60th Anniversary State-of-the-Art Reviews

Predicting Fracture in Civil Engineering

Steel Structures: State of the Art
Amit Kanvinde, M.ASCE 1

Abstract: Fracture is an extreme limit state in steel structures, often precipitating structural failure or serious loss of function. Methods
to predict fracture in civil structures include traditional approaches developed in other disciplines (mechanical or aerospace engineering)
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.

subsequently adopted in structural engineering, as well as approaches to characterize earthquake-induced fractures originating within civil
engineering. Developed over nearly six decades, the state of the art is composed of theories and models that address fracture over multiple
scales and are targeted toward disparate application scenarios. The paper examines these approaches from a structural engineering standpoint,
considering trade-offs in accuracy and expense, while identifying areas for improvement. Traditional approaches (including linear elastic and
elastic plastic fracture mechanics) are presented, followed by newer local approaches that are better suited for scenarios where traditional
approaches are inapplicable. By simulating micromechanisms such as microvoid growth as well as granular cleavage, local approaches
address fracture under large-scale yielding, ultralow-cycle fatigue (which occurs during earthquakes), and low-stress triaxiality, all of which
are important in civil structures. The physical basis for these approaches is outlined, with a summary of best practices for calibration and
application. However, these local approaches have limitations as well, and often require substantial resources for successful implementation.
With this background, optimal fracture assessment strategies are outlined for common structural scenarios, considering accuracy and cost.
Limitations of the entire fracture modeling framework are summarized because they pertain to mainstream adoption within structural
engineering research and practice. As the profession moves toward accurate performance characterization, it is anticipated that research
will accelerate to overcome these limitations. DOI: 10.1061/(ASCE)ST.1943-541X.0001704. © 2016 American Society of Civil Engineers.
Author keywords: Fracture; Steel structures; Metal and composite structures.

Introduction (structural) engineering, e.g., earthquake-induced ultralow-cycle

Failure of steel structures triggered by fracture is a much-feared
situation across various engineering disciplines. Beginning with the
Liberty Ship fractures in the 1940s (Sheldon 2010) that catalyzed
the initial development of fracture mechanics as a field of study,
catastrophic structural failures are periodic reminders of the conse-
quences of fracture. Prominent examples include the fracture of the
Alexander Kielland drilling rig in 1980 (Almar-Naess et al. 1984)
resulting in 123 deaths, and steel building fractures during the 1994
Northridge and 1995 Kobe earthquakes. Because some of these
earlier fractures occurred outside the realm of civil infrastructure,
fracture mechanics initially evolved within the mechanical, aero-
space, and nuclear engineering disciplines. Within civil engineer-
ing, fracture was largely addressed through prescriptive detailing
or through design checks in which applied stresses (calculated
through classical strength of materials–type approaches) were
checked against empirical stress capacities (American Welding
Society 2015). The Northridge and Kobe earthquakes underscored
the problems with these empirical approaches, providing impetus
for studying fracture within a context specific to civil (especially
earthquake) engineering. This resulted in adaptation of existing
methodologies to civil engineering (Barsom and Rolfe 1999), as
well as the development of new models and methods (Kanvinde
and Deierlein 2007) that address problems peculiar to civil
1
Professor, Dept. of Civil and Environmental Engineering, Univ. of
California, Davis, CA 95616. E-mail: kanvinde@ucdavis.edu
Note. This manuscript was submitted on February 4, 2016; approved
on September 21, 2016; published online on November 11, 2016.
Discussion period open until April 11, 2017; separate discussions must
be submitted for individual papers. This paper is part of the Journal of
Structural Engineering, © ASCE, ISSN 0733-9445
fatigue (ULCF). The subsequent evolution of fracture mechanics
within structural engineering has been influenced by various factors
including (1) its application scenario, e.g., sharp versus blunt
cracked components, monotonic versus cyclic loading, and limited
versus extensive yielding within the component; (2) computational
resources available at the time of its development; and (3) estab-
lished trade practices and standards for material testing and char-
acterization. Not surprisingly, fracture mechanics sometimes
appears to be a patchwork of competing theories and methodolo-
gies whose suitability for addressing structural fracture under vari-
ous conditions is somewhat unclear. In view of this, the primary
objective of this paper is to present fracture prediction approaches
from a structural engineering standpoint with consideration of the
following factors: (1) application scenario, (2) trade-offs between
expense and accuracy, (3) limitations and areas for refinement, and
(4) improvements to professional practices to enable more effective
fracture simulation.
The next section provides a general overview of fracture in steel
structures, describing the responsible micromechanisms across
multiple scales. The subsequent section outlines methodologies for
fracture prediction, recognizing that all ultimately seek to charac-
terize common physical mechanisms, albeit in varying contexts and
at different scales, such that these methods are suited for different
application scenarios. The next section examines scenarios (mate-
rials, loading types, geometry) common to structural engineering
and the suitability of the different fracture models to address them.
The paper then summarizes limitations of the state of the art as they
pertain to structural engineering, outlining areas for improvement
in the modeling methods themselves and in professional practices
that are necessary to fully leverage these methods. Finally, notewor-
thy applications of fracture mechanics are discussed to illustrate its

© ASCE 03116001-1 J. Struct. Eng.

J. Struct. Eng., 03116001

 growing acceptance in professional practice. The focus of the paper
is to review the state of the art for predicting fracture in civil steel
structures, and not the state of the art of fracture mechanics itself.
This is an important distinction because the former is subject to
subjectivities in material characterization owing to trade practices,
as well as a relative lack of resources (compared to other disciplines
and applications) for effective application of some of the most ad-
vanced fracture mechanics approaches. Consequently, the paper
emphasizes approaches that are practical for structural engineering
while discussing the advanced models to provide fuller context.
Fracture in Steel Structures: General Overview
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
From a structural engineering standpoint, fracture is interpreted as
the separation of material leading to loss in load-carrying capacity.
The goal of fracture mechanics is to predict this event in terms of
applied loads or deformations. The fracture event itself may be de-
composed into phenomena leading up to it: (1) concentration of
stress and strain fields over small regions of the component due
to stress increasers such as cracks or notches, or other phenomena
such as local buckling or necking; and (2) subsequent activation of
microstructural phenomena such as microvoid nucleation, growth,
and coalescence, or alternatively granular cleavage within the local-
ized zone, to initiate and propagate fracture. Simulation of the for-
mer (development of strain-stress fields and their concentration) is
typically within the scope of continuum finite-element simulation
with appropriate sophistication (large deformations, multiaxial
plasticity). On the other hand, the latter (microstructural phenom-
ena resulting in material separation) must be inserted into the analy-
sis as failure or damage criteria. This may be performed in an
uncoupled sense, in which the failure criterion is applied after the
stress-strain fields have been computed, with the presumption that
once fracture initiates propagation and failure are inevitable. Alter-
natively, this may be done in a coupled manner, where material
separation is explicitly simulated to capture its interactions with the
surrounding stress and strain fields. Typically, this is done through
one of the following constructs: (1) creation of free surfaces in
the simulation model (Baldwin and Rashid 2013), (2) a softening
constitutive model (e.g., Gurson 1977) along with finite-element
removal, or (3) a cohesive zone model (Elices et al. 2002). These
approaches enable the simulation of ductile crack growth and its
transition to unstable, brittle propagation (Ruggieri et al. 1996).
The section on crack propagation presented subsequently provides
further discussion of these methods. The fracture criterion itself is
usually expressed as a comparison between toughness capacity
(which is assumed a material property and may be defined in multi-
ple ways) and a corresponding fracture toughness demand. The
former is determined through laboratory tests, whereas the latter
(which is a function of applied loads and geometry) is determined
through computational simulation, or, for more regular geometries
and materials, through closed-form equations. The multiplicity of
fracture criteria arises from the scales at which they are applied and
the mechanisms they seek to simulate.
Fracture criteria that are expressed at the higher (macro) scale
cannot be generalized easily across various geometric configura-
tions or loading types. These include experimental methods, in
which laboratory measurements of load or deformation correspond-
ing to fracture are used as direct estimates of the fracture capacity of
the structural component (e.g., resulting in empirical fragility func-
tions). Lower-scale approaches utilize continuum stress or strain
fields (rather than global deformations) and form the disciplinary
core of fracture mechanics. Within this disciplinary core, at one end
of the spectrum is traditional fracture mechanics where parameters
© ASCE 03116001-2
J. Struct. Eng., 03116001
such as the stress intensity factor (Anderson 1995) or the J-integral
(Rice 1968) are calculated from crack tip stress fields and used as
indicators of fracture toughness demands to be compared with their
counterpart capacities, which are determined from laboratory tests.
At the other end, local fracture criteria are represented as critical
combinations of stress and/or strain at the location of crack initia-
tion (rather than from the remote stress-strain field). In contrast to
the higher-scale models, the lower-scale models can be generalized
across a wider range of geometrical configurations, material prop-
erties, and loading types (e.g., monotonic versus cyclic). However,
they require simulation of stress-strain fields at a high resolution,
entailing high computational cost. Moreover, they require depth of
expertise in continuum finite-element simulation, large deforma-
tion mechanics, and constitutive modeling. Given these trade-offs,
the optimal model depends on the structural configuration, type of
loading, and desired level of accuracy relative to resources available
for material testing and simulation. This paper provides an assess-
ment of various methodologies against this backdrop, considering
(1) their fidelity to the physics of fracture, (2) their accuracy and
generality across various scenarios, and (3) their cost and practical-
ity. The next section describes the physical phenomena responsible
for fracture at the microstructural scale and their correspondence
with macrostructural response.
Mechanisms and Physics of Fracture in
Structural Steel
Referring to the previous section, fracture criteria seek to establish
conditions (at the component or continuum scale) under which
microstructural fracture mechanisms are activated. This section
provides an overview of these mechanisms. When the steel micro-
structure (consisting of metal grains along with carbide and sulfide
inclusions) is subjected to sufficient stress or strain, material sep-
aration occurs through one of two processes or their combination:
(1) cleavage due to tensile stress rupturing molecular bonds within
or at boundaries of grains, and (2) microvoid growth and coales-
cence due to inelastic flow of material around initiated microvoids,
leading to their enlargement and subsequent coalescence to form
the fracture surface. Figs. 1(a and b) schematically illustrate these
processes, as well as the resultant fracture surfaces. Cleavage frac-
ture does not require significant inelastic dissipation to propagate,
whereas microvoid growth and coalescence is accompanied by
plastic dissipation. The implication of this at the macrostructural
scale is that cleavage often propagates in an unstable manner (with-
out increase in applied loads), whereas microvoid controlled frac-
ture, requiring higher energy, propagates slowly and stably with
respect to applied loads. As a result, cleavage is often referred to
as brittle, whereas tearing is often considered ductile. While true
in general, the suddenness of fracture at the macroscale, often per-
ceived as brittleness, is not necessarily associated with cleavage
and is controlled by factors such as strain energy stored in the en-
tire component, loading rate, and its interaction with the energy
required to propagate the crack locally. For example, under high
loading rates or load-controlled conditions, even a locally ductile
fracture mechanism (microvoid growth) can result in highly non-
ductile performance (sudden fracture) at the structural scale.
The preferred process depends on three factors: temperature,
loading rate, and stress constraint (stress triaxiality). Lower temper-
atures, higher loading rates, and high stress constraint inhibit plastic
flow and increase stress, increasing propensity toward cleavage.
Conversely, higher temperatures, slow loading rates, and low stress
constraint promote plastic flow, favoring microvoid growth and co-
alescence. The Charpy V-notch (CVN) temperature transition curve
J. Struct. Eng.
 Intact Microvoid nucleation and Coalescence
matrix growth and fracture
Dimpled fracture surface

