k + σ min ,k )
The DS1 algorithm requires that the complete loading
∆ε
history be known prior to cycle counting, such that the data can 2
be initially reordered and filtered within the algorithm. The DS1
Strain Amplitude (Log Scale)

ε’f
method is used in the Level 2 Method C assessment. It may also
be used for proportional low-cycle fatigue applications with an
σ’f c
appropriate strain-life damage model.
Ε
After reordering and filtering the loading history, three b

consecutive loading points ( q −2 L, q −1L, q L ) are used to


σ’f − σm
Ε
calculate the beginning and ending loading ranges for each
loading event, as defined in Equations (12) and (13).
= X abs  q L − q −1L  (12) 2Nf
Cycles to Failure (Log Scale)
= Y abs  L − L 
q −1 q −2
(13) Figure 3 – Visualization of the Morrow mean stress correction that
If X ≥ Y , count Y as a cycle, remove its loading points is used in the strain-life Equation (14) and Brown-Miller Equation
(7). Nominally compressive stresses increase fatigue life and
from the history, reinitialize the time indices, and compare the
nominally tensile stresses decrease fatigue life.
ranges for the next three loading points. The procedure is
continued until there are no longer three data points to compare. A nominally compressive mean stress increases fatigue life,
The cycle ranges are then used directly in a strain-life fatigue while a nominally tensile mean stress decreases fatigue life. An

5 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


additional damage model that accounts for mean stress effects directly with Hooke’s law, (v) if it is outside the current yield
through the maximum stress per cycle is the Smith-Watson- surface solve the system of 12 equations for the 12 plastically
Topper equation (15). corrected actual elastic-plastic stress and strain components, (v)
∆ε k (σ ′ f )2 update and translate the yield surface(s) using the material
( )
2b
σ max ,k = 2 N f ,k + σ ′ f ⋅ ε ′ f (2 N f ,k )b+ c (15) model stipulations, and (vi) if the calculated actual loading
2 E
point exceeds the limiting yield surface then radially return back
The strain ranges and mean/max stresses from the Rainflow
onto the limit surface using bisection.
cycle count are used to solve the damage model(s) for the
The intention is to use the plasticity correction method to
number of permissible cycles, per cycle, N f ,k , that is the basis
obtain the elastic-plastic loading history from the linear elastic
for the strain-life curve (ε − N ) . The cumulative fatigue
FEA simulation to limit the computational cost of a FEA
damage can then be calculating using Miner’s rule, Equation
simulation and to reduce any mesh sensitivity errors. It is
(16), where nk is the number of repetitions of each cycle, and
assumed that all elastic loading histories are obtained at the
M is the total number of identified cycles. The resulting
local notch/crack root location and are not nominal strains (i.e.
damage factor D f is the cumulative fatigue damage for the
at some location away from the notch/crack root location).
entire loading history block of interest.
M
Additionally, Neuber’s rule stipulates that plasticity remains
n
= D f ∑ k ≤ 1.0 (16) local to the notched region. Additional details for the plasticity
k =1 N f ,k correction procedures that are implemented in Part 14 and
To perform a uniaxial fatigue damage assessment the WRC 550, with in-depth examples, are included in a
loading history should be proportional in nature, the user should complimentary paper [5]. Here, it is shown that Neuber’s rule is
have the necessary material parameters for the damage model or increasingly conservative for increasing plastic strain, when
an experimental fatigue curve for the given material, and the used appropriately.
elastic-plastic loading history (stress and strain) should be An example of the multiaxial plasticity correction is
known for all critical locations of interest. If the linear elastic compared to FEA for a cyclic pressure of 150 MPa that is
loading history is known then the uniaxial plasticity correction applied to a notched beam in the x-direction, away from the
procedure should be used prior to cycle counting. If either the local notch. The resulting elastic-plastic stress-strain response in
elastic-plastic stress or strain is unavailable, then the uniaxial Figure 4 illustrates that, even for small local strains (i.e
constitutive relations should be used to calculate the other ε < 0.01 for the 150 MPa example), Neuber’s method is
taking into account material memory and cycle closure effects. conservative. However, it is demonstrated in Section 4 that the
Uniaxial damage prediction examples for various models, fatigue life predictions for the two approaches are comparable;
materials, and loading histories are included in WRC 550. as the strain ranges and mean stress effects follow similar
trends, and only the first loading event is conservative (the first
3.4 Multiaxial Plasticity Correction loading event is not included in fatigue calculations).
For multiaxial states of stress (either proportional or non-
proportional), the incremental Neuber’s rule, along with an
appropriate elastic-plastic constitutive relation, can be used to
estimate the elastic-plastic stress and strain loading histories
from the purely elastic stress and strain histories as defined in
Equations (17-18), respectively. Here, σ ije , ε ije , σ ij , ε ij are the
elastic and actual stress/strain components, and ∆σ ije , ∆ε ije ,
∆σ ij , ∆ε ij are the elastic and actual stress/strain component
increments. The cyclic plasticity model is the combined
isotropic/kinematic-hardening Chaboche model with multiple
back stresses, where the total back stress is the sum of the
individual back stresses [12].
ε ije ∆σ ije + ∆ε ijeσ ije= ε ij ∆σ ij + ∆ε ijσ ij (17) Figure 4 –Stress vs. strain response in the x-direction, comparing
results from a fully elastic-plastic FEA simulation and the
1  (1 + ν ) ∆σ ij − ν ⋅ δ ij ( ∆σ 11 + ∆σ 22 + ∆σ 33 )  incremental Neuber’s method for a cyclic pressure of 150 MPa .
∆ε ij =   (18)
E  + 1.5 ⋅ d λ ⋅ ( Sij − aij ) / σ eq  3.5 Multiaxial Cycle Counting
The general multiaxial plasticity correction procedure Perhaps the largest contribution to the new Part 14 is the
must: (i) obtain the linear elastic loading history at the notch improved cycle counting procedures that are presented in
root, (ii) initialize the material model and the yield surface(s), enough detail for practical numerical implementation. It seems
(iii) calculate whether the increment is within or outside the most appropriate to choose a set of standardized multiaxial
yield surface (elastic or plastic), (iv) if it is elastic, update it methods that are straight-forward to implement, well-

