Journal of China Pressure Vessel Technology 1 (2003) 71-79

Fitness-for-service assessment of in-service pressure components:
petrochemical components
Chon L. Tsai
*
Department of Industrial, Welding and Systems Engineering, The Ohio State University
1248 Arthur E. Adams Drive, Columbus, OH 43221, USA



Abstract

Using the finite element analysis method, the fitness-for-service assessment procedure provided in API 579 [1] was evaluated with
respect to its assumption of uniform residual stress thought the joint thickness. In this paper, a welding-induced residual stress distribu-
tion in the plane that contains an elliptical surface crack was first obtained by a thermo-elastic-plastic modeling procedure. A
three-dimensional fracture mechanics model was then conducted with the incorporation of the estimated residual stress distribution.
The plasticity correction factors were determined due to the effect of residual stress. This paper demonstrated that incorporating the
actual residual stress distribution in the fracture analysis would reduce the failure driving force as a crack grows deeper due to the fact
that welding-induced residual stress is compressive in the middle thickness of the joint. The current API 579 assessment procedure for
crack-like flaws in weldments assumes a uniform residual stress at the level of yield stress plus 10 ksi. This assumption may result in
significant conservatism in calculating the toughness ratio.
Keywords: Fitness-for-service flaw assessment; Failure assessment diagram; API 579 assessment standard; Residual stress effect;
Semi-elliptical surface cracks.



1. Introduction

When significant flaws are found in welded petro-
chemical equipment during inspection, the engineering
critical assessments and decisions on repair, replace-
ment or continued operation are usually required in a
relatively short time. To assure continued safe opera-
tion without overconservatism, sufficient planning and
engineering analysis need to be carried out. Fit-
ness-for-service (FFS) assessments are quantitative
engineering evaluations that demonstrate the structural
integrity of a flawed or damaged component. They are
multi-disciplinary engineering analyses for decision on
(1) fit for continued safe operation, which ensures pre-
sent integrity of the component given a current state of
damage, operating loads, and environment condition;
(2) safety margin and re-rating for continued service,
which includes the limiting operating condition to
avoid failure of equipment containing a known or pos-
tulated flaw; (3) alter, repair, monitor, or replace; and
(4) guidance on inspection intervals. Some FFS (e.g.
BS PD6493 or BS7910 [2]) also intends to project re-
maining life for run-repair-replace decision based on
future operating conditions and environmental com-
patibility. The rational judgment should be based on a
balanced consideration among three factors of per-
formance, safety and cost.
The FFS assessment procedures can be used for
re-rating of pressure vessels designed and constructed

*
Corresponding author.
E-mail address: tsai.1@osu.edu (Chon L. Tsai).
to the ASME Boiler and Pressure Vessel Code, piping
systems designed and constructed to the ASME B31.3
Piping Code, and above-ground storage tanks designed
and constructed to API 650 and API 620 Specifications.
The rapidly becoming standard for conducting FFS
assessments is API 579 [1], Fitness-for-Service, which
describes standardized FFS assessment procedures for
pressurized equipment used in industry and supple-
ments the inspection assessment techniques in API 510,
API 570 and API 653.
API 579 provides three assessment levels for the
crack-like flaws to meet specific requirements of a
given application circumstance. The level I assessment
is limited to crack-like flaws in pressurized cylinders,
spheres, or flat plates away from all structural discon-
tinuities. The level II assessment applies for general
shell structures including crack-like flaws located at
structural discontinuities. Detailed information on ma-
terial properties and loading conditions are required,
and a stress analysis is typically performed to deter-
mine the state of stress at the location of the flaw. The
level III assessment is to evaluate those cases not
meeting Level I or Level II assessment requirements,
or flaws that may grow in service due to loading or
environmental conditions.
The methodology and assumptions for API 579 level
II assessment procedure include flaw rectangulation,
multiple flaw containment, flaw recategorization,
stress categorization, stress linearization, reference
stress solution compendium, stress intensity factor
compendium, partial safety factors for primary stress,
Chon L. Tsai / Fitness-for-service assessment of in-service pressure components: petrochemical components 72
material toughness and flaw size, secondary and resid-
ual stress reduction factor when the reference secon-
dary or residual stress is greater than material’s yield
stress, and plasticity interaction factor for secondary or
residual stress.
Typical scenarios associated with the petrochemical
components requiring a FFS assessment include: gen-
eral metal loss, local metal loss, brittle fracture, pitting
corrosion, blisters and laminations, crack-like flaws
including HIC and SOHIC, and fire damage. In API
579, the Fracture Mechanics-based analysis is used
extensively for assessing the crack-like flaws. The
fracture assessment procedures are based on the con-
cept of the Failure Assessment Diagram (FAD),
which correlates the fracture toughness behaviors,
dubbed “toughness ratio, K
r
=(K
I
P
+ρK
IP
SR
)/K
mat
”, of
the pressure component to its rupture characteristics,
dubbed “load ratio, L
r
p

p
ref

ys
”, when overloading in
the unfractured net section exists. The primary refer-
ence stress, σ
p
ref
, is based on factored primary stress
distribution and factored flaw size. The reference stress
of secondary or residual stress, σ
SR
ref
, distribution is
based on secondary or residual stress and factored flaw
size, which is included only in the calculation of the
stress intensities. K
I
P
is the stress intensity based on
factored primary reference stress distribution. K
IP
SR
is
the stress intensity based on secondary or residual
stress from elastic-plastic finite element analysis and
factored flaw size. K
mat
is factored material toughness,
σ
ys
is material’s yield stress, and ρ is plasticity interac-
tion factor as function of the factored reference stresses
(i.e. primary and secondary or residual stresses). Fig. 1
shows ρ versus L
r
P
relations for various L
r
SR
values.
The relationship can be written as

SR
IP
P
I I p
r
K
K K
L

= ) ( ρ (1)

The interaction of both modes of failure is repre-
sented by the failure locus.

] 7 . 0 3 . 0 )[ 14 . 0 1 (
6
65 . 0
2 p
r
L p
r r
e L K

+ − = (2)

The assessment of a flaw generates a single point on
the FAD. If the point lies within the failure locus the
structure is considered safe. If the point lies outside the
failure locus, structure failure is possible.
In estimating the toughness ratio of welded pressure
components, the effect of welding-induced residual
stresses needs to be incorporated in the analysis, al-
though it is not required for the determination of the
load ratio due to its secondary nature of the internal
stresses associated with the crack growth behaviors.
The current API 579 procedure for assessing the
crack-like flaws in the welded pressure components
assumes either a constant residual value at the mini-
mum material’s yield stress level (i.e. yield stress plus
10 ksi) or a linear stress distribution through the wall
thickness containing the flaw with the maximum stress
at material’s yield stress level (i.e. yield stress plus 10
ksi). A plasticity interaction, dubbed “ρ-factor”, is
used in API 579 to account for the effect of residual
stresses when the primary loading is applied (Fig. 1).
Fig. 2 shows an example of the upper bound solu-
tion for the nonself-equilibrating residual stress esti-
mate.
The residual stress assumptions used in the current
API 579 are known to be conservative. This paper dis-
cusses the effect of the residual stress distributions on
the FAD when the actual residual stress distribution in
the weldment is incorporated in the fracture analysis.
Semi-elliptical surface cracks of shallow and deep as-
pect ratios in a welded plate were analyzed using the
finite element analysis (FEA) method. It was found
that the crack depth to thickness ratio influences the
magnitude of ρ-factor substantially for the
semi-elliptical surface cracks in the welded plate.
The actual residual stress distribution has more pro-
found effect on deep cracks than shallow cracks.















Fig. 1. API 579 Plasticity Interaction Factor, ρ as function of
L
r
P
.














Fig. 2. API 579 Upper Bound Solution (nonself-equilibrating
stress distribution).