0.1 mm
(a)
Crack propagation
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.

Grain A

Grain B
Rupture through grains
River patterns and shiny facets
(b)

Fig. 1. Micromechanisms of fracture and corresponding failure surfaces: (a) microvoid growth and coalescence; (b) cleavage; transgranular cleavage
shown

Dynamic predate the advent of the discipline of fracture mechanics, often

Static Curve
Curve
considered coincident with the introduction of the Griffith (1921)
Absorbed energy

Upper shelf –
microvoid growth
Temperature shift
Transition region –
combination of microvoid
growth and cleavage
Lower shelf –
predominantly cleavage
Test temperature
Fig. 2. Characteristic temperature transition response of structural
steels
well known to structural engineers is a direct macrostructural mani-
festation of this. Fig. 2 shows a typical CVN curve, plotting the
absorbed CVN energy (as a measure of toughness) against temper-
ature. Referring to the figure, three distinct regions are observed. At
low temperatures, fracture is controlled by cleavage resulting in low
toughness. At intermediate temperatures, a transition region is ob-
served, where some ductile tearing may precede cleavage, resulting
in increased toughness. Finally, at higher temperatures an upper
shelf is observed, where fracture is controlled entirely by ductile
tearing. Fig. 2 also shows the effect of loading rate on CVN, such
that an increased loading rate lowers toughness, especially in the
transition region. With this background, the next section discusses
fracture simulation methodologies for steel structures.
Models for Predicting Fracture in Steel Structures
Empirical methods to characterize fracture (or more generically,
structural failure) have been prevalent for several decades and
criterion for fracture, based on strain energy release rate. The ap-
plication of these methods involves calculating a force or deforma-
tion demand through structural analysis and then comparing it to a
corresponding capacity, determined through (ideally full-scale)
component testing. This approach is accurate (assuming the tested
components are sufficiently similar in scale and construction to the
field prototypes), but expensive and difficult to generalize to dis-
similar structure components. Nevertheless, it is used when (1) there
is a low tolerance for inaccuracy or modeling uncertainty that may
be introduced due to lower-scale methods, and (2) sufficient resour-
ces are available for large-scale testing. Successful applications of
this approach include the multiuniversity campaign of large-scale
testing and analysis during the SAC steel project (Malley 1998),
motivated by the Northridge moment frame fractures. The design
considerations arising from these were based primarily on the ex-
perimental observations. Accordingly, standards such as AISC 358
(AISC 2010a) reference prequalified connections (FEMA 2000)
that ensure similarity between the tested specimens and archetype
connections, recognizing that purely experimental observations are
difficult to generalize. Although these methods are important, their
intellectual basis is self-evident. This section focuses on the
specialized fracture mechanics approaches.
Traditional Fracture Mechanics
The fundamental concern of traditional fracture mechanics is
precracked bodies. Reasons for this are practical as well as aca-
demic. From a practical standpoint, fracture is often triggered
around sharp cracks and discontinuities in structures (and less often
in smooth details). Academically, sharp cracks are interesting be-
cause they imply a stress singularity (stress and strain approaching
infinity) at the crack tip (Fig. 3), thereby precluding any stress- or
strain-based failure criterion. To overcome this problem, traditional
fracture mechanics formulates the fracture criterion in terms of
parameters that characterize the stress field in the vicinity of the

© ASCE 03116001-3 J. Struct. Eng.