6 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


established in the literature, and identify cycles in a similar mode of cracking is unknown, several damage models should
manner to the proposed uniaxial cycle counting procedures. be considered and the most conservative chosen.
Thus, we have chosen the Wang-Brown method for cycle Langlais extended the critical plane approach to
counting the entire multiaxial loading tensor, and the Critical incorporate multiple damage parameters (multi-channel
Plane method for projecting the loading history onto candidate approach) [16]. Here, it is assumed that the main channel is
planes and calculating the plane of maximum fatigue damage. used for counting and the auxiliary channel(s) may occur
The Wang-Brown method accounts for multi-directional crack anywhere within the main-channel cycle range. For example,
fatigue damage, while the Critical Plane method is more the Brown-Miller main channel is the shear strain range ∆γ max
appropriate for unidirectional fatigue damage. and the auxiliary channel is the normal strain range on the shear
Wang Brown – This method counts half-cycles as paths of plane ∆ε N . Alternatively, in the Fatemi-Socie model, the main
increasing equivalent stress/strain in the relative equivalent channel is the shear strain range ∆γ max , while the auxiliary
stress/strain space. This counting procedure may be performed channel is the maximum stress on the shear plane σ N ,max .
for either stress or strain, and that cycle counting is performed
only once as the entire loading tensor is counted [13].
First, the loading history is reordered to begin with the
loading point of maximum Mises stress/strain. The relative
equivalent stress/strain, q LReq , is then calculated with respect to
the first point of the reordered loading history as defined in
Equations (19) and (20) (shown here for strain).
0.5
( ∆L11 − ∆L22 )2 + ( ∆L22 − ∆L33 )2 
1  
Leq =  + ( ∆L11 − ∆L33 ) 
q R 2
(19)
2 (1 + ν eff )  
 +Cγ ⋅ ( ∆L12 + ∆L13 + ∆L23 ) 
2 2 2