L
r
P

ρ
Lr
SR
=0.5
Lr
SR
=1.0
Lr
SR
=1.5
Lr
SR
=2.0
Lr
SR
=2.5
Lr
SR
=3.0
Lr
SR
=3.5
Lr
SR
=4.0
Lr
SR
=4.5
Lr
SR
=5.0
L
r
SR
=0.5
L
r
SR
=5.0
x x
Transverse Residual Stress
Longitudinal
Seam Weld
Circumferential
Girth Weld
PIPE WELDS (t=1”)
r
ys
r
i
σ σ =
r
ys
r
o
σ σ =
r
ys
r
i
σ σ =
r
ys
r
o
σ σ 2 . 0 =
Thumb Nail
Toe Crack
Journal of China Pressure Vessel Technology 1 (2003) 71-79 73
2. Estimating residual stresses

In a welded joint, the expansion and contraction
forces act on the weld metal and its adjacent base metal.
As the weld metal solidifies and fuses with the base
metal, it is in its maximum expanded state. However,
at this point, the weld metal and its adjacent base metal
are at high temperatures and have little strength or
rigidity. The volume expansion causes local thickening
in the weld area, but is incapable of causing significant
amount of plastic strains in the cooler joint
neighborhoods. On cooling, it attempts to contract to
the volume it would normally occupy at the lower
temperature, but it is restrained from doing so by the
adjacent cooler base metal. Stresses develop within the
weld, finally reaching the yield strength of the weld
metal. At this point the weld stretches, or yields, and
thins out, thus adjusting to the volume requirements of
the lower temperature. But only those stresses that
exceed the yield strength of the weld metal are relieved
by this accommodation. By the time the weld reaches
room temperature, the weld will contain locked-in
tensile stresses (residual stresses) of yield magnitude
and the base metal away from the weld is usually in
compression with smaller magnitude. The internal
tensile and compressive forces are in equilibrium with
the joint deforming to comply with the strain
compatibility. The residual stress distributions and the
amount of weld distortion depend on the final state of
the plastic strain distributions and their compatibility in
the joint.
The welding-induced incompatible inelastic strains
in the weldment during the heating and cooling weld
cycles include transient thermal strains, cumulative
plastic strains, and the final inherent shrinkage strains.
At any instant during welding, the incompatible
thermal strains resulting from the nonlinear
temperature distributions generate the mechanical
strains, which lead to the incremental plastic strains in
the weldment if yielding occurs. The incremental
plastic strains accumulate over the periods of heating
and cooling. Upon completion of the welding cycles,
the cumulative plastic strains interact with the
weldment stiffness and the joint rigidity resulting in the
final state of residual stresses and distortion of the
weldment. This final state of the inelastic strains,
which are always compressive, is referred to as the
"inherent shrinkage strains."
Welding-induced, incompatible plastic strains
(assuming a 2-D plane-strain condition for illustration
purpose) at each heating or cooling time increment may
be described mathematically as follows:

)] , ( ) , ( [ ) (
1
) (
2 2
y x g y x g
E
y x
∆ + − ∇

− = + ∇ αθ
ν
σ σ (3)

where
2
∇ is the Laplacian operator, σ
x
and σ
y
are
thermal stress components in the respective x and y
directions, E is Young’s modulus, ν is Poisson’s ratio, α
is thermal expansion coefficient, θ is a temperature
function, g(x,y) is a cumulative plastic strain function,
and ∆g(x,y) is the plastic strain increment function over
each thermal loading step. The plastic strain functions
may be written as follows:

) (
1
2
1
) , (
2
2
2
2
2
2
2
2
p
y
p
x
p
xy
p
y
p
x
E
y x
x y
E
y x g
ε ε
ν
ν
ε ε
ε
ν
+ ∇


|
|
.
|

\
|
∂ ∂




+



=
(4)

) (
1
) (
2
) (
) (
1
) , (
2
2
2
2
2
2
2
2
p
y
p
x
p
xy
p
y
p
x
E
y x
x y
E
y x g
ε ε
ν
ν
ε ε
ε
ν
∆ + ∆ ∇


|
|
.
|

\
|
∂ ∂
∆ ∂


∆ ∂
+

∆ ∂

= ∆
(5)

The Laplacian thermal strains are governed by the
rate of enthalpy change in the weldment. Upon com-
pletion of the heating and cooling cycles the cumula-
tive plastic strains in the weldment are usually com-
pressive and become the inherent shrinkage strains,
g
I
(x,y), that interact with the structural rigidity to result
in residual stresses (σ
x
R
and σ
y
R
) and distortion (ε
x
D

and ε
y
D
). Residual stresses may be written with a pos-
sible reverse yielding, g
R
(x,y) as follows:

) , ( ) , ( ) (
2
y x g y x g
I R R
y
R
x
− = + + ∇ σ σ (6)

Distortion may be written in the form of final total
strains as follows:

( ) | |
PR
x
R
z
R
y
R
x
I
x
D
x
E
ε σ σ ν σ ε ε + + − = −
1
(7)

( ) | |
PR
y
R
x
R
z
R
y
I
y
D
y
E
ε σ σ ν σ ε ε + + − = −
1
(8)

( ) | |
t cons
E
PR
y
PR
x
R
y
R
x
R
z
I
y
I
x
D
z
tan
1
) (
= − −
+ − = + −
ε ε
σ σ ν σ ε ε ε
(9)

PR
xy
R
xy
I
xy
D
xy
E
γ τ
ν
γ γ +
+
= −
) 1 ( 2
(10)

where the superscripts: D represents distortion strains,
I represents the inherent (cumulative) shrinkage strains,
R represents residual stresses, and PR represents plas-
tic strains due to reverse yielding.
Chon L. Tsai / Fitness-for-service assessment of in-service pressure components: petrochemical components 74
With known distortion strains determined from the in-
herent shrinkage plastic strains, the distortion shape of a
weldment can be determined by integrating these strains
with respect to spatial coordinate variables.
Equations 3 through 10 demonstrate that the cumu-
lative plastic strains govern the final state of residual
stresses and weldment distortion. Therefore, an engi-
neering approach to estimating welding-induced resid-
ual stresses or distortion is to establish the relation-
ships between these plastic strains and variables asso-
ciated with welding process, joint design, and struc-
tural detail. Some physical phenomena occurs during
welding may not be described using these theoretical
equations, which effects can only be analyzed by the
numerical simulations. During the heating process all
the strains in the molten pool relax to the nil strain
state. Upon solidification the weld metal shrinks from
the melting temperature resulting in high shrinkage
stresses. The inherent shrinkage strains should include
these shrinkage strains. However, depending upon the
relative stiffness of the shrinkage zone and the elas-
tic-resistance zone in the base metal, only those inher-
ent strains within twice the yield magnitude are effec-
tive in causing the final state of residual stresses and
distortion.
The inherent shrinkage strains are primarily caused
by material softening and nonlinear thermal gradients
in the cooler areas. The inherent shrinkage strains are
uniform along the weld length, except in the areas of
arc start and stop. They are nearly uniform within the
softening area in the direction transverse to the weld.
The strain magnitude decreases at a steep slop to zero
within a short distance from the edges of the softening
zone. The peak temperature attained in the weldment
are uniquely related to the longitudinal inherent
shrinkage strains due to relatively large stiffness ratio
between the soften zone and its cooler surroundings.
The peak temperature distribution can be used to esti-
mate the longitudinal inherent shrinkage strains with
good accuracy. The longitudinal residual stresses and
the longitudinal cambering can therefore be determined
from the peak temperature distributions. Buckling can
also be predicted in large thin-plate weldments. The
peak temperatures alone are insufficient to determine
the transverse inherent shrinkage strains because of the
smaller stiffness ratio and its sensitivity to the joint
thickness and the external constraint conditions. Nev-
ertheless, with a correction procedure considering the
material incompressibility, the modified transverse
inherent shrinkage strains can be used to estimate with
good accuracy the transverse residual stresses and an-
gular distortion of the weldment.
To get the baseline through-thickness residual stress
distribution in a multiple-pass weld plate, Fig. 3 shows
the typical cross-section of a complete-joint-penetra-
tion groove weld. Eleven weld passes are deposited in
the sequence as indicated by the numerical numbers.
The welding condition is summarized in Table 1.
The FEA model using the ABAQUS [3]

software
program simulates a welding arc moving along the
center plane of a plate (ASTM A516 Grade 70 steel
with matching electrode, 6”x6”x1”). In this study,
temperature-dependent thermal and mechanical prop-
erties of the steel were incorporated in the numerical
analysis. Several physical phenomena associated with
the welding process, including moving arc source,
filler metal addition, and plastic strain relaxation due to
melting, were modeled. Results analyzed from this
modeling study included peak temperature, cumulative
plastic strains, and total strains.
Assuming two ellipsoidal functions for arc heat dis-
tribution ahead and behind the arc, respectively, the
heat input, calculated from products of arc efficiency,
welding current, and arc voltage, was applied to the
weld joint of the plate. The heat distribution was uni-
form through the plate thickness. Assuming that the
heat intensity along the boundary of the ellipsoids
equals to 5% of the maximum value at the arc center,
the generic ellipsoidal heat function may be written as
follows:

|
|
.
|

\
|

+
− −
=
76 . 5
3
76 . 5
) ( 3
76 . 5
6
) , , (
2
2
2
y St x
Exp
H
VI f
t y x q
β β π
η
&
(11)

where ) , , ( t y x q& is power density (W/in
3
); t is time
elapsed after arc initiation (second); x, y are global
coordinate axes with origin at the arc initiation center
(x is along weld axis and y is in the direction trans-
verse to weld) (in); H is plate thickness (in); I is aver-
age welding current (A); S is arc travel speed (in/s); V
is arc voltage (volt); f is weight factor (1.4 behind the
arc and 0.6 ahead the arc); η is arc efficiency (0.55); β
is the lag factor that determines the slop of the heat
distribution function in x-direction (14.0 behind the arc
and 1.0 ahead the arc). The 5.7 factor is a square of the
semi-axes of the ellipsoidal boundaries. The weight
and lag factors were calibrated with published experi-
mental data [3]. Fig. 4 shows the energy distraction of
the moving source (Eq. 11) A similar approach dubbed
“double ellipsoidal heat source model” was introduced
by Argyris et. al.[5] in 1982 and Goldak et. al.[6] in
1986.
Assuming an elliptical surface crack develops at the
weld toe in the mid section during service life of a
pressure component, Fig. 5 shows the through-thick-
ness residual stress distribution in the crack plane at a
direction transverse to this plane estimated from the
aforementioned process modeling analysis. The re-
sidual stress is tensile at the yield stress level near the
weld face due to weld shrinkage. It also shows tensile
in the root portion of the joint due to plate bending (i.e.
caused by offset between weld centroid and the neutral
axis of the plate thickness). To reach an equilibrium
Journal of China Pressure Vessel Technology 1 (2003) 71-79 75
condition of this internal stress distribution without any
external restraints, the residual stress in the middle
portion is compressive.



Fig.3. Weld cross-section details.



Table 1. Welding Condition
Pass
No.
Current
(ampere)
Voltage
(volt)
Speed
(in./min.)
1 190 25 7.9
2-3 215 26 11.1
4-5 250 27 11.1
6 190 25 7.9
7-9 220 26 11.1
10-11 250 27 11.1





Fig. 4. Energy distribution of a double-ellipsoidal moving heat
source.



















Fig. 5. Through thickness residual stress distribution in the crack
plane.
3. Fracture analysis

The fracture model used a quarter-joint model.
Three layers of 20-node isoperimetric brick elements
(C3D20R in ABAQUS) surrounded the crack tip re-
gion and the elements in the first layer were degener-
ated into triangular prisms along the crack front.
Mid-side nodes were moved to the quarter-point posi-
tion on the edges of the first layer elements to provide
a strain singularity and improve the modeling of the
strain field adjacent to the crack tip. Fig. 6 shows the
mesh pattern of the crack model. The mesh was also
modeled that the effect of the boundary conditions on
the far faces of the model was negligible to the solu-
tions. The far field tension was applied to the surface
crack (thumbnail crack) model.
In order to develop the baseline information regard-
ing the three-dimensional J-integral analysis, the elas-
tic-plastic FEA fracture analysis of the
three-dimensional thumbnail cracks of various depths
and crack aspect ratios in a plate having dimensions
7”x1.25”x0.5” were performed. No residual stress was
considered. Various stress levels were applied to the
cracked specimens. The J-integral values showed path
independency for all integration loops except the first
two-loops near the crack tip. Fig. 7 shows a compari-
son of stress intensity factor K
I
along the crack front
between the converted K
I
values from the numerically
calculated J-integral and the results reported by Raju
and Newman [7]. The crack aspect ratio is 0.2, crack to
depth ratio is 0.1 for this comparison.
In order to accommodate the basic property of the
path-independent J-integral considering the effect of
plastic deformations (i.e. residual stress) a FORTRAN
subroutine was developed in accordance with the nu-
merical procedure proposed by Simoto et al. [8], Aoki
el al. [9] and Shih et al. [10]. A sensitivity and accuracy
study was performed. Fig. 8 shows a plot of strain en-
ergy density release rate (J-integral) values for a
welded plate using the integration contours at different
distance from the crack tip. The primary load is 10 ksi
and the secondary stress is taken from Fig. 5. The
J-integral values of 27 psi-in and 45 psi-in were calcu-
lated for the cases with and without considering the
residual stress.


Fig. 6. Mesh pattern of a 3-D thumbnail crack model.
Transverse Residual Stress (ksi)
-50 -40 -30 -20 -10 0 10 20 30 40 50
1.0
0.8
0.6
0.4
0.2
0.0
48ksi
per API
D
i
s
t
a
n
c
e

f
r
o
m

B
o
t
t
o
m

(
i
n
c
h
)

Chon L. Tsai / Fitness-for-service assessment of in-service pressure components: petrochemical components 76











Fig. 7. Comparison of stress intensity factors between the 3-D
numerical analysis of a thumbnail crack with the Raju and
Newman solution [7].


Fig. 8. Path-independency of J-integration considering the effect
of residual stresses and numerical corrections[8-10].

4. Fitness-for-service assessments

The general failure assessment diagram (FAD) for
crack-like flaws provided by API 579 is defined by a
failure assessment curve, which is based on a theoreti-
cal analysis by Ainsworth [11] in 1986 for deriving the
plasticity correction factor (ρ-factor) using the concept
of a reference stress. It was assumed an infinite
cracked plate, tension loading, plane stress conditions
and perfectly plastic materials. The failure assessment
curve is expressed as a functional relationship between
the toughness ratio, K
r
and the load ratio, L
r
, in which
L
r
is defined as ratio of the reference stress considering
the primary stress only. The functional relationship
may be written as

ys
p
ref
r
r
s
I
p
I
mat
r
L where
L f K K
K
K
σ
σ
ψ
=
= − + =
..
) ( ] [
1
(12)

K
I
p
and K
I
s
are the stress intensities resulting from the
primary stress (i.e. applied stress) and the secondary
stress (i.e. residual stress), respectively. K
mat
is the fac-
ture toughness of material and σ
sy
is the yield stress of
material. σ
p
ref
is the reference stress due to the primary
stress, which is defined in API 579.
The load function is shown in Fig. 9 and may be
written as follows:

)] 65 . 0 exp( 7 . 0 3 . 0 )[ 14 . 0 1 ( ) (
6 2
r r r
L L L f − + − = (13)

Ainworth [11] used ψ instead of ρ representing the
plasticity correction factor, which is

(
(
(
(
(
¸
(

¸

+
+
+
− =
) (
) (
) (
ys
s
ref
s
ref
p
ref
s
ref
p
ref
yss
s
ref
p
ref
r
f
f
L f
σ
σ
σ
σ
σ σ
σ
σ σ
ψ (14)

Again, the superscripts “p” and “s” means primary and
secondary stresses, respectively. This correction func-
tion is generally considered conservative.
Hooton and Budden [12] in 1995 modified the Ani-
worth plasticity correction factor considering plastic
relaxation due to crack opening. This reduces the cor-
rection factor when high residual stress magnitude ex-
ists, but it decreases the correction factor for interme-
diate values of residual stresses.
In 1997, Pan and Li [13] modified the Ainworth
plasticity correction factor using the effective crack
length concept. This has been reported improving the
accuracy in the assessment of small flaws under high
tensile residual stresses.
In all three cases referenced above they assumed re-
sidual stresses being tensile at yield magnitude and
uniform through the thickness. This assumption is
generally believed to be the major source contributing
to the conservatism in the current FFS assessment pro-
cedure provided in API 579. This section will quantify
this conservatism by comparing the FEA fracture
analysis based on the true residual stress distributions
shown in Fig. 5 and the API 579 FFS assessment
guidelines based on the assumption of uniform residual
stress distribution through the joint thickness.
The 3-D FEA fracture analysis first determines the
plasticity correction factor due to the primary stress,
the secondary stress, and the combined stresses. The
difference in the toughness ratio (
IP
I
P
IP
P
I
K
K
or
K
K
, where
the subscripts “I” and “IP” mean elastic and elas-
tic-plastic analyses, respectively and the superscript
"P" means the stress intensity considering the primary
stress only), calculated with and without considering
the residual effect, defines the numerical plasticity cor-
rection factor, or modified plasticity interaction factor,
and dubbed “ρ
r
”. Fig. 10 shows a typical graphical
presentation of ρ
r
for a crack with properties as:
depth=0.1”, length=0.4”, thickness=1”. This defines
the numerical plasticity correction factor.
0 20 10 30 40 50 60 70 80 90
0
100
200
300
400
500
600
Crack Front Location (degrees)
S
t
r
e
s
s

I
n
t
e
n
s
i
t
y

F
a
c
t
o
r

(
l
b
/
i
n
2
-
i
n
1
/
2
)

Raju and Newman solution (1979)
ABAQUS 3-D crack modeling
Crack Depth Ratio, a/t=0.1
Crack Aspect Ratio, a/2c=0.2
Journal of China Pressure Vessel Technology 1 (2003) 71-79 77









Fig. 9. API 579 FAD assessment curve [1].