J. Struct. Eng., 03116001


Increasing KI
σ
r
Singular stress fields
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
Fig. 3. Crack tip singularity implicit in linear elastic fracture mechanics
crack, rather than a discrete location within the body. The implicit
assumption is that these parameters are correlated with crack tip
stress and strain fields that activate the microstructural processes
described previously. Two popular parameters associated with tra-
ditional fracture mechanics are the stress intensity factor (denoted
K I ), and the contour J-integral (denoted J I ). The K I forms the
core of linear-elastic fracture mechanics (LEFM), whereas the
latter is arguably the most important parameter within elastic-
plastic fracture mechanics (EPFM). Other EPFM parameters such
as the crack tip opening displacement (CTOD) or crack tip open-
ing angle (CTOA) are not discussed here, albeit extensive discus-
sion on these may be found in textbooks (e.g., Anderson 1995;
Saxena 1997).
Both K I and J I may be interpreted as fracture toughness de-
mand parameters determined from the stress field in a loaded struc-
tural component. The former (K I ) is interpreted as the amplitude
of the singularity of the stress field near the crack tip in an elastic
body, shown in Fig. 3. Eq. (1) indicates that (1) the stress field may
be expressed as a series expansion in terms p offfiffiffithe distance r ahead
of the crack tip, such that K I scales the 1= r term, which repre-
sents the singularity because σ → ∞ as r → 0; and (2) a higher K I
(shown by the dashed line in Fig. 3) implies higher stresses remote
from the crack trip, although both curves approach infinity near the
crack tip
KI
σ ¼ pffiffiffiffiffiffiffi
ffi þ higher order terms ð1Þ
2πr
The J-integral (J I ) = central parameter of EPFM. This param-
eter [proposed by Rice (1968)] may be interpreted as the strain en-
ergy release rate per unit area extension of a crack in a nonlinear
elastic body. Similar to K I , J I may be interpreted as a stress inten-
sity parameter as well, as demonstrated by Hutchinson (1968) and
Rice and Rosengren (1968), defining the amplitude of a crack tip
singularity. Thus, traditional fracture mechanics (LEFM or EPFM)
rely on the notion that the amplitude of the stress field singularity
defined by K I or J I (rather than pointwise stress values themselves)
may be used as a fracture parameter, such that both toughness de-
mands and capacities may be expressed in terms of this parameter.
Following this, the fracture criteria in traditional fracture mechanics
may be expressed as
© ASCE 03116001-4
J. Struct. Eng., 03116001
K I ≥ K IC or J I ≥ J IC ð2Þ
In the preceding inequality, the left-hand side (K I or J I ) repre-
sents the toughness demand in a structural component, which is
determined through finite-element simulations or sometimes (for
more regular geometries) through analytical or regressed formulas
[refer to Rooke and Cartwright (1976) for K I solutions and Kumar
et al. (1984) for J I solutions]. The right-hand side (K IC or J IC ) is
assumed to be a material property determined through standardized
testing [ASTM E399 (ASTM 2013a) for K IC, ASTM E1820
(ASTM 2013c) for J IC ]. This departure from continuum-based
failure theories (which are based on pointwise stress evaluations)
is necessitated by the stress singularity. In reality, the stresses (and
strains) at the crack tip are regularized (meaning they are not equal
to infinity) due to (1) the formation of a plastic zone around the
crack tip, (2) microstructural features that invalidate continuum me-
chanics over these small scales, or (3) finite crack widths. Even
today, it is computationally expensive to simulate these features
with the aim of characterizing stress and strain fields in the close
vicinity of the crack tip. The usefulness of a far-field or remote
parameter such as the stress intensity factor is even more under-
standable considering the computational resources available when
LEFM and EPFM were first developed.
Traditional fracture mechanics is valid only as long as the tough-
ness parameters retain a unique correspondence with the stress-
strain fields in the vicinity of the crack tip, even if these fields
are not consistent with the underlying theory (e.g., linear elasticity)
used to formulate the parameters. In general, this is true when the
size of the plastic zone at the crack tip is significantly smaller than
the region of K I or J I dominance (i.e., the region where the singular
stress-strain fields consistent with K I or J I dominate the overall
stress field). In practical terms, this results in size limitations on
calibration specimens for K IC and J IC and corresponding size lim-
itations for structural components within which they may be used to
predict fracture. For example, LEFM is valid if the characteristic
dimension (e.g., flange thickness, crack length) of a component
is greater than approximately 2.5 × ðK IC =Fy Þ2. This indicates over-
all elastic response before fracture. EPFM is valid when the speci-
men size is greater than 10 × ðJ IC =Fy Þ, which typically allows for
ductile fracture at the macroscale. The size limitations for EPFM
are more liberal than those for LEFM because EPFM formulations
admit material nonlinearity, thereby preserving the correspondence
between J I and crack tip stress fields for a higher level of plasticity.
Because linear elasticity is a special case of nonlinear elasticity,
EPFM is always valid when LEFM is, but not vice versa, meaning
J I -based fracture assessment may be conducted even when the
response is expected to be linear elastic or brittle (following the
preceding size limitations). On the other hand, K I -based assess-
ment may be used only in the case of elastic or brittle fracture.
A corollary of this is that for linear elastic response, J I and K I are
equivalent. In fact, they may be uniquely related such that J I ¼
K 2I × ð1 − ν 2 Þ=E. An analogous relationship may be used to relate
K IC and J IC . In summary, traditional fracture mechanics ap-
proaches may be used when (1) there is a sharp crack or flaw
in the structural component, (2) the spatial extent of yielding is
small with respect to component size, and (3) the loading is
monotonic.
Following the previous discussion, traditional fracture mechan-
ics [Eq. (2)] require (1) either regressed formulas or finite-element
simulation to determine K I or J I in the structural component, and
(2) laboratory tests to calibrate K IC and J IC . Commercial and aca-
demic finite-element programs (e.g., Abaqus; Dodds et al. 2012)
are able to simulate stress and strain fields required for the deter-
mination of K I or J I subject to two considerations. First, the
J. Struct. Eng.

Rosette of “Quarter point” crack tip elements
(a)
Crack
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
Highly refined Crack tip elements 0.005 mm
(b)
Fig. 4. Meshes for sharp crack simulation: (a) zero width crack with
quarter-point elements; (b) finite width crack with highly refined crack
tip mesh
material constitutive response (von Mises plasticity for steel) must
be calibrated accurately, especially for high strains (greater than
0.5, i.e., 50% strain). Second, the finite-element mesh in the vicin-
ity of the crack tip must simulate the high stress and strain gradients
in this region. Fig. 4 shows examples of two acceptable mesh pat-
terns. Fig. 4(a) shows a crack tip rosette in which the element for-
mulations are enriched by moving the center node to the quarter
point toward the crack tip; in this manner, the elements can reflect
the singular stress field near the crack tip. In this case, the crack is
“Atomically Sharp”
fatigue precrack
(a)
Fig. 5. Specimens commonly used for fracture model calibration: (a) compact tension specimen for K IC, JIC ; (b) circumferentially notched tension
(CNT) specimen; (c) inclined notch butterfly (IN) specimen; (d) rectangular notched (RN) specimen
© ASCE 03116001-5
J. Struct. Eng., 03116001
modeled with zero width. Fig. 4(b) shows an alternative, in which
the crack is modeled as a finite width gap, with refinement of the
mesh to capture stress and strain gradients. The latter approach is
acceptable if the crack width is significantly lower than the critical
CTOD (∼0.25 mm for structural steels). Modern software plat-
forms such as Abaqus and WARP3D provide on-board capabilities
for K I and J I computations.
Laboratory tests for the determination of K IC and J IC are
informed by the standards ASTM E399 (ASTM 2013a), and
ASTM E1820 (ASTM 2013c), respectively, requiring sharp-
cracked specimens such as the compact tension specimen shown
in Fig. 5(a). These specimens must conform to strict geometrical
limits to ensure small-scale yielding (SSY). Additionally, these
specimens require the application of fatigue loading prior to testing
to produce a zero-width precrack. These processes add significant
effort and cost to fracture toughness characterization. Owing to
these costs, standard K IC and J IC tests are rarely conducted for civil
structures (neither for materials specification nor for performance
assessment), and most design standards reference the CVN fracture
toughness, which may be determined more economically through
impact testing. For example, the AISC seismic provisions [AISC
341-10 (AISC 2010b)] require minimum material toughness for
various scenarios in terms of CVN. While these requirements es-
tablish a baseline level of material toughness, they cannot be used
for quantitative fracture prediction. This necessitates the use of cor-
relations between K IC and CVN to substitute direct measurements
of K IC . One such correlation, proposed by Barsom and Rolfe
(1969) is shown in Eq. (3)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
K ID ¼ 0.646 × CVN × E ð3Þ
In this equation K ID is the critical stress intensity factor at a
dynamic rate of loading (strainrate ≈ 105 mm=mm=s). This is sig-
nificantly lower compared with K IC for seismic or static loading
(strainrate ≈ 0 − 0.1 mm=mm=s), implying that using the correla-
tion is conservative. However, if the full Charpy transition curve
(similar to that previously shown in Fig. 2) is available, then the
correlation of Eq. (3) may be used along with a temperature shift
(Fig. 2) to generate an equivalent curve for K IC. It is also pertinent
θN
45°
(b) (c) (d)
J. Struct. Eng.
 Yield ULCF Low T history effects Cleavage
to mention here the Master Curve approach based on the work by

Yesd
Yes
Yes

Yes
No

No
No

No

No
Wallin (1998), which establishes the temperature dependence of
fracture toughness in an efficient manner through a limited number

Ability to simulate fracture under

of K IC tests. The Master Curve approach is standardized as per

Loading
ASTM E1921 (ASTM 2013b).

Yes
Yes

Yes
No
No

No

No

No
No
Local Approach to Fracture
Traditional fracture mechanics requires the correspondence of

Yes

Yes
No
No

No

No
No

No
No
its parameters with local stress fields under specified conditions
(small-scale yielding, monotonic loading). It is ineffective when
these conditions are violated. In contrast, local approaches con-

Yes

Yes
No
No

No

No

No

No
No
struct fracture criteria directly in terms of the stress and strain fields
(at the crack tip or in the zone of high distress). By relying on the
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.

Yesc
Yesc

Yes

Yes
Yes

Yes

Yes

No
No
local stresses, these approaches are less constrained by require-
ments of specimen or component size, yielding, or the necessity of

sharp cracked for length scale

CNT, inclined notch butterfly

a sharp crack or flaw, and therefore are more generally applicable.