∆Lij = q
Lij − qF Lij for i, j = 1, 2,3 (20)
Figure 5 – Relative equivalent strains with respect to the turning
Here, qF is the time index of the reference point for the relative points of the uncounted blocks demonstrating the eight counted
equivalent stress/strain. The parameter ν eff is the elastic-plastic reversals for the Wang Brown loading history.
effective Poisson’s ratio, as per Equation (21). It is assumed that
ν eff = 0.4 if the elastic strain increment is approximately The traditional transformation rule of the Cauchy stress
equivalent to the plastic strain increment. The two available tensor is considered and illustrated in Figure 6 [14]. Sufficient
literature examples that are reproduced in WRC 550 use increments should be considered, such that the plane of
ν eff = 0.5 , as extensive plasticity was observed [13-14]. maximum damage is accounted for (e.g. ∆θ = 5 and ∆φ = 5 ).
σ eq After transforming the tensor onto the candidate plane of
ν eff = 0.5 − (0.5 − ν ) (21)
E ⋅ ε eq interest, the multi-channel components should be calculated for
the entire translated loading history. After using the traditional
For each relative equivalent loading block, there is a half- Rainflow algorithm to cycle count the main channel component,
cycle (reversal) that is identified as the path of increasing the auxiliary components are determined for each cycle. The
relative equivalent stress/strain. There are also segments of the candidate plane’s fatigue damage is then calculated for each
loading block that remain uncounted, as they do not result in a cycle and Miner’s rule is used to determine the cumulative
monotonically increasing equivalent stress/strain. These damage on that plane, as per Equation (22). The critical plane is
remaining loading blocks must be counted separately with their then defined as the candidate plane with maximum cumulative
initial point defined as the turning point for the newly fatigue damage, as per Equation (23).
considered loading block. This process is continued until all M
nk
loading paths in the equivalent stress/strain space are counted. =
D f ,θ ,φ ∑N
k =1
≤ 1.0 (22)
The eight loading paths identified in the original Wang-Brown f ,k

example (equivalent-strain) are illustrated in Figure 5. 1


Critical Plane – This method was developed by D f = max  D f ,θ ,φ  , and N f = (23)
θ ,φ Df
Bannantine and Socie, wherein the fatigue damage is estimated
by translating the elastic-plastic loading history onto all possible Since the translated loading history is constant for all cycles on
candidate planes and determining the plane of maximum a single plane, the critical plane method only accounts for
damage [15]. In all critical plane methods the damage model is fatigue damage in a single direction on each plane. When
incorporated into the counting procedure. Thus, the accuracy of significant multi-directional fatigue damage is observed, it is
the predictions is in the choice of the damage model. If the suggested to use the Wang-Brown method.

7 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


The multiaxial strain-life damage models that are
z
considered include the Brown-Miller and Fatemi-Socie models,
z’ as per Equations (26-27). The crack predicted by the Brown-

x’ Miller damage model is assumed to initiate in Mode II. The
′ qσ
13 N
Brown-Miller model also requires an additional auxiliary
y’ parameter to account for the normal mean stress effects on the