Fig. 10. Modified plasticity interaction factor, ρ
r
.

The stress intensity ratio shows the relative signifi-
cance of the plasticity effect on the calculation of the
stress intensity values. The plasticity interaction factor
shows the relative significance of the residual stress
effect. When considering the primary stress only, very
little plasticity is developed at low applied loads. The
plasticity increases the stress intensity when the ap-
plied load becomes high.
The residual stress alone does not result in much
plasticity effect on the stress intensity. However, when
residual stress is considered together with the applied
load, it quickly yields the crack region and increases
the stress intensity. This is responsible for the differ-
ence between the cases with or without considering the
residual stress, and hence the ρ
r
factor. As the magni-
tude of the applied load becomes more than 60% of
material's yield stress, the residual stress effect dimin-
ishes. When the applied load ratio is greater than 1.0,
the stress intensity calculated from the elastic analysis
overcomes the plasticity effect due to residual stress.
The plasticity caused by the combined stresses actually
increases the demand of plastic work instead of driving
the crack.
The numerical plasticity correction factor, ρ
r
may be
plotted against the load ratio, L
r
. Fig. 11 shows two
sets of plots for a shallow crack (depth/thickness ra-
tio=0.1) and a deeper crack (depth/thickness ratio=0.2),
respectively. Each plot shows the functional relation-
ship for three different crack aspect ratios
(depth/length ratios=0.05, 0.25 and 0.50). The residual
stress distribution has more significant effect on shal-
low cracks because of high tensile stress near the weld
face. For deeper cracks the sensitivity of crack geome-
try is relatively insignificant, which shows the residual
stress distribution has less effect on the plasticity in-
teraction due to reduced stress magnitude at the crack
front.
For the shallow crack (depth/thickness ratio=0.1),
the flat crack shape (wider crack) shows higher plastic-
ity interaction at lower load ratios. However, it de-
creases when the primary stress is greater than 40% of
material’s yield stress and becomes lower values than
the semi-circular crack when the load ratio is greater
than 0.6 or 0.7, depending upon the depth/length ratio.
When the crack length reduces for a given crack
depth (elliptical and circular) the significance of the
plasticity interaction shifts towards higher applied
loads. More strain energy from the applied load would
be required to develop the plasticity effect for shorter
cracks of a given depth.
With the ρ
r
factors determined for the crack condi-
tions interested, the stress intensity ratios may be cal-
culated with consideration of the applied load only,
which is then amended by this plasticity correction
factor as follows:

(1)
r
P
IP
P
I
IP
I
K
K
K
K
ρ − = for distributed residual
stress as shown in Fig. 5. (15)

(2)
579 API
P
IP
P
I
IP
I
K
K
K
K
ρ − = for uniform residual
stress as defined by API 579 (16)

(3)
P
IP
P
I
IP
I
K
K
K
K
= without the plasticity
correction (17)

Fig. 12 shows comparisons of the numerically de-
termined stress intensity ratios, with and without cor-
rections due to the plasticity interaction. The crack
conditions are: depth/length ratio=0.25 for both cases
and depth/thickness ratios=0.1 and 0.2, respectively.
The figures reflect the effect of plasticity interaction as
shown in Fig. 11. When the actual residual stress dis-
tribution is incorporated in the FFS analysis for a shal-
low crack, the effect of correction on plasticity interac-
tion is relatively insignificant in choosing the correc-
tion factors (API or FEA) because the residual stress at
the crack tip is high for shallow crack. The K
I
/K
IP

curve is the stress intensity ratio considering both the
applied and the residual stresses. For deeper crack, the
difference between the correction factors is shown be-
cause that API assumes high residual stress magnitude
throughout the thickness. In this figure, K
I
P
/K
IP
P
is
plotted and the corrected curves are those with the re-
spective ρ
r
subtracted from this stress intensity curve
of primary load only.
L
r

K
r

0.0 0.2 1.0 0.8 0.6 0.4
0.0
1.2 1.4 1.6 1.8 2.0 2.2
0.2
0.8
0.6
0.4
1.2
1.0
Unacceptable region
Acceptable region
Cut-off for steels with a yield
plateau
Cut-off for ASTM A508
Cut-off for C-Mn steels
Cut-off for stainless steels
0.0 0.2 0.4 0.6 0.7 1.0
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
Load Factor Due to Primary Load,
P
S
t
r
e
s
s

I
n
t
e
n
s
i
t
y

R
a
t
i
o

p
IP
P
I
K
K
IP
I
K
K
a/t=0.1 and a/2c=0.25
Elastic and Elastic-Plastic
FEA Crack Analyses
1.2
Chon L. Tsai / Fitness-for-service assessment of in-service pressure components: petrochemical components 78
When the crack-driving force is to be calculated us-
ing the API 579 assessment procedure, the plasticity
correction factor based on the assumption of an uni-
form residual stress distribution through the joint
thickness is added to the standard calculation of the
toughness ratio. The numerically determined failure
driving forces considering the residual stress effect are
smaller than calculated based on the API procedure.
Fig. 13 shows the driving force comparisons consid-
ering the effect of uniform residual stress and actual
residual stress for crack depth/length ratio=0.25 and
crack depth/thickness ratio= 0.1, 0.2, 0.3 and 0.4. Al-
though the parametric analyses (i.e. results not pre-
sented in this paper) show actual residual distribution
reduces the crack driving force as compared to the API
579 assessment procedure, the degree of reduction de-
pends on the crack geometric aspect ratio and the crack
depth/joint thickness ratio.










(a) depth/thickness ratio=0.1









(b) depth/thickness ratio=0.2
Fig. 11. Effect of crack dimensions on modified plasticity inter-
action factor, ρ
r
as function of load ratio obtained by FEA frac-
ture analysis.














(a) K
I
/K
IP
Vs. L
r
(a/t=0.1)















(b) K
I
/K
IP
Vs. L
r
(a/t=0.2)

Fig. 12. Stress intensity ratios with or without corrections.












Fig. 13. Comparison of failure driving forces: API 579 proce-
dure and from the FEA analyses. Crack depth/length ratio=0.25.

5. Concluding remarks

The current API 579 assessment procedure for
crack-like flaws in weldments assumes an uniform
residual stress at the level of yield stress plus 10 ksi.
This assumption may result in significant conservatism
in calculating the toughness ratio. This paper demon-
strates that using the actual residual stress distribution
will reduces the failure driving force as a crack grows
deeper due to the fact that welding-induced residual
stress is compressive in the middle thickness of the
joint. For failure assessment of critical weld joints,
actual residual stress distribution is an important factor
to be incorporated in the FFS assessment.
In the general FFS assessment procedure, the frac-
ture parameter for calculating the toughness ratio may
use (1) stress intensity factor (K); (2) stain energy re-
lease rate (G or J); (3) strain energy density factor (S);
or crack-tip opening displacement (CTOD).