Sharp-cracked CTa, SENBb

tension (CNT) along with

Circumferentially notched
On the other hand, they are contingent on effective simulation of

(IN), rectangular notched

Sharp cracked similar to

(RN) specimens (Fig. 5)
Specimens (Fig. 5)
the stress and strain fields over extremely small scales (on the order
of steel grain size ≈ 0.1 mm). In turn, this presents challenges in

LEFM or EPFM
terms of computational demand and simulation of material constit-
utive response. Additionally, the notion of material homogeneity is

calibration
and others
also somewhat questionable at this scale, sometimes necessitating
the simulation of heterogeneous microstructural features such as
individual grains.
Unlike traditional fracture mechanics, local approaches seek to

but best practices developed

simulate specific mechanisms of fracture. The assumption is that

No established standards;
J IC , CTODC , CTOAC E1820 (ASTM 2013c),
fracture will initiate and propagate in accordance with the mecha- Calibration

ASTM standards

E1290 (ASTM 1999),

E2472 (ASTM 2012)
nism (microvoid growth or cleavage) that is more favorable based
on stress state, loading rate, and temperature. Because these mech-
anisms are dependent on material microstructure, numerous local

in research
models have been developed for specific materials; a comprehen-
sive review is available in Besson et al. (2006). In this paper, popu-
E399

lar models for two common mechanisms (microvoid coalescence

and cleavage) are presented. Table 1 summarizes the various local
models discussed in this section, along with their features, param-

B1 , B2 , B3 , B4 , n
eters that require calibration, and experiments required to cali-

Only if plastic zone is small, i.e., specimen or component size limitations are satisfied.


SWDM , λ, l
VGM , λ, l
Parameters


A1 , A2 , l

brate them.
l

σ f , λ
σu , m
,
K IC

Dcritical
VGM
Dcritical

Dcritical
Local Fracture Criteria for Microvoid Growth and
Coalescence
Void growth and coalescence around sulfide or carbide particles
is a primary fracture mechanism in structural steel. Criteria used
Eqs. (7) and (8) Microvoid growth, collapse,

Microvoid growth, collapse,

Micromechanism simulated

to describe this response are typically based on understanding of

microvoid or cavity growth in elastoplastic continua. Early ana-
coalesecence, and shear
Microvoid growth and

lytical work by McClintock (1968), and Rice and Tracey (1969)

Not associated with

Microvoid growth,

coalescence, shear

shows derivations for the cases of cylindrical and spherical

micromechanism

cavities, respectively. Their findings are similar, and are expressed

coalescence

coalescence

in Eq. (4)
Cleavage
Table 1. Summary of Selected Fracture Criteria

dR=R ¼ C1 × expðC2 × TÞ · dεp ð4Þ

Including spatial sampling of weak links.
Eqs. (5) and (6)

Referring to Eq. (4), the main observations are that (1) incremen-
Equations

Eq. (11)

Eq. (12)

Eq. (10)

Eq. (13)
Eq. (15)
Eq. (2)
Eq. (2)

tal void growth dR=R is proportional to the incremental equivalent

plastic strain dεp , and (2) the proportionality is a strong function
Single-edge notched bending.

of the stress triaxiality T ¼ σm =σe, where σm and σe are the mean

(or hydrostatic), and effective (or von Mises) stress, respectively.
The original derivations by Rice and Tracey (1969) suggest a value
Bao and Wierzbicki
Cyclic void growth
Void growth model

Ritchie Knott Rice

Compact tension.

of 0.283 for C1 and 1.5 for C2. The validity of this relationship
Stress-weighted
Stress-modified

has been demonstrated through numerical void cell simulations

damage model
critical strain

(Faleskog et al. 1998; Cooke 2015, Fig. 6) and through numerous

Criterion

Beremin

studies that have explored the response of cavities (or microvoids)

LEFM
EPFM

model

and their effect on ductile rupture under a range of stress states,

loading regimes, and material microstructures. Some examples
b

d
a

© ASCE 03116001-6 J. Struct. Eng.