φ 12
plane of γ max , where σ N ,mean,k = σ mean,k / 2 . As an attempt to
θ y
improve the Brown-Miller damage model for shear plane
dominant cracks, Fatemi-Socie proposed that the normal stress
(rather than the normal strain) on the plane of maximum shear
x be considered. Neither of the above two models were developed
Figure 6 – Transformation of the Cartesian coordinate system to for tensile (Mode I) cracks, which is the intention of the Smith,
the primed coordinate system. The coordinate axis x′ is normal to Watson, and Topper (SWT) model that is defined in Equation
the candidate plane in consideration. (15) [14]. For multiaxial loading histories the strain range is
replaced with the maximum principal strain range.
3.6 Multiaxial Fatigue Damage
∆γ max,k ∆ε N ,k  σ ′ − 2σ N ,mean ,k 
( ) ( )
bk ck
To improve our decision regarding the selection of an = + 1.65  f  2 N f ,k + 1.75ε ′f 2 N f ,k (26)
appropriate fatigue damage model, we must understand the 2 2  E 
various forms of crack nucleation and growth. Loading a ∆γ max,k  σ  σ′f
(2N ) + 3 ⋅ ε ′f ( 2 N f ,k ) (27)
bk ck
component under tension results in opening or Mode I cracks.  1 +=
k f N ,max 

σ′f  3 ⋅G
f ,k
Mode II cracks are a result of in-plane shear loading, while 2 
Mode III cracks are a result of out-of-plane shear loading. A
crack that is growing radially and tangentially is a mixed Mode 4. LOW-CYCLE FATIGUE APPLICATION
II and III crack. After crack nucleation, the crack will initially This section presents a detailed example of a notched beam
begin to grow along the plane of maximum shear stress (i.e. subject to a cyclic loading pressure for comparison of the
Mode II). As the crack grows larger it will tend to grow fatigue lives computed based on the elastic-plastic solution from
perpendicular to the principal tensile stress (i.e. Mode I) [14]. FEA and the multiaxial plasticity correction procedure of Part
While preparing Part 14 and WRC 550, inconsistencies 14 [17]. The newest additions to Part 14, the Wang-Brown
were observed in the literature with respect to choosing the (Level 2 Method A/B) and Critical Plane (Level 3 Method A/B)
appropriate damage parameter for a given damage model. This cycle counting procedures with the Brown-Miller and Fatemi-
is especially true for the interpretation of the maximum shear Socie strain-life damage models are compared. Constant
strain and normal strain on the plane of maximum shear. The amplitude, cyclic, and uniformly distributed pressures of
global maximum shear and normal strain on the plane of global 150 MPa and 350 MPa are applied to the beam face in the x-
maximum shear for a general tensor are defined as direction, opposite the notch root, for a total of 50 combinations
γ=
max ε max(
(i )
− ε min
(i )
, and ) (24) of loading/unloading events. The material model in both the
FEA and the plasticity correction procedure is the Chaboche
ε N =0.5 ⋅ (ε (i )
max + ε min
(i )
), (25) model with a single back stress component. In-depth details and
where ε (i )
max and ε (i )
minare the maximum and minimum principal discussions for the FEA and incremental plasticity procedures
strain components, respectively. An alternative approach are included in WRC 550 and the complimentary Neuber’s
considers γ max representing the largest shear component and method paper [3,5].1 Here, we are concerned with the effect of
ε N the normal component of the transformed tensor on each the plasticity procedure on the simulated elastic-plastic loading
candidate plane. The two approaches do not necessarily history and the resulting fatigue life predictions.
produce the same γ max and ε N , and the later of the two was The material is AISI 1010 rolled steel ( E = 210 GPa and
derived for critical-plane methods only, which approximate the ν = 0.3 ), and plane strain conditions are assumed. The plastic
plane of global maximum shear by testing all possible planes. material properties and initial back stresses for the Chaboche
Additionally, there is a misconception as to how the constitutive model are listed in Table 1. An initial equivalent
cumulative fatigue damage is calculated. Some define the plastic strain of p0 = 0.43 (which leads to an initial yield
critical plane with maximum cumulative damage as the plane surface size of 411 MPa ) is assumed.
that maximizes the shear strain alone. However, it is more 1) Figures that are illustrated in the complimentary Neuber’s method paper and
correctly defined as the plane that maximizes the entire damage not included in this manuscript include: (i) the notched beam geometry, (ii) the
discretization of the FEA domain, (iii) the elastic loading history for 150 MPa ,
parameter (whatever the parameter may be). In WRC 550 the (iv) the elastic loading history for 350 MPa , (v) the localized zone of plasticity
effects of the various interpretations are considered, verified concentrated local to the notch root, (vi) the 50 loading event stress-strain
response for FEA vs. Neuber’s rule for 150 MPa (shown here in Figure 4 for
with known literature examples, and the most appropriate for the first 4 loading blocks), and (vii) the 50 loading event stress-strain response
low-cycle fatigue applications are properly identified. for FEA vs. Neuber’s rule for 350 MPa [5].