Acknowledgements

This paper was written based on the research data
obtained during a Ph.D. dissertation study of Dr. Beom
No Lee, who was an advisee of the author of this paper.
The dissertation title is “Effect of Residual Stresses on
Load Factor Due to Primary Load, L
r
P

a/t=0.1
Shallow
a/2c=0.05
a/2c=0.25
a/2c=0.50
flat crack
circular crack
elliptical crack
ρ
r

a/2c=0.05
a/2c=0.25
a/2c=0.50
flat crack
circular crack
elliptical crack
ρ
r

Load Factor Due to Primary Load, L
r
P

a/t=0.2
Crack Depth Increases
S
t
r
e
s
s

I
n
t
e
n
s
i
t
y

R
a
t
i
o

Load Factor Due to Primary Load, L
r
P
0.0 0.2 0.4 0.6 0.8 1.0 1.2
a/t=0.1 & a/2c=0.25
Shallow Crack
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
General FAD
IP
I
K
K
579 API P
IP
P
I
K
K
ρ − r P
IP
P
I
K
K
ρ −
1.4
Load Factor Due to Primary Load, L
r
P
0.0 0.2 0.4 0.6 0.8 1.0 1.2
S
t
r
e
s
s

I
n
t
e
n
s
i
t
y

R
a
t
i
o

a/t=0.2 & a/2c=0.25
Crack Depth Increases
0.0
0.2
0.4
0.6
0.8
1.0
1.2
General FAD
P
IP
P
I
K
K
) 579 ( API r P
IP
P
I
K
K
ρ −
r P
IP
P
I
K
K
ρ −
Load Factor Due to Primary Load, L
r
P

0.0 0.2 0.4 0.6 0.8 1.0 1.2
T
o
u
g
h
n
e
s
s

R
a
t
i
o
,

K
r

a/2c=0.25
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0.4
a/t=0.4
a/t=0.2
0.3
0.2
0.1
0.1
0.3
F FA AD D E En nv ve el lo op p
Journal of China Pressure Vessel Technology 1 (2003) 71-79 79
Fracture Behavior of Weldments,” by Beam No Lee,
The Ohio State University (2001) (Advisor: Chon L.
Tsai)

References

[1] API RP-579, Section 9: Assessment of Crack-Like Flaws,
Recommended Practice for Fitness-for-Service, 2000.
[2] BS PD 6493 (1991)/BS 7910 (1999), Guidance on Methods
for Assessing the Acceptability of Flaws in Fusion Welded
Structures.
[3] ABAQUS, commercial FEA software program by Hibbitt,
Karlsson & Sorensen, Inc.
[4] Feng ZL, Cheng WT, Chen YS. Development of New
Modeling Procedures for 3D Welding Residual Stress and
Distortion Assessment. EWI CRP Report SR9818, No-
vember 1998.
[5] Argyris JH, Szimmat J, William KJ. Computational As-
pects of Welding Stress Analysis, Computer Methods in
Applied Mechanics and Engineering 1982; 33: 635-666.
[6] Goldak J, Bibby M, Moore J, House R, Patel B. Computer
Modeling of Heat Flow in Welds Model for Welding Heat
Source. Metallurgical Transactions B 1986; 17B: 587-600.
[7] Newman Jr. JC, Raju IS. Stress Intensity Factor Equations
for Cracks in Three-dimensional Finite Bodies Subject to
Tension and Bending Loads. NASA Technical Memoran-
dum 85793, April 1984.
[8] Simoto K, Sakata M. On the Path Independent Integral-J.
Eng Frac Mech 1980; 13: 841-850.
[9] Aoki S, Kishimoto K, Sakata M. Elastic-Plastic Analysis of
Crack in Thermally-Loaded Structures. Eng Frac Mech
1982; 16(3): 405-413.
[10] Shih CF, Moran B, Nakamura T. Energy Release Rate
Along a Three-Dimensional Crack Front in a Thermally
Stressed Body. Int J Frac 1986; 30: 79-102.
[11] Ainsworth RA. The Assessment of Defects in Structures of
Strain Hardening Materials. Eng Frac Mech 1984; 19:
633-642.
[12] Hooton DG, Budden PJ. R6 Developments in the Treat-
ment of Secondary Stresses. ASME BVP Conference, Fa-
tigue and Fracture Mechanics in Pressure Vessel and Pip-
ing 1995; 304: 503-509.
[13] Pan HL, Li PN. A Procedure for Treating Secondary Stress
in Failure Assessment Diagram. Eng Frac Mech 1997;
58(1): 149-159.



0 ρ LrSR=0. LrSR=5.0 LrSR=0. is based on factored primary stress distribution and factored flaw size. Semi-elliptical surface cracks of shallow and deep aspect ratios in a welded plate were analyzed using the finite element analysis (FEA) method. σSRref. 1 shows ρ versus LrP relations for various LrSR values. and fire damage.0 LrSR=2. the effect of welding-induced residual stresses needs to be incorporated in the analysis.14 Lrp )[0. Tsai / Fitness-for-service assessment of in-service pressure components: petrochemical components material toughness and flaw size. 1. which is included only in the calculation of the stress intensities. The fracture assessment procedures are based on the concept of the Failure Assessment Diagram (FAD).3 + 0.2σ ys The assessment of a flaw generates a single point on the FAD.7e −0. local metal loss. The relationship can be written as assumes either a constant residual value at the minimum material’s yield stress level (i. and plasticity interaction factor for secondary or residual stress. Fig. Kr=(KIP+ρKIPSR)/Kmat”. is used in API 579 to account for the effect of residual stresses when the primary loading is applied (Fig.5 LrSR=3. pitting corrosion. If the point lies within the failure locus the structure is considered safe. The residual stress assumptions used in the current API 579 are known to be conservative. and ρ is plasticity interaction factor as function of the factored reference stresses (i.e. brittle fracture. although it is not required for the determination of the load ratio due to its secondary nature of the internal stresses associated with the crack growth behaviors. distribution is based on secondary or residual stress and factored flaw size. API 579 Plasticity Interaction Factor. In API 579.0 LrSR=4.5 LrSR=4. The primary reference stress. KIP is the stress intensity based on factored primary reference stress distribution. primary and secondary or residual stresses).0 LrSR=1. The reference stress of secondary or residual stress. crack-like flaws including HIC and SOHIC. 2. Fig. The actual residual stress distribution has more profound effect on deep cracks than shallow cracks. This paper discusses the effect of the residual stress distributions on the FAD when the actual residual stress distribution in the weldment is incorporated in the fracture analysis. If the point lies outside the failure locus. σys is material’s yield stress. 2 shows an example of the upper bound solution for the nonself-equilibrating residual stress estimate. when overloading in the unfractured net section exists.5 LrSR=1. of the pressure component to its rupture characteristics.5 LrSR=2.e. ρ as function of LrP. K r = (1 − 0. σpref. yield stress plus 10 ksi). KIPSR is the stress intensity based on secondary or residual stress from elastic-plastic finite element analysis and factored flaw size. It was found that the crack depth to thickness ratio influences the magnitude of ρ-factor substantially for the semi-elliptical surface cracks in the welded plate. the Fracture Mechanics-based analysis is used extensively for assessing the crack-like flaws. Fig.65 Lr ] 2 p6 (2) PIPE WELDS (t=1”) x x r σ or = σ ys r σ or = 0. The current API 579 procedure for assessing the crack-like flaws in the welded pressure components r σ ir = σ ys r σ ir = σ ys Thumb Nail Toe Crack Transverse Residual Stress Circumferential Girth Weld Longitudinal Seam Weld Fig.0 LrSR=3.e. . 1). Typical scenarios associated with the petrochemical components requiring a FFS assessment include: general metal loss.72 Chon L. dubbed “load ratio. API 579 Upper Bound Solution (nonself-equilibrating stress distribution). blisters and laminations.5 K I − K IP ρ (L ) = SR K IP p r (1) LrP The interaction of both modes of failure is represented by the failure locus. Lrp=σpref/σys”. yield stress plus 10 ksi) or a linear stress distribution through the wall thickness containing the flaw with the maximum stress at material’s yield stress level (i. In estimating the toughness ratio of welded pressure components. which correlates the fracture toughness behaviors. structure failure is possible. dubbed “ρ-factor”. Kmat is factored material toughness. secondary and residual stress reduction factor when the reference secondary or residual stress is greater than material’s yield stress.5 LrSR=5. dubbed “toughness ratio. A plasticity interaction.