J. Struct. Eng., 03116001


Coupon scale simulations for Single void cell to develop
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
stress state determination fundamental void mechanics
Fig. 6. Void cell simulations in support of micromechanics-based fracture criteria
include work by Srivastava and Needleman (2012) examining void
growth and collapse in monocrystals and Basu and Benzerga
(2015) investigating the loading path dependence of void growth
controlled ductile fracture. Other refinements include incorporation
of cavity shape effects on fracture, e.g., Pardoen and Hutchinson
(2000). These represent only some of the advances in a field that
continues to evolve rapidly. Cumulatively, these studies have re-
sulted in (and continue to refine) insights regarding dependencies
of void response on key aspects of the stress-strain fields and their
histories. Consequently, the large majority of local fracture models
are based on these insights. Due to the multiplicity of conditions for
which these insights have been developed, variants of local models
that rely on them are similarly numerous. This paper focuses on a
subset of these that have been recently examined and applied to
structural steel components.
Local fracture criteria that seek to simulate microvoid-induced
ductile fracture based on the previous research may be broadly clas-
sified into two categories: (1) damage mechanics–based criteria, in
which damage (which may be interpreted as proxy for void size) is
evaluated as a history variable quantity based on evolving stresses
or strains, for example, by integrating the left-hand side of Eq. (4);
or (2) state-based criteria, in which a critical fracture strain is
expressed as a function of stress state. Within the damage mechan-
ics category, the models may be further classified as coupled or
uncoupled:
• Coupled models are able to simulate interactions between
the fracturing material and the surrounding stress field by incor-
porating damage as a quantity that affects material softening.
A classic example is the Gurson (1977) model, which
simulates material softening response due to void growth.
Numerous coupled damage mechanics models have been
subsequently developed to target specific aspects of response,
which are triggered under different stress states or material
microstructure. For example, Kiran and Khandelwal (2015)
incorporate the effect of void elongation along with dilation
in such a model, whereas others (e.g., Ruggieri et al. 1996)
have incorporated material-specific length scales into such
models.
• Uncoupled models apply the fracture model as a failure criterion
(through a postprocessing check) to stress and strain fields
that are computed based on a constitutive model that does
not simulate damage or softening. As a result, these are some-
what less generalizable as compared with the coupled models,
especially if the strain distribution (or localization) leading to
fracture is driven by damage itself.
© ASCE 03116001-7
J. Struct. Eng., 03116001
Multivoid cells to
characterize intervoid
interactions and localization
The use of coupled models (which are essentially constitutive
models with softening material response) results in mesh sensitivity
of the solution, making them somewhat problematic for application
in civil or structural simulations, where resources for mitigation of
mesh sensitivity [e.g., the use of regularization schemes (Enakoutsa
et al. 2007)] may be inaccessible or unfeasible. Accordingly, the
more accessible uncoupled models are discussed in this paper. In
this context, recent efforts by Kiran and Khandelwal (2014) to de-
velop weakly coupled models that balance the accuracy of coupled
models with the convenience of uncoupled ones are particularly
notable.
Within the damage mechanics category, the void growth model
(VGM) is relatively well known. This model, which has been used
by Panontin and Sheppard (1995) and Kanvinde and Deierlein
(2006), among others, as a fracture criterion predicts fracture when
the following condition [based on a direct integration of Eq. (4)] is
satisfied:
Z
lnðR=R0 Þ ≥ C1 × expðC2 × TÞ · dεp ≥ lnðRcrit =R0 Þ ð5Þ
loading
Simplifying further, this may be expressed as
DVGM ≥ Dcritical
VGM ð6Þ
In the previous equation, the void growth damage DVGM ¼
∫ loading expðC2 × TÞ · dεp , and its critical value Dcritical
VGM ¼
lnðRcrit =R0 Þ=C1 . Simplification in this manner allows for the
calculation of the void growth damage as a demand parameter
based solely on simulation [C2 is often considered to be a material-
independent parameter with the value 1.5 (Kanvinde and Deierlein
2006)], whereas Dcritical
VGM may be interpreted as a material property
(or a damage capacity) requiring calibration. For structural steels,
Dcritical
VGM is in the range of 1.00–5.00. Kanvinde and Deierlein (2007)
extended the void growth model to ULCF, which typically occurs
during seismic loading. Unlike high-cycle fatigue, which occurs in
bridges or mechanical components (involving thousands or even
millions of cycles), ULCF is characterized by a small (less than 20)
number of cycles with high strain amplitudes (several times yield).
In terms of micromechanisms, ULCF bears a close resemblance
to fracture (because it involves cyclic growth and collapse of voids)
rather than high cycle fatigue [which involves the processes of
dislocation slip and decohesion (Suresh 1998)]. The cyclic void
growth model (CVGM) (Kanvinde and Deierlein 2007) has a dam-
age functional form similar to the VGM, except it accounts for
cyclic deterioration, as in Eq. (7)
J. Struct. Eng.
 DCVGM ≥ Dcritical
CVGM ð7Þ
In the preceding, the damage index DCVGM accounts for
reversed cyclic loading, such that
Z Z
DCVGM ¼ expð1.5 × jTjÞ · dεp − expð1.5 × jTjÞ · dεp
T≥0 T<0
ð8Þ
Eq. (8) separates the loading history into positive and negative
triaxiality cycles, reflecting that voids grow and shrink during al-
ternating cycles of positive and negative triaxiality. The damage
capacity may be determined according to the following expression:
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
Dcritical critical
CVGM ¼ DVGM × expð−λ · εacc Þ ð9Þ
According to the preceding expression, the damage capacity de-
creases with loading as a function of the accumulated plastic strain
εacc . At the initiation of loading, εacc ¼ 0, and as a result Dcritical
CVGM ¼
Dcritical
VGM with the implication that for monotonic loading the CVGM
is identical to the VGM. Consequently, as summarized in Table 1,
the CVGM requires the calibration of two parameters (in addition
to a length scale parameter, discussed subsequently), one of which
is the damage capacity of the monotonic VGM, i.e., Dcritical VGM ; the
other is λ, characterizing the rate of cyclic deterioration. For struc-
tural steels, this parameter is in the range 0.0–1.0. Myers et al.
(2009) demonstrate an application of CVGM to large-scale struc-
tural tests. The original derivations by McClintock (1968) and Rice
and Tracey (1969) were developed for high triaxiality (T > 1.0)
owing to the importance of fracture at crack tips. Low triaxiality
is commonly found at protruding corners of structures (e.g., tips
of beam flanges) or in high-shear conditions that are common
in civil structures. To address this, Smith et al. (2014) proposed
the stress-weighted damage model (SWDM), whose fracture
criterion is
Z usually taken as 1.5. The parameter A1 is often the only material
DSWDM ¼ eλ·εacc × ðe1.3T − e−1.3T Þ · e0.5X · dεp ≥ Dcritical
SWDM
loading
ð10Þ
The preceding criterion is similar to the CVGM or the VGM in
that it is based on the damage index DSWDM , which must be com-
pared with its critical value Dcritical
SWDM . However, the SWDM criterion
offers improvements compared with the CVGM in two respects:
• It reflects a lower damage rate for low triaxiality such that at
zero triaxiality the damage rate [the integrand in Eq. (10)] is 0.
This is consistent with experimental data by Bridgman (1964)
and Smith et al. (2014) and is not accounted for by the CVGM
and similar models.
• It incorporates the effect of the deviatoric parameter (or Lode
angle parameter) X, which may be determined as X ¼
pffiffiffi
3 3 · J 3 =ð2 · J 3=2
2 Þ, where J 2 and J 3 are the second and third
invariants of the deviatoric stress tensor. This parameter may be
interpreted as an additional indicator of stress state, such that
X ¼ 1 indicates an axisymmetric stress state (e.g., at the cen-
ter of a cylindrical bar) and X ¼ 0 indicates plane strain. This
parameter has been shown through experimental data (Bao and
Wierzbicki 2004; Xue 2007) to strongly influence fracture.
Functional forms derived by Rice and Tracey (1969) and
McClintock (1968) are developed only for the axisymmetric
condition with X ¼ 1. Moreover, a majority of experimental
specimens used to develop and calibrate models based on this
functional form (e.g., the CVGM and VGM) are cylindrically
notched specimens with an axisymmetric stress state at the
© ASCE 03116001-8
J. Struct. Eng., 03116001
center (where fracture initiates). This suggests that generalizing
these models to nonaxisymmetric stress states is potentially in-
accurate; the SWDM addresses this. As summarized in Table 1,
the SWDM requires the calibration of two parameters (in addi-
tion to the length scale), similar to the CVGM. One is the
Dcritical
SWDM , and the other is λ (identical in interpretation and values
as for the CVGM).
Other notable models in this category (uncoupled damage
mechanics models addressing low triaxiality) include those by
Wen and Mahmoud (2016a) for monotonic loading and Wen and
Mahmoud (2016b) for cyclic loading. These models are similar to
the SWDM discussed previously, but propose somewhat different
functional forms for damage accumulation based on different data
sets. Similar developments have been made in the area of coupled
damage mechanics models as well; for example, Nahshon and
Hutchinson (2008) present such a model for shear-dominated,
low-triaxiality stress states.
The models discussed previously require the numerical integra-
tion of stress and strain quantities at each continuum location to
apply the fracture criterion. On the other hand, state-based models
are based on the instantaneous stress and strain fields (rather than
their histories). For example, a class of models [termed stress-
modified critical strain (SMCS)] take the following general form:
εp ≥ εcritical
p ¼ A1 · expð−A2 · TÞ ð11Þ
The preceding criterion (in which A1 and A2 are calibrated
parameters) may be evaluated without numerical integration (as
required in the VGM, CVGM, and SWDM models). In fact, the
functional form shown in Eq. (11) arises directly through integra-
tion of Eq. (5) for the VGM, assuming that the triaxiality is T, a
constant over the loading process. Consequently, SMCS-type mod-
els are accurate when the temporal variations in triaxiality are
modest. The variants of the SMCS models are typically distin-
guished through the calibrated values of A1 and A2 ; the latter is
property that requires calibration. Popular variants of the SMCS-
type models include those by Hancock and Mackenzie (1976),
Johnson and Cook (1983), and Rousselier (1987).
The SMCS models are state-based counterparts to the damage-
mechanics models of VGM or SWDM. Consequently, in addition
to their inaccuracy with respect to loading histories, they cannot
accurately describe the effects of low triaxiality or the Lode param-
eter. Analogous state-based counterparts for low-triaxiality fracture
have been developed as well. Major contributions in this area have
been made by current and former members of the Wierzbicki
research group at the Massachusetts Institute of Technology. For
example, Bao and Wierzbicki (2004) presented a fracture criterion
targeted toward crashworthiness assessment of vehicles, where
low-triaxiality stress states are frequently encountered. This cri-
terion is similar to the SMCS in terms of its history independence
εp ≥ εcritical
p ¼ B1 · expð−B2 · TÞ − ½B1 · expð−B2 · TÞ
− B3 · expð−B4 · Tފ · ð1 − X 1=n Þn ð12Þ
The parameters B1 , B2 , B3 , B4 , and n are calibrated material
parameters. Fig. 7(a) shows the failure locus (i.e., the strain capac-
ity as a function of the stress triaxiality) for the SMCS model
[Eq. (11), which does not consider the effect of the Lode param-
eter], as well as Eq. (12), which does. Referring to Fig. 7(b),
Eq. (12) can capture the effect of the Lode parameter (meaning
it can distinguish between axisymmetric and plane strain situa-
tions even for the same stress triaxiality), whereas the SMCS-type
models cannot. However, similar to the SMCS models, these are
J. Struct. Eng.
 ε pfracture
Triaxiality T
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
(a)
Fig. 7. Fracture locus for (a) SMCS-type models, invariant to deviatoric parameter [Eq. (11)]; and (b) Bao and Wierzbicki (2004) model accounting
for deviatoric parameter dependence [Eq. (12)]
unsuitable if the stress state changes significantly over the loading
history; in this case the damage mechanics–type models are more
appropriate. Some recent research that has focused on developing
and refining models in this general area includes Roth and Mohr
(2016), Bai and Wierzbicki (2010), and Xue (2007).
Local Criteria for Cleavage Fracture
Cleavage fracture occurs due to the rupture of molecular bonds
within (or on the boundaries of) metal grains, with the implication
that it is stress (rather than strain) controlled. It is more probable
when the material yield stress is increased either due to low tem-
perature, high loading rates, or high stress constraint. Cleavage is
typically triggered around weak links (such as inclusions) in the
material microstructure, and its onset is quite random, depending
on the manner in which the stress field samples these weak-links.
Two popular local criteria are discussed here: one is commonly
termed the ritchie knott rice (RKR) model (Ritchie et al. 1973),
and the other is termed the Beremin model (Beremin 1983).
The fracture criterion for the RKR model is a simple stress check,
as shown in Eq. (13)
σmax ≥ σf ð13Þ
In this equation, σmax = maximum (tensile) principal stress at
any location, and σf = fracture stress. As per the RKR model, the
previous condition must be satisfied over a characteristic distance
(sometimes denoted λ ) to ensure adequate sampling of material
weak links. The RKR model is deterministic in its formulation,
but acknowledges randomness in the cleavage process through the
sampling parameter. The Beremin model (Beremin 1983), popu-
larly known as the Weibull stress method, formalizes this random-
ness in a rigorous way, such that the fracture criterion is described
in a probabilistic manner, wherein the probability of fracture may
be expressed as
Pf ¼ 1 − exp½−ðσw =σu Þm Š ð14Þ
In this equation, σu and m are material parameters (sometimes
termed Weibull moduli), whereas the Weibull stress σw is a demand
quantity; more specifically, an averaged stress causing a consistent
probability of failure accounting for weakest-link sampling over the
entire component. It is determined as follows:
© ASCE
J. Struct. Eng., 03116001
ε pfracture
Axisymmetric
stress state
Plane strain
(b)
 1=m
Z
σw ¼ ð1=V 0 Þ σm
max · dV ð15Þ
Ω
In Eq. (15), V 0 = reference volume, which may be arbitrarily
chosen in concert with the material parameters σu and m. The
underpinnings of this approach and calibration methods are dis-
cussed at length in Beremin (1983), and by Matos and Dodds
(2001) in a structural engineering context (fracture of beam-
column connections in the Northridge earthquake). In the context
of this paper, it suffices to observe that (1) once V 0 , σu , and m
have been selected, Eqs. (14) and (15) may be evaluated to es-
timate the probability of cleavage fracture; and (2) this probability
reflects the spatial sampling of presumed weakest links in the
material, interacting with a nonhomogenous stress field in the
loaded component.
As discussed previously, cleavage fracture is preferred when
the stresses are elevated due to a combination of constraint, tem-
perature, and loading rate. For most structural components, as-
suming that the loading rate and temperature are relatively
constant during loading, fracture often initiates due to the micro-
void coalescence process (especially if sharp cracks are absent).
However, as loading progresses, stresses rise as a result of two
phenomena: (1) formation of crack surfaces due to microvoid
growth, resulting in an increase in stress triaxiality; and (2) strain
hardening. This increases the probability of triggering cleavage
fracture. Moreover, because cleavage fracture is not associated
with additional plastic dissipation, transition to cleavage usually
implies unstable propagation, as observed in many buildings after
the Northridge earthquake.
The local models described previously are applied through
stress and strain fields determined through continuum simulations.
This means that in theory they can be applied at any continuum
point, which has zero volume. This interpretation is meaning-
less unless there is a finite volume associated with this point
because fracture occurs only when microstructural features (such
as void inclusions, which have finite dimensions) are activated.
The Beremin model addresses this explicitly. For the void growth
models, a characteristic length (or volume) parameter must often be
included in the model definition; this length is on the order of 2–10
times the material grain size. Once this length scale is established,
any of the model criteria described Table 1 may be implemented
based on stress and strain fields determined from appropriately
03116001-9 J. Struct. Eng.
 Increasing loading Application to Common Structural Engineering
parameter, such as Situations
deformation or J-integral
D / Dcrit This section examines the suitability of various models (discussed
in the previous section) for common structural engineering scenar-
ios. The discussion is premised on the notion that the ultimate ob-
Solid line indicates stress field jective of fracture modeling is to determine a global (or structural)
deformation corresponding to force or deformation at which fracture (and associated loss of
fracture strength) occurs. This may be required for a component that is
being designed or is already in service, or sometimes for develop-
ing broader insights into effects of detailing or material properties
in support of design standards. Referring to prior discussion, differ-
ent fracture models require disparate resources for simulation as
well as material characterization, resulting in varying levels of ac-
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.