8 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


Table 1: Material properties for the Chaboche combined Table 2: Material and fatigue properties for carbon steel AISI 1010
isotropic/kinematic-hardening model [17].
(Uniform Material Law – Part 14 [1]).
σ0 σ max b C1 γ1 α11 0 α 22 0 α 33 0 ( )
τ ′= σ ′ / 3 , γ ′= 3 ⋅ ε ′, and σ y ≈ σ ′
200 2000 ε′ σ′
0.26 25500 81 128 -181 53 E UTS b c G kf
MPa MPa 2.1E5 365 8E4
0.59 547.5 -0.087 -0.58 1
For both loading events, Nueber’s assumption of local MPa MPa MPa
plasticity with nominally elastic stresses away from the notch As an example, the contour of cumulative fatigue damage for
root is satisfied [7]. Both of the predictions using Neuber’s the FEA elastic-plastic loading history, with a cyclic pressure of
method are conservative, and the larger load demonstrates the 150 MPa and the Brown-Miller strain-life model, is illustrated
increasing over-conservative effect of the approximation with in Figure 7. The model assumes plane strain conditions, thus the
increasing plasticity. The plasticity correction results for the original loading plane is recovered when φ = 90o . A cross-
150 MPa cyclic pressure follow the same trends as observed by section of the damage on this plane for all θ is shown in Figure
the FEA model (ratchetting and mean stress limit approaching 8a. The maximum damage occurs on the planes
zero). For the 350 MPa pressure, however, the two methods =φ 90 = o
, θ 30o and
= φ 90 =o
, θ 150o . Figure 8b illustrates
have an initial large relative error (the FEA model shows the fatigue damage along the φ = 90o plane for the Brown-
ratchetting, while Neuber’s rule shows the Bauschinger effect), Miller model and a cyclic pressure of 350 MPa . The
but approach the same limiting behavior of stress and strain. predictions using Neuber’s rule are overlaid on both curves for
Thus, longer fatigue life predictions would tend toward more comparison. It is clear that Neuber’s rule over-predicts the
comparable results between the two methods. damage, especially for the case with a larger load.
For both multiaxial loading cases and both material
models, the fatigue lives from the Brown-Miller and Fatemi-
Socie damage models for the Wang-Brown and Critical Plane
cycle counting procedures are compared. For all four scenarios
the damage parameter definition presented in Socie is used (i.e.,
the shear strain component on the selected candidate plane and
the normal strain component that is normal to the plane) [14]. It
is also assumed that the maximum damage on each plane occurs
at the maximum value of the damage parameter represented as
the combination of shear and normal strain. The material-
specific fatigue parameters are listed in Table 2, and the
cumulative damage and fatigue lives are summarized in Table 3.
The damage factor is the cumulative fatigue damage for the
entire loading history of interest (all 50 cycles), and the fatigue Figure 7 – Cumulative fatigue damage for the FEA (Abaqus)
life is the number of permissible repetitions of the entire loading loading history produced from the 150 MPa pressure using the
history block. Critical Plane method with the Brown-Miller damage model.

Table 3: Comparison of the fatigue damage between the cyclic response from the fully elastic-plastic FEA simulation (Abaqus) and the
incremental approach (Neuber), for horizontal cyclic pressures of 150 MPa and 350 MPa , the Wang-Brown and Critical Plane methods,
and the Brown-Miller and Fatemi-Socie damage models.