But only those stresses that exceed the yield strength of the weld metal are relieved by this accommodation. but is incapable of causing significant amount of plastic strains in the cooler joint neighborhoods. The welding-induced incompatible inelastic strains in the weldment during the heating and cooling weld cycles include transient thermal strains. θ is a temperature function. Upon completion of the welding cycles. y) = − g I ( x. the cumulative plastic strains interact with the weldment stiffness and the joint rigidity resulting in the final state of residual stresses and distortion of the weldment. Residual stresses may be written with a possible reverse yielding. Estimating residual stresses In a welded joint. Upon completion of the heating and cooling cycles the cumulative plastic strains in the weldment are usually compressive and become the inherent shrinkage strains. This final state of the inelastic strains. As the weld metal solidifies and fuses with the base metal.y) as follows: R ∇ 2 (σ xR + σ y ) + g R ( x. it is in its maximum expanded state.y) is the plastic strain increment function over each thermal loading step. y) + ∆g ( x. the weld metal and its adjacent base metal are at high temperatures and have little strength or rigidity. α is thermal expansion coefficient. . The internal tensile and compressive forces are in equilibrium with the joint deforming to comply with the strain compatibility. The plastic strain functions may be written as follows: g ( x. thus adjusting to the volume requirements of the lower temperature. that interact with the structural rigidity to result in residual stresses (σxR and σyR) and distortion (εxD and εyD). cumulative plastic strains. The incremental plastic strains accumulate over the periods of heating and cooling. gR(x.y) is a cumulative plastic strain function. y) (6) Distortion may be written in the form of final total strains as follows: ε xD − ε xI = D I εy −εy = 1 R R σ x − ν σ y + σ zR + ε xPR E [ ( )] (7) 1 R PR σ y − ν σ zR + σ xR + ε y E [ ( )] (8) I ε zD − (ε xI + ε y ) = −ε PR x −ε PR y 1 R R σ z −ν σ xR + σ y E = cons tan t [ ( )] (9) D I γ xy − γ xy = 2(1 + ν ) R PR τ xy + γ xy E (10) (3) where ∇ 2 is the Laplacian operator. σx and σy are where the superscripts: D represents distortion strains. at this point. and PR represents plastic strains due to reverse yielding.Journal of China Pressure Vessel Technology 1 (2003) 71-79 73 2. which are always compressive. incompatible plastic strains (assuming a 2-D plane-strain condition for illustration purpose) at each heating or cooling time increment may be described mathematically as follows: ∇ 2 (σ x + σ y ) = − E ∇ 2 (αθ ) − [ g ( x.y). The residual stress distributions and the amount of weld distortion depend on the final state of the plastic strain distributions and their compatibility in the joint. At this point the weld stretches. I represents the inherent (cumulative) shrinkage strains. ν is Poisson’s ratio. Stresses develop within the weld." Welding-induced. On cooling. The volume expansion causes local thickening in the weld area. At any instant during welding. and ∆g(x. g(x. but it is restrained from doing so by the adjacent cooler base metal. y ) = E 1 −ν 2 p p  ∂ 2ε p ∂ 2ε y ∂ 2 ε xy x  + −2  ∂y 2 ∂x∂y ∂x 2      νE p − ∇ 2 (ε xp + ε y ) 1 −ν 2 ∆g(x. E is Young’s modulus. gI(x. is referred to as the "inherent shrinkage strains. or yields. finally reaching the yield strength of the weld metal. and the final inherent shrinkage strains. y)] 1 −ν thermal stress components in the respective x and y directions. it attempts to contract to the volume it would normally occupy at the lower temperature. By the time the weld reaches room temperature. However. y) = 2 p p 2 p ∂ 2 (∆ε xy )  E  ∂ (∆ε x ) ∂ (∆ε y )   −2 + 2  2 2 ∂x∂y  1−ν  ∂y ∂x  (4) νE 2 p − ∇ (∆ε xp + ∆ε y ) 1−ν 2 (5) The Laplacian thermal strains are governed by the rate of enthalpy change in the weldment. R represents residual stresses. and thins out. which lead to the incremental plastic strains in the weldment if yielding occurs. the weld will contain locked-in tensile stresses (residual stresses) of yield magnitude and the base metal away from the weld is usually in compression with smaller magnitude. the expansion and contraction forces act on the weld metal and its adjacent base metal. the incompatible thermal strains resulting from the nonlinear temperature distributions generate the mechanical strains.

76β 2 5.55). and structural detail.penetration groove weld. including moving arc source.[6] in 1986. filler metal addition. Equations 3 through 10 demonstrate that the cumulative plastic strains govern the final state of residual stresses and weldment distortion. Assuming that the heat intensity along the boundary of the ellipsoids equals to 5% of the maximum value at the arc center. Nevertheless. Upon solidification the weld metal shrinks from the melting temperature resulting in high shrinkage stresses. temperature-dependent thermal and mechanical properties of the steel were incorporated in the numerical analysis. The longitudinal residual stresses and the longitudinal cambering can therefore be determined from the peak temperature distributions. To reach an equilibrium . 5 shows the through.e. and plastic strain relaxation due to melting. The inherent shrinkage strains should include these shrinkage strains. depending upon the relative stiffness of the shrinkage zone and the elastic-resistance zone in the base metal. It also shows tensile in the root portion of the joint due to plate bending (i. Several physical phenomena associated with the welding process. 6”x6”x1”). an engineering approach to estimating welding-induced residual stresses or distortion is to establish the relationships between these plastic strains and variables associated with welding process. They are nearly uniform within the softening area in the direction transverse to the weld. the heat input. V is arc voltage (volt). The welding condition is summarized in Table 1. t is time elapsed after arc initiation (second). were modeled. The inherent shrinkage strains are primarily caused by material softening and nonlinear thermal gradients in the cooler areas. the modified transverse inherent shrinkage strains can be used to estimate with good accuracy the transverse residual stresses and angular distortion of the weldment. the generic ellipsoidal heat function may be written as follows: & q( x. except in the areas of arc start and stop. joint design.7 factor is a square of the semi-axes of the ellipsoidal boundaries. y are global coordinate axes with origin at the arc initiation center (x is along weld axis and y is in the direction transverse to weld) (in). During the heating process all the strains in the molten pool relax to the nil strain state. Therefore. caused by offset between weld centroid and the neutral axis of the plate thickness).74 Chon L. Fig. β is the lag factor that determines the slop of the heat distribution function in x-direction (14. H is plate thickness (in). η is arc efficiency (0. the distortion shape of a weldment can be determined by integrating these strains with respect to spatial coordinate variables. To get the baseline through-thickness residual stress distribution in a multiple-pass weld plate. S is arc travel speed (in/s). respectively. Assuming an elliptical surface crack develops at the weld toe in the mid section during service life of a pressure component.0 ahead the arc). Buckling can also be predicted in large thin-plate weldments. The peak temperature attained in the weldment are uniquely related to the longitudinal inherent shrinkage strains due to relatively large stiffness ratio between the soften zone and its cooler surroundings. was applied to the weld joint of the plate. t ) =  − 3( x − St)2 − 3 y 2  6 fηVI  + Exp  5. with a correction procedure considering the material incompressibility. The residual stress is tensile at the yield stress level near the weld face due to weld shrinkage. y.thickness residual stress distribution in the crack plane at a direction transverse to this plane estimated from the aforementioned process modeling analysis. The strain magnitude decreases at a steep slop to zero within a short distance from the edges of the softening zone. t ) is power density (W/in3). Fig. x. which effects can only be analyzed by the numerical simulations. welding current.[5] in 1982 and Goldak et. Some physical phenomena occurs during welding may not be described using these theoretical equations. The 5. 4 shows the energy distraction of the moving source (Eq. 3 shows the typical cross-section of a complete-joint. I is average welding current (A). and total strains. cumulative plastic strains. calculated from products of arc efficiency. al. y. and arc voltage. The FEA model using the ABAQUS [3] software program simulates a welding arc moving along the center plane of a plate (ASTM A516 Grade 70 steel with matching electrode. However. Eleven weld passes are deposited in the sequence as indicated by the numerical numbers. Fig. The peak temperatures alone are insufficient to determine the transverse inherent shrinkage strains because of the smaller stiffness ratio and its sensitivity to the joint thickness and the external constraint conditions. f is weight factor (1. 11) A similar approach dubbed “double ellipsoidal heat source model” was introduced by Argyris et. The weight and lag factors were calibrated with published experimental data [3]. Tsai / Fitness-for-service assessment of in-service pressure components: petrochemical components With known distortion strains determined from the inherent shrinkage plastic strains. Assuming two ellipsoidal functions for arc heat distribution ahead and behind the arc.76    (11) & where q( x. only those inherent strains within twice the yield magnitude are effective in causing the final state of residual stresses and distortion. The peak temperature distribution can be used to estimate the longitudinal inherent shrinkage strains with good accuracy. The heat distribution was uniform through the plate thickness. Results analyzed from this modeling study included peak temperature.0 behind the arc and 1. The inherent shrinkage strains are uniform along the weld length.76πHβ 5.4 behind the arc and 0.6 ahead the arc). In this study. al.