curacy for different applications. Following this, the primary objec-

l r = Distance ahead of
crack or notch tip
Fig. 8. Determination of material length scale or volume parameter l
constructed finite-element simulations and calibrated material
parameters. Fig. 8 shows this approach schematically.
Because the local models are an active area of research
(with a multiplicity of functional forms and so on), methods
to calibrate their parameters are, in general, not standardized.
Consequently, calibration is based on (1) experimental coupons
(such as cylindrically notched tension coupons with varying
notch dimensions, Fig. 5) to interrogate various stress states, and
(2) complementary finite-element simulations. The stress-strain
data from the finite element model corresponding to the instant
of fracture observed in the experiments is then used to back-
calibrate the parameters of the local model by enforcing the frac-
ture criterion at this instant. Some studies (e.g., Myers et al.
2010, 2014) have focused exclusively on methods for parameter
estimation). Table 1 and Fig. 5 summarize calibration specimens
for various models.
tive of this discussion is not to advocate for a particular modeling
approach, but rather to provide perspective for optimal modeling
strategies, recognizing that the optimal strategy is influenced by
available resources for calibration experiments and computational
resources, as well as the desired accuracy in prediction, noting that
a full-scale test of the component is the most accurate but also likely
the most expensive approach. The section describes conditions
commonly occurring in structural engineering and examines vari-
ous modeling options for each. Fig. 9 schematically illustrates
structural components where these conditions may be encountered.
Referring to Fig. 9, the primary classification of the scenarios is
whether they include sharp cracks or smooth transitions in geom-
etry. Subclassifications pertain to whether the components are mon-
otonic or cyclic, and whether there is the expectation of large-scale
yielding prior to fracture. These are not all possible combinations,
primarily because some (e.g., blunt notched conditions with large-
or small-scale yielding) are essentially equivalent in terms of their
amenability to modeling approaches. Before discussing each sce-
nario, it is useful to establish the connotations of sharp crack versus
smooth transitions and small- versus large-scale yielding.
Components may be classified as sharp cracked if they contain
cracks (or flaws) that are introduced without removal of material
(sometimes referred to as atomically sharp cracks). Such cracks are
typically the result of high cycle fatigue where cracks grow under

High (sharp cracked) Intermediate (notches, holes local Low (corners and shear bands)
buckles )
Stress triaxiality
Monotonic

Cyclicity

Partial Joint Penetration with unfused Shear bands in block shear

Net section fracture around holes
weld root
Seismic
ULCF

Seismic beam column connection with

Cyclically buckling brace with local Seismic column base plate connection
flaw at weld root
High Cycle Fatigue buckling induced fracture with corner fracture
(Vehicular/ mechanical)

Fig. 9. Common structural engineering situations in the space of stress triaxiality and load cyclicity

© ASCE 03116001-10 J. Struct. Eng.