Critical Plane Wang-Brown


Brown-Miller
Brown-Miller Fatemi-Socie Brown-Miller Fatemi-Socie
(Global Max-Shear)
Damage Factor 0.00626 0.01054 0.01987 0.03078 0.00964
150 MPa
Fatigue Life 159.74 94.88 50.33 32.49 103.73
FEA
Damage Factor 0.06558 0.17423 0.1897 0.48437 0.0959
350 MPa
Fatigue Life 15.25 5.74 5.27 2.06 10.43
Damage Factor 0.00715 0.01249 0.02321 0.03596 0.01065
150 MPa
Fatigue Life 139.86 80.06 43.08 27.81 93.9
Neuber
Damage Factor 0.10501 0.2757 0.30037 0.8704 0.14848
350 MPa
Fatigue Life 9.52 3.63 3.33 1.15 6.73

9 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


with global maximum shear Brown-Miller is the lower bound.
All of the fatigue predictions, for this example, are within 21%
relative error of the average of the two.
Comparing the fatigue predictions from the elastic-plastic
FEA model and Neuber’s multiaxial plasticity correction, it is
clear that even though the shapes of the stress-strain curve are
very different, the fatigue lives are all within 9% and 28%
relative error of the average of the two for the 150 MPa and 350
MPa loading pressures, respectively. Additionally, the Fatemi-
Socie model demonstrates that the maximum normal stress
influences the fatigue damage more than the normal strain range
(a) in the Brown-Miller model, in this example, thus predicting
lower lives.

5. CONCLUSIONS
The intention of this paper is to familiarize the ASME
community with the procedures that are included in the new
Part 14 and WRC Bulletin 550. After reading this document the
reader should feel comfortable implementing the algorithms and
reproducing the examples that are presented. Hopefully, it is
apparent that these methods are not a one-stop-shop for
calculating fatigue damage, and that operator-experience and
expert knowledge is valuable for determining the appropriate
(b) model. Part 14 and WRC 550 aim to minimize this ambiguity.
Figure 8 – a) FEA (Abaqus) and Neuber’s rule cumulative fatigue In this manuscript we highlight the difficulties surpassed and
lives for the 150 MPa pressure on the φ = 90o plane using the key findings obtained while preparing Part 14 and WRC 550.
Critical Plane method with the Brown-Miller damage model. b) The As mentioned previously, we have limited our scope to
same comparison as (a), except for the 350 MPa pressure. low-cycle fatigue applications that are encountered in the
refining and petrochemical industry. The new Part 14 and WRC
A damage factor of one ( D f = 1) implies that the crack Bulletin 550 propose the most recent and well-established
initiates after one repetition of the complete 50 cycle loading uniaxial and multiaxial methods necessary for a complete
block. For instance, the FEA model using the Brown-Miller life fatigue damage assessment. This summarizes a major
prediction for the cyclic pressure of 150 MPa suggests that the advancement in the direction of standardization that is a
component initiates after approximately 160 repetitions of the necessity for practitioners in the industry. In this manuscript we
loading block ( N f ≈ 160 ) . Note, however, that due to the have included an in-depth demonstration and comparison of the
significant plasticity, it is not likely that the loading block would most significant Part 14 additions (i.e. the multiaxial Wang-
be repeated, and the loading history would actually converge Brown and Critical-Plane cycle counting algorithms) for a
toward zero mean stresses, increasing the cumulative fatigue more-specific low-cycle fatigue application than that presented
life, such that the current predictions would be conservative. in WRC 550.
The Wang-Brown method assumes that the orientation of
fatigue damage changes within the loading block for each cycle. REFERENCES
Thus the maximum damage for each cycle occurs on different [1] ASME and API, API 579-1/ASME FFS-1 2016 Fitness-
planes. The Critical Plane method, however, assumes that the For-Service. The American Society of Mechanical
entire loading block accumulates fatigue damage in a single Engineers, American Petroleum Institute, 2016.
direction (all cycles accumulate damage on the same plane). [2] D.A. Osage. Fatigue Assessment for In-Service
The results in Table 3, thus, correctly demonstrate the bounding Components–A New Part for API 579-1/ASME FFS-1
limits of the two models. Wang-Brown is overly conservative, Fitness-For-Service. Procedia Engineering, 133 (2015):
while Critical Plane is much less conservative. 320-347.
The original Wang-Brown Damage model that only cycle [3] A. Stenta, E. Gassama, D. Spring, C. Panzarella, and J.
counts the entire loading tensor once was developed to use the Cochran, Standardization of Fatigue Methods for Use in
true global maximum shear strain as the main component of the API 579-1/ASME FFS-1. Welding Research Council,
damage model (only having to perform the damage calculation Bulletin 550, 2016.
once). In this regard, the Critical Plane with Brown-Miller [4] D.A. Osage, and J.S. Sowinski, ASME Section VIII –
damage prediction is the upper bound and the Wang-Brown Division 2 Criteria and Commentary. ASME PTB-1,
ASME, New York, N.Y., 2013.