A sensitivity and accuracy study was performed. The primary load is 10 ksi and the secondary stress is taken from Fig. 8 shows a plot of strain energy density release rate (J-integral) values for a welded plate using the integration contours at different distance from the crack tip. The mesh was also modeled that the effect of the boundary conditions on the far faces of the model was negligible to the solutions. In order to accommodate the basic property of the path-independent J-integral considering the effect of plastic deformations (i. Three layers of 20-node isoperimetric brick elements (C3D20R in ABAQUS) surrounded the crack tip region and the elements in the first layer were degenerated into triangular prisms along the crack front.25”x0. 5.e. Energy distribution of a double-ellipsoidal moving heat source.9 11.) 7. Fracture analysis The fracture model used a quarter-joint model.2 0. The far field tension was applied to the surface crack (thumbnail crack) model./min. Aoki el al.8 0.0 0. 1 2-3 4-5 6 7-9 10-11 Table 1. 3.Journal of China Pressure Vessel Technology 1 (2003) 71-79 75 condition of this internal stress distribution without any external restraints. Mid-side nodes were moved to the quarter-point position on the edges of the first layer elements to provide a strain singularity and improve the modeling of the strain field adjacent to the crack tip. [10].2. Welding Condition Current Voltage (ampere) (volt) 190 25 215 26 250 27 190 25 220 26 250 27 Speed (in. Pass No.1 11. In order to develop the baseline information regarding the three-dimensional J-integral analysis.1 11. 4. Fig. [9] and Shih et al. . The crack aspect ratio is 0. No residual stress was considered.4 0.1 7. Fig. Mesh pattern of a 3-D thumbnail crack model. Fig. Fig.5” were performed.1 for this comparison. [8]. Through thickness residual stress distribution in the crack plane. Distance from Bottom (inch) 1. Weld cross-section details. The J-integral values of 27 psi-in and 45 psi-in were calculated for the cases with and without considering the residual stress. residual stress) a FORTRAN subroutine was developed in accordance with the numerical procedure proposed by Simoto et al. crack to depth ratio is 0. the elastic-plastic FEA fracture analysis of the three-dimensional thumbnail cracks of various depths and crack aspect ratios in a plate having dimensions 7”x1. 6. Fig.0 -50 -40 -30 -20 -10 0 10 20 48ksi per API 30 40 50 Transverse Residual Stress (ksi) Fig.1 Fig. 5.9 11.6 0.3. The J-integral values showed path independency for all integration loops except the first two-loops near the crack tip. 6 shows the mesh pattern of the crack model. 7 shows a comparison of stress intensity factor KI along the crack front between the converted KI values from the numerically calculated J-integral and the results reported by Raju and Newman [7]. the residual stress in the middle portion is compressive. Various stress levels were applied to the cracked specimens.

10 shows a typical graphical presentation of ρr for a crack with properties as: depth=0.e. The failure assessment curve is expressed as a functional relationship between the toughness ratio. In 1997. It was assumed an infinite cracked plate.1”. Fitness-for-service assessments The general failure assessment diagram (FAD) for crack-like flaws provided by API 579 is defined by a failure assessment curve. plane stress conditions and perfectly plastic materials. The difference in the toughness ratio ( Kr = 1 [ K Ip + K Is ] − ψ = f ( Lr ) K mat p σ ref where. defines the numerical plasticity correction factor.. residual stress).14L2 )[0. respectively. ψ = f (Lr ) − σ yss p s σ ref + σ ref  )  s σ ref   p σ ref + σ s  ref   ) f(  σ ys    (14) Fig. This reduces the correction factor when high residual stress magnitude exists. which is defined in API 579.7 exp(−0. which is 90 100 0 0 10 20 30 40 50 60 70 80 Crack Front Location (degrees) f( p s σ ref + σ ref  Fig. but it decreases the correction factor for intermediate values of residual stresses. and dubbed “ρr”. In all three cases referenced above they assumed residual stresses being tensile at yield magnitude and uniform through the thickness. Kmat is the facture toughness of material and σsy is the yield stress of material. Lr. Pan and Li [13] modified the Ainworth plasticity correction factor using the effective crack length concept. where P K IP K IP KIp and KIs are the stress intensities resulting from the primary stress (i. applied stress) and the secondary stress (i. . which is based on a theoretical analysis by Ainsworth [11] in 1986 for deriving the plasticity correction factor (ρ-factor) using the concept of a reference stress. a/t=0.4”. the secondary stress. This defines the numerical plasticity correction factor. This correction function is generally considered conservative. This has been reported improving the accuracy in the assessment of small flaws under high tensile residual stresses. 8.1 Crack Aspect Ratio. length=0. calculated with and without considering the residual effect. 9 and may be written as follows: 400 f (Lr ) = (1− 0. or modified plasticity interaction factor. respectively and the superscript "P" means the stress intensity considering the primary stress only). and the combined stresses. Path-independency of J-integration considering the effect of residual stresses and numerical corrections[8-10]. thickness=1”. 4. Fig.3 + 0.65L6 )] r r Raju and Newman solution (1979) ABAQUS 3-D crack modeling (13) 300 200 Ainworth [11] used ψ instead of ρ representing the plasticity correction factor.2 Stress Intensity Factor (lb/in2-in1/2) 500 The load function is shown in Fig. in which Lr is defined as ratio of the reference stress considering the primary stress only. the subscripts “I” and “IP” mean elastic and elastic-plastic analyses. 5 and the API 579 FFS assessment guidelines based on the assumption of uniform residual stress distribution through the joint thickness. This assumption is generally believed to be the major source contributing to the conservatism in the current FFS assessment procedure provided in API 579. This section will quantify this conservatism by comparing the FEA fracture analysis based on the true residual stress distributions shown in Fig. respectively. σpref is the reference stress due to the primary stress. The functional relationship may be written as Again. tension loading.76 600 Chon L. a/2c=0. 7. The 3-D FEA fracture analysis first determines the plasticity correction factor due to the primary stress. Tsai / Fitness-for-service assessment of in-service pressure components: petrochemical components Crack Depth Ratio. Comparison of stress intensity factors between the 3-D numerical analysis of a thumbnail crack with the Raju and Newman solution [7].e. the superscripts “p” and “s” means primary and secondary stresses.Lr = σ ys (12) K IP K or I . Hooton and Budden [12] in 1995 modified the Aniworth plasticity correction factor considering plastic relaxation due to crack opening. Kr and the load ratio.