J. Struct. Eng., 03116001

 relatively low-amplitude cyclic stresses through the microstructural
processes of slip and decohesion of dislocations (Suresh 1998).
Cracks in steel bridge decks are a common civil engineering exam-
ple. Sharp cracks may also be formed when fracture initiates at
a blunt notch or in a smooth structural detail. Some types of flaws
(e.g., those produced by lack of fusion in welded components) or
in partial joint penetration (PJP) welds may be considered sharp for
practical purposes, although they are not atomically sharp in theory.
Following work by McMeeking and Parks (1979), if the crack
width is significantly less than the CTOD, which is on the order of
0.25 mm for low-carbon structural steels, of a material, then it may
be conservatively treated as sharp for the purposes of fracture
prediction.
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
The notion of SSY is interpreted in the context of its effect on
the accuracy of traditional fracture mechanics; it is generally under-
stood to mean that the plastic zone is well confined (typically in the
vicinity of the crack tip) such that traditional fracture mechanics
parameters (K I , J I , or CTOD) bear unique correspondence with
crack tip stresses. Thus, SSY for structural components may be
defined because it is for the validity of standard K IC or J IC tests.
With this background, various structural engineering scenarios are
discussed.
Scenario Type 1: Structural Components with
a Sharp Crack
For structural components with a sharp crack, two subscenarios are
distinguished by whether the component is loaded monotonically
or cyclically. Fig. 9 shows a PJP welded connection with a lack-
of-fusion flaw, and a pre-Northridge moment connection with a
pre-existing flaw, both of which may be considered structural com-
ponents with a sharp crack.
Scenario 1a: Sharp-Cracked Structural Component
Subjected to Monotonic Loading
An example of this is a PJP welded butt splice (Fig. 9). In these
types of situations, the appropriate fracture model is motivated by
the extent of yielding. The specimen size limitations for the use
of linear-elastic fracture mechanics [as outlined in ASTM E399
(ASTM 2013a)] or elastic-plastic fracture mechanics [as outlined
in ASTM E1820 (ASTM 2013c)] may be used as general guide-
lines to assess the size of the structural component (given material
yield stress and fracture toughness) in the context of K IC or J IC
validity. If determined to be valid, standard LEFM or EPFM
approaches (outlined in prior discussion) are suitable, where the
toughness demand is determined from finite-element simulations
and the capacity is determined either through standardized testing
or through correlations such as the one outlined in Eq. (3).
In sharp-cracked structural components where traditional
fracture mechanics approaches are inapplicable (this is common
in structural components with sharp but shallow cracks and high
material toughness), local criteria are necessary. Specifically, this
may include the history-based VGM [Eq. (6)] or its state-based
counterpart [Eq. (11)]. Chi et al. (2006) and Kanvinde et al. (2008)
indicated that for these situations, either of these approaches is
more effective compared with traditional approaches. However,
because sharp-cracked situations involve steep strain gradients near
the crack tip, the fracture predictions are sensitive to the character-
istic distance l , which must be calibrated accurately; the aforemen-
tioned references describe procedures to accomplish this. Because
cleavage fracture is always possible, it is also prudent to apply the
corresponding criterion (e.g., the Weibull approach or the RKR
model) concurrently with the ductile fracture models.
© ASCE 03116001-11
J. Struct. Eng., 03116001
Scenario 1b: Sharp-Cracked Component Subjected to Cyclic
Loading
In this case, the geometric configuration is similar to Scenario 1a,
except that the loading is cyclic. In this situation, the nature of
cyclic loading is the primary factor in selecting the optimal mod-
eling strategy. If the situation can be categorized as high cycle
fatigue (a bridge deck under vehicular loading or a mechanical
component), then fatigue crack growth may be calculated through
established procedures such as the Paris fatigue law (Paris and
Erdogan 1963). For every computed crack length and applied load
(either an overload or the peak tensile load of the cyclic loading),
fracture may be evaluated by assuming this to be equivalent to
Scenario 1a (i.e., a precracked body under monotonic loading),
with the final cycle being considered as the monotonic load. With
the context of Scenario 1a, traditional EPFM and LEFM or local
approaches may be applied (as discussed in the previous section),
depending on the level of anticipated yielding.
If the component is subjected to earthquake-type ultralow
cycle fatigue, then large-scale yielding is anticipated even during
the cyclic loading, with a small (<20) number of cycles to failure.
This situation may be interpreted as one of fracture due to cycli-
cally induced damage, rather than one of fatigue crack growth.
In this case, local criteria such as SWDM [Eq. (10)] or CVGM
[Eq. (8)] may be applied in a manner similar to the VGM for
Scenario 1a.
Scenario Type 2: Structural Components without a
Sharp Crack
The possibility of fracture without the presence of a sharp crack is
of special interest in civil structures. This may occur in zones of
stress or strain concentration around holes or blunt notches, in
zones of local buckling, or simply in smooth structural members
subjected to large inelastic strains. The selection of modeling
strategy for structural components without a sharp crack is more
straightforward than for cracked components because traditional
fracture mechanics are invalidated at the outset due to the absence
of a sharp crack and the high likelihood of large-scale yielding.
Moreover, without the presence of a sharp crack (and the asso-
ciated increase in stress triaxiality) it is unlikely that fracture will
initiate due to cleavage (recall that high triaxiality raises the prin-
cipal stresses that are responsible for cleavage). This implies that
local criteria for ductile fracture are usually applicable. Within
these local criteria, the specific criterion may be selected based
on combination of factors including (1) nature of loading (mon-
otonic or ULCF), (2) stress state, defined by triaxiality and Lode
parameter, and (3) evolution of stress-state through loading his-
tory, affecting whether history-based or state-based criteria are
used. Criteria such as the SWDM are most general, being valid
for cyclic and monotonic loading, a range of stress states, and
also loading history effects. However, these models also require
the calibration of additional parameters, which are redundant in
specific scenarios (high triaxiality, monotonic loading). Notable
applications of these models include safety assessments of liquid
storage tanks under seismic loading (Prinz and Nussbaumer
2012) and bolted steel connections sensitive to block shear (Wen
and Mahmoud 2015).
Even in smooth components, the initiation of fracture (which
may be predicted by a local criterion for ductile fracture) reflects
material separation and the formation of a sharp crack. Subsequent
to this, fracture propagates as a sharp crack, which typically in-
cludes some extent of ductile tearing followed by transition to
cleavage.
J. Struct. Eng.
 Limitations of the Current Framework and Areas for
Future Research
The limitations of individual fracture approaches (LEFM, EPFM,
and local criteria) have been detailed in the foregoing sections. This
section focuses on issues within the modeling framework in its
entirety. These broader issues impede the widespread application
of these well-developed models within the structural engineering
profession:
1. Nonuniqueness of constitutive parameter fitting: Several models
outlined in this paper (especially the local models, which rely on
the local continuum strains) require finite-element simulations
that must be able to simulate large strains (on the order of
0.5–1.0, i.e., 50–100%), which occur in highly localized regions
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
of necking or local buckling. In turn, these finite-element simu-
lations rely on multiaxial constitutive models (typically with
a von Mises yield surface and cyclic hardening properties).
Calibration of these models for large strains (greater than 0.3
or so) cannot be performed using stress-strain curves from stan-
dard tests because the stress state in standard coupons becomes
nonhomogenous at lower strains (0.1–0.2) due to necking or
buckling. Necessitated by this, calibration is often performed
by matching load-displacement curves of calibration specimens
to those obtained through complementary finite-element simu-
lations. In these situations, multiple parameter sets produce
strain fields that match the measured load-displacement re-
sponse, resulting in nonunique parameter fits. This means that
multiple parameter fits may match the load-displacement curves
accurately, albeit with different internal strain fields. Conse-
quently, the calibrated parameter set may not (in general) reflect
true material response, leading to large errors in determination
of stress-strain fields, and consequently fracture. Cooke and
Kanvinde (2015) describe these issues.
2. Simulation of crack propagation: The primary focus on this
paper is on fracture initiation, with the assumption that in most
structural engineering situations the initiation of fracture must
be prevented for acceptable performance. However, initiation
does not always result in propagation and failure, and may be
arrested depending on the material toughness, stress gradient
(crack may run into zone of compression), strain energy stored
in the component, rate of loading, and the onset of cleavage
fracture (which is quite random). Simulating crack propagation
requires the integration of fracture criteria along with numerical
schemes that can simulate this separation within finite-element
models. This is challenging for many reasons, especially
because simulated crack propagation may be highly mesh-
dependent in terms of crack path as well as the extent of pro-
pagation itself. Some common approaches for simulating crack
propagation include the following:
• Introduction of free surfaces in the simulation model. This
may be done explicitly, for example, through the insertion
of traction-free surfaces into the model topology, with atten-
dant mathematical regularization, i.e., mitigation of stress
singularities that arise at a sharp crack tip (Baldwin and
Rashid 2013). Alternatively, this may be accomplished
through the use of a field order parameter, which establishes
a smooth transition between ruptured intact material phases
and modifies material governing equations in a manner that
reflects crack initiation and propagation. Notable references
in this area include Miehe et al. (2016), and Raina and Miehe
(2016). Ambatti et al. (2015) provides a general summary of