10 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo


[5] D. Spring, E. Gassama, A. Stenta, J. Cochran, C.
Panzarella. On the Applicability of Neuber’s Rule for Low-
Cycle Fatigue. ASME 2016 Pressure Vessels and Piping
Conference. American Society of Mechanical Engineers,
2016. (Submitted)
[6] Dong, P., Pei, X., Xing, S., and Kim, M.H., A Structural
Strain Method for Low-Cycle Fatigue Evaluation of
Welded Components. International Journal of Pressure
Vessels and Piping, 119 (2014) 39-51.
[7] H. Neuber, “Theory Of Notch Stresses: Principles For
Exact Calculation Of Strength With Reference To
Structural Form And Material,” 1958.
[8] J. Draper, Modern Metal Fatigue Analysis. 2008.
[9] Y.-L. Lee, M. E. Barkey, and H.-T. Kang, Metal Fatigue
Analysis Handbook: Practical Problem-Solving
Techniques for Computer-Aided Engineering. 2011.
[10] Endo, T., M. Matsuishi, K. Mitsunaga, K. Kobayashi, and
K. Takahashi. Rain flow method, the proposal and the
applications. Memoirs of Kyusyu Institute of Technology
28 (1974): 33-62.
[11] S. D. Downing and D. F. Socie, Simple Rainflow Counting
Algorithms. International Journal of Fatigue, vol. 4, no. 1,
pp. 31–40, 1982.
[12] J. L. Chaboche, A Review of Some Plasticity and
Viscoplasticity Constitutive Theories. International Journal
of Plasticity, vol. 24, no. 10, pp. 1642–1693, 2008.
[13] C. H. Wang and M. W. Brown, Life Prediction Techniques
for Variable Amplitude Multiaxial Fatigue—Part 1:
Theories. Journal of Engineering Materials and
Technology, vol. 118, no. 3, pp. 367–370, 1996.
[14] D. F. Socie and G. B. Marquis, Multiaxial Fatigue. 2000.
[15] J. A. Bannantine and D. Socie, A Variable Amplitude
Multiaxial Fatigue Life Prediction Method. American
Society of Testing and Materials, pp. 249–275, 1991.
[16] T. Langlais, Computational Methods for Multiaxial
Fatigue Analysis. University of Minnesota, 1999.
[17] I. Doghri, Fully Implicit Integration And Consistent
Tangent Modulus In Elasto-Plasticity. International
Journal of Numerical Methods in Engineering., vol. 36,
no. 22, pp. 3915–3932, 1993.

11 Copyright © 2016 by ASME

Downloaded From: http://proceedings.asmedigitalcollection.asme.org/pdfaccess.ashx?url=/data/conferences/asmep/90475/ on 03/19/2017 Terms of Use: http://www.asme.org/abo

You might also like