6 0. 10. For deeper cracks the sensitivity of crack geometry is relatively insignificant.1 and a/2c=0.4 a/t=0. KIP/KIPP is plotted and the corrected curves are those with the respective ρr subtracted from this stress intensity curve of primary load only.2 1.2 Lr Fig.0 0. When the crack length reduces for a given crack depth (elliptical and circular) the significance of the plasticity interaction shifts towards higher applied loads.2 1.6 0. and hence the ρr factor. 5. the stress intensity calculated from the elastic analysis overcomes the plasticity effect due to residual stress. 11.0.2 0. the difference between the correction factors is shown because that API assumes high residual stress magnitude throughout the thickness. The plasticity caused by the combined stresses actually increases the demand of plastic work instead of driving the crack.0 Acceptable region 0. When the actual residual stress distribution is incorporated in the FFS analysis for a shallow crack. 0. . respectively. which shows the residual KI KP = IP − ρ r for distributed residual K IP K IP stress as shown in Fig. 11 shows two sets of plots for a shallow crack (depth/thickness ratio=0. With the ρr factors determined for the crack conditions interested. The residual stress distribution has more significant effect on shallow cracks because of high tensile stress near the weld face. the effect of correction on plasticity interaction is relatively insignificant in choosing the correction factors (API or FEA) because the residual stress at the crack tip is high for shallow crack. Lr.6 or 0. 1. with and without corrections due to the plasticity interaction.6 1.7. For deeper crack.0 K IP p K IP KI K IP Elastic and Elastic-Plastic FEA Crack Analyses 0. (15) (2) KI KP = IP − ρ API 579 for uniform residual K IP K IP stress as defined by API 579 (16) (3) KI KP = IP K IP K IP correction without the plasticity (17) Fig. 9.2 0. However. it decreases when the primary stress is greater than 40% of material’s yield stress and becomes lower values than the semi-circular crack when the load ratio is greater than 0. depending upon the depth/length ratio. The stress intensity ratio shows the relative significance of the plasticity effect on the calculation of the stress intensity values.8 2.8 77 Unacceptable region Cut-off for steels with a yield plateau Cut-off for ASTM A508 Cut-off for C-Mn steels Cut-off for stainless steels Kr 0.05. stress distribution has less effect on the plasticity interaction due to reduced stress magnitude at the crack front. 12 shows comparisons of the numerically determined stress intensity ratios. This is responsible for the difference between the cases with or without considering the residual stress. The residual stress alone does not result in much plasticity effect on the stress intensity.0 2.2 Load Factor Due to Primary Load. The crack conditions are: depth/length ratio=0. When considering the primary stress only. it quickly yields the crack region and increases the stress intensity. very little plasticity is developed at low applied loads.0 0.6 0. The plasticity increases the stress intensity when the applied load becomes high.8 1. The plasticity interaction factor shows the relative significance of the residual stress effect.0 0.8 0. More strain energy from the applied load would be required to develop the plasticity effect for shorter cracks of a given depth. When the applied load ratio is greater than 1.1) and a deeper crack (depth/thickness ratio=0. The numerical plasticity correction factor. The KI/KIP curve is the stress intensity ratio considering both the applied and the residual stresses. ρr may be plotted against the load ratio.6 0.50).4 0.4 1. Fig.2). As the magnitude of the applied load becomes more than 60% of material's yield stress.0 1. when residual stress is considered together with the applied load.0 1.Journal of China Pressure Vessel Technology 1 (2003) 71-79 1.1 and 0. respectively. Each plot shows the functional relationship for three different crack aspect ratios (depth/length ratios=0. the stress intensity ratios may be calculated with consideration of the applied load only. Modified plasticity interaction factor. which is then amended by this plasticity correction factor as follows: (1) Fig. the flat crack shape (wider crack) shows higher plasticity interaction at lower load ratios. For the shallow crack (depth/thickness ratio=0. the residual stress effect diminishes. ρr.1).2 Stress Intensity Ratio 1. In this figure. However.25 and 0.2 0.25 1.4 0.4 0.2 P 0.4 0. API 579 FAD assessment curve [1]. The figures reflect the effect of plasticity interaction as shown in Fig.2.25 for both cases and depth/thickness ratios=0.0 0.7 1.

LrP (b) depth/thickness ratio=0.4 0.25 a/2c=0.0 FAD Envelop 0. who was an advisee of the author of this paper. (3) strain energy density factor (S). 1.0 1. Fig. The dissertation title is “Effect of Residual Stresses on .4. or crack-tip opening displacement (CTOD). Stress intensity ratios with or without corrections. Comparison of failure driving forces: API 579 procedure and from the FEA analyses.2 Crack Depth Increases Fig.4 0. Kr 1. Crack depth/length ratio=0. (2) stain energy release rate (G or J).1 & a/2c=0. the degree of reduction depends on the crack geometric aspect ratio and the crack depth/joint thickness ratio.2 0.6 0.6 0. This paper demonstrates that using the actual residual stress distribution will reduces the failure driving force as a crack grows deeper due to the fact that welding-induced residual stress is compressive in the middle thickness of the joint. results not presented in this paper) show actual residual distribution reduces the crack driving force as compared to the API 579 assessment procedure.25 flat crack Toughness Ratio.8 1. 13.2 0. 11.4 Stress Intensity Ratio 1.1 0.3 elliptical crack circular crack 1. LrP Load Factor Due to Primary Load.4 0.2 0.05 a/2c=0.1 0. circular crack flat crack elliptical crack 5. For failure assessment of critical weld joints. Beom No Lee.4 a/2c=0.4 0.2 0. ρr as function of load ratio obtained by FEA fracture analysis.2 1.0 0. the fracture parameter for calculating the toughness ratio may use (1) stress intensity factor (K). a/t=0.2 Load Factor Due to Primary Load.25 and crack depth/thickness ratio= 0.1) This paper was written based on the research data obtained during a Ph. Lr (a/t=0.2 Load Factor Due to Primary Load.2 a/t=0.2 a/t=0.05 a/2c=0.4 0.6 0. This assumption may result in significant conservatism in calculating the toughness ratio.0 0.D.8 1.25. In the general FFS assessment procedure. Tsai / Fitness-for-service assessment of in-service pressure components: petrochemical components When the crack-driving force is to be calculated using the API 579 assessment procedure. actual residual stress distribution is an important factor to be incorporated in the FFS assessment.25 Crack Depth Increases Stress Intensity Ratio K IP − ρr P K IP 1.4 1.8 0.2) Fig.3 and 0.2 Fig. 0. 1. dissertation study of Dr.6 0.2 0. the plasticity correction factor based on the assumption of an uniform residual stress distribution through the joint thickness is added to the standard calculation of the toughness ratio.2 Load Factor Due to Primary Load. The numerically determined failure driving forces considering the residual stress effect are smaller than calculated based on the API procedure.6 0.4 0. L (a) KI/KIP Vs.8 0.0 0.78 Chon L.0 K IP P K IP K IP − ρ r ( API 579 ) P K IP General FAD 0.2 & a/2c=0.3 0.0 P r 1.0 0.50 Load Factor Due to Primary Load.6 0. Concluding remarks The current API 579 assessment procedure for crack-like flaws in weldments assumes an uniform residual stress at the level of yield stress plus 10 ksi.0 P 1.2 0.0 a/t=0. 12. Lr (a/t=0.0 0. LrP (a) depth/thickness ratio=0.25 K IP Shallow Crack − ρ API 579 P K IP K IP − ρr P K IP KI K IP General FAD 0. Although the parametric analyses (i.2 0.e. Lr (b) KI/KIP Vs.2 0.1.4 ρr a/2c=0.25 a/2c=0.50 a/t=0. Acknowledgements ρr a/2c=0.2. 0. 13 shows the driving force comparisons considering the effect of uniform residual stress and actual residual stress for crack depth/length ratio=0.8 0.8 1.0 0.4 0.1 a/t=0. Effect of crack dimensions on modified plasticity interaction factor.1 Shallow 1.

Computer Modeling of Heat Flow in Welds Model for Welding Heat Source. Bibby M. House R. [13] Pan HL. Eng Frac Mech 1982. [8] Simoto K. ASME BVP Conference.” by Beam No Lee. 19: 633-642. Li PN. EWI CRP Report SR9818. Section 9: Assessment of Crack-Like Flaws. Szimmat J. [3] ABAQUS. Computational Aspects of Welding Stress Analysis. Development of New Modeling Procedures for 3D Welding Residual Stress and Distortion Assessment. Kishimoto K. [12] Hooton DG. 33: 635-666. [11] Ainsworth RA. . Eng Frac Mech 1997. [5] Argyris JH. Raju IS. 2000. Sakata M. [4] Feng ZL. Budden PJ. JC. Metallurgical Transactions B 1986. Cheng WT. April 1984. November 1998. Guidance on Methods for Assessing the Acceptability of Flaws in Fusion Welded Structures. NASA Technical Memorandum 85793. Computer Methods in Applied Mechanics and Engineering 1982. Energy Release Rate Along a Three-Dimensional Crack Front in a Thermally Stressed Body. 16(3): 405-413. [6] Goldak J. William KJ. Karlsson & Sorensen. commercial FEA software program by Hibbitt. Recommended Practice for Fitness-for-Service. [2] BS PD 6493 (1991)/BS 7910 (1999). Chen YS. 58(1): 149-159. The Ohio State University (2001) (Advisor: Chon L. The Assessment of Defects in Structures of Strain Hardening Materials. Stress Intensity Factor Equations for Cracks in Three-dimensional Finite Bodies Subject to Tension and Bending Loads.Journal of China Pressure Vessel Technology 1 (2003) 71-79 79 Fracture Behavior of Weldments. R6 Developments in the Treatment of Secondary Stresses. Sakata M. Int J Frac 1986. 304: 503-509. Elastic-Plastic Analysis of Crack in Thermally-Loaded Structures. 13: 841-850. [10] Shih CF. Patel B. 30: 79-102. Eng Frac Mech 1984. Tsai) References [1] API RP-579. Nakamura T. 17B: 587-600. Eng Frac Mech 1980. On the Path Independent Integral-J. [7] Newman Jr. A Procedure for Treating Secondary Stress in Failure Assessment Diagram. Fatigue and Fracture Mechanics in Pressure Vessel and Piping 1995. Moore J. Inc. Moran B. [9] Aoki S.

Sign up to vote on this title
UsefulNot useful