phase field modeling, as this approach is known.
• The use of a softening constitutive model, along with finite-
element removal. The Gurson (1977) model is possibly the
© ASCE 03116001-12
J. Struct. Eng., 03116001
most well known of these. However, significant work has
been done in developing and refining these general ap-
proaches; these have been discussed previously in this paper
in the context of coupled damage mechanics models.
• The use of cohesive zone models in which closing tractions
are applied on the crack faces behind the opened crack tip;
these provide regularization of the stress field, but suffer
from other limitations in terms of establishing fracture cri-
teria. Elices et al. (2002) provide a comprehensive review of
cohesive zone models.
3. Heterogeneity and spatial randomness of material properties:
In structural components, the areas of high stress and strain (and
high stress-strain gradients), which occur at discontinuities such
as connections, often intersect areas of highly heterogeneous
material properties due to welding. Welding introduces sharp
gradients in material toughness, such that the weld and base
metal (with higher toughness) sandwich a low-toughness heat-
affected zone. This interaction creates additional challenges for
simulation of fracture, especially using local criteria. A metho-
dology to assess fracture (with high confidence and tractability
of implementation) in these situations remains elusive.
4. Trade practices for specifying fracture toughness: The structural
engineering profession relies primarily on the Charpy V-notch
energy measured at specific temperatures as the primary indica-
tor of steel toughness. From the perspective of fracture simula-
tion, this raises two issues: (1) the standards (e.g., AISC 2010b)
establish only minimum requirements for fracture toughness,
such that the actual CVN energy is not known, and (2) even
assuming that the CVN is known, it cannot be conveniently con-
verted to an equivalent K IC or J IC, for example, due to the un-
certainty associated with the temperature shift (if the full CVN
curve is unavailable). The resulting uncertainty in the fracture
toughness results in conservative design, which while reassuring
from the standpoint of safety, is not compatible with accurate
simulation of expected structural performance. In this regard,
it is also relevant to note that nominally identical steel samples
[e.g., ASTM A572 (ASTM 2015)] often have large variation in
fracture properties necessitating the use of material-specific
toughness data for fracture assessment.
5. Standardization of local approaches: The calibration methods
for the local models and approaches to implement these models
into research and advanced engineering practice have not been
standardized. Nevertheless, best practices have been developed
for these, albeit within individual research projects; these best
practices are described in preceding sections. In this regard, it is
noted that guidelines for seismic performance assessment of
frame structures [e.g., FEMA P695 (FEMA 2009)] have greatly
accelerated the adoption of advanced simulation techniques into
mainstream research and practice. Standardization of methods
for fracture simulation in civil structures has the potential to
be similarly transformative.
Even if fracture occurs during structural failure, fracture may
not be the controlling event. For example, Fell et al. (2009) dem-
onstrated (for cyclically buckling steel braces) that local strains
increase very rapidly with respect to applied global displacement
after the onset of local buckling, with the implication that after this
instant, global displacement corresponding to fracture is virtually
insensitive to characterizations of local fracture strain (for example,
through a local criterion). In such cases, it is more beneficial to
predict local buckling accurately rather than the fracture that usu-
ally follows it. Similar situations may be encountered in tensile
loading, where the onset of necking amplifies strain, and predicting
this event (rather than the fracture that follows it) is more important.
This suggests why, for several situations (such as net section
J. Struct. Eng.
 fracture of bolted connections), the use of the ultimate strength
(associated with necking) is an excellent predictor of fracture.
In closing, it is informative to discuss the adoption of fracture
mechanics into structural engineering practice. Although this began
with the SAC Steel Project nearly two decades ago, recent years
have seen an increasing acceptance of fracture mechanics into sim-
ulation frameworks and design provisions, as well as specific con-
struction projects. For example, recent guidelines for structural
performance evaluation [e.g., ATC 114 (ATC 2016), under prepa-
ration] include explicit fracture mechanics–based equations for
simulation of structural components such as beam-column connec-
tions or splices. Similarly, design standards such as the upcoming
seismic provisions (AISC 2010b) allow the use of PJP welded spli-
ces, in part based on fracture mechanics simulations of Stillmaker
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
et al. (2016). Fracture mechanics simulation was also heavily used
for the design and material specification of key components as well
as the spire of the (currently under construction) Wilshire Grand
Tower in Los Angeles, which will become the tallest building
on the West Coast of the United States upon completion in 2017.
Summary
This paper provides an overview of the state of the art in fracture
prediction for steel structures in civil infrastructure. This consists
of models developed over decades for varied application scenarios
and within different technological and computational settings, re-
sulting in the absence of a unifying methodology for fracture pre-
diction across all scenarios. In view of this, the mission of the paper
is to contextualize what often seems like a patchwork of discordant
methodologies, with the ultimate objective of providing guidance
for effective simulation of fracture in the structural engineering pro-
fession. For this purpose, it is critical to develop an understanding
of the fracture micromechanisms themselves and the relationship of
various models to these mechanisms. The various models include
the traditional approaches of linear-elastic and elastic-plastic frac-
ture mechanics, which are phenomenological, and the more recent
local approaches, which seek to simulate the micromechanisms at a
lower scale. The local approaches do not suffer from some limita-
tions of the traditional approaches, but are challenging to effec-
tively implement within a practical setting. Optimal strategies for
predicting fracture in common structural scenarios are outlined,
recognizing trade-offs between the complexity and cost of im-
plementation and the accuracy of prediction. While major ad-
vances have been made in terms of fracture criteria for a range of
conditions (stress states, loading types, micromechanisms), the
framework itself has significant limitations. With advances in com-
putational technology, fracture mechanics continues to evolve to
address these limitations and other issues faced by the structural
engineering profession. Complementarily, because the profession
emphasizes accurate performance assessment rather than prescrip-
tive code-based design, it is anticipated that fracture mechanics will
meaningfully integrate into the design process.
Acknowledgments
The author would like to thank previous and current students and
collaborators whose work has contributed to this paper through
research or engaging discussions. These include Gregory Deierlein
of Stanford University, Benjamin Fell of California State Uni-
versity, Sacramento, Andrew Myers of Northeastern University,
Christopher Smith of the National Institute of Standards and Tech-
nology, Ryan Cooke of Forell Elsesser Engineers, and Kimberly
Stillmaker and Vincente Pericoli of the University of California
© ASCE 03116001-13
J. Struct. Eng., 03116001
at Davis. The author is also grateful to sponsors of his research over
the years, including the National Science Foundation (Grants
#CMMI 0825155, 118634, and 143400) and the American Institute
of Steel Construction. Recently, the author worked with Peter
Maranian of Brandow Johnston and Associates and Leonard
Joseph of Thornton Tomasetti Engineers for the fracture mechanics
design of the Wilshire Grand Tower; he is grateful to these indivi-
duals for stimulating discussions and collaboration. The opinions
expressed in this paper are solely of the author.
References
ABAQUS [Computer software]. Dassault Systèmes, Waltham, MA.
AISC. (2010a). “Prequalified connections for special and intermediate steel
moment frames for seismic applications.” AISC 358, Chicago.
AISC. (2010b). “Seismic provisions for structural steel buildings.” AISC
341, Chicago.
Almar-Naess, A., Haagensen, P. J., Lian, B., Moan, T., and Simonsen, T.
(1984). “Investigation of the Alexander L. Kielland failure—
Metallurgical and fracture analysis.” J. Energy Resour. Technol.,
106(1), 24–31.
Ambatti, M., Gerasimov, T., and De Lorenzi, L. (2015). “Phase-field
modeling of ductile fracture.” Comput. Mech., 55(5), 1017–1040.
American Welding Society. (2015). “Structural welding code.” AWS
D1.1:2015, Miami.
Anderson, T. L. (1995). Fracture mechanics, 2nd Ed., CRC Press,
Boca Raton, FL.
ASTM. (1999). “Standard test method for crack-tip opening displacement
(CTOD) fracture toughness measurement.” ASTM E1290, West
Conshohocken, PA.
ASTM. (2012). “Standard test method for determination of resistance to
stable crack extension under low-constraint conditions.” ASTM
E2472 12e1, West Conshohocken, PA.
ASTM. (2013a). “Standard test method for determination of reference
temperature, T o , for ferritic steels in the transition range.” ASTM
E399, West Conshohocken, PA.
ASTM. (2013b). “Standard test method for linear-elastic plane strain frac-
ture toughness K IC of metallic materials.” ASTM E1921, West Consho-
hocken, PA.
ASTM. (2013c). “Standard test method for measurement of fracture
toughness.” ASTM E1820, West Conshohocken, PA.
ASTM. (2015). “Standard specification for high-strength low-alloy
columbium-vanadium structural steel.” ASTM A572/A572M, West
Conshohocken, PA.
ATC (Applied Technology Council). (2016). “Development of accurate
models and efficient simulation capabilities for collapse analysis to
support implementation of performance based seismic engineering.”
ATC 114, Redwood City, CA.
Bai, Y., and Wierzbicki, T. (2010). “Application of extended Mohr-
Coulomb criterion to ductile fracture.” Int. J. Fract., 161(1), 1–20.
Baldwin, A. T., and Rashid, M. M. (2013). “A geometric model of deco-
hesion in solid continua.” Int. J. Fract., 180(2), 205–221.
Bao, Y., and Wierzbicki, T. (2004). “On fracture locus in the equivalent
strain and stress triaxiality space.” Int. J. Mech. Sci., 46(1), 81–98.
Barsom, J. M., and Rolfe, S. T. (1969). “Correlations between KIC and
charpy V-notch test results in the transition-temperature range.” Impact
Testing of Metals, ASTM STP 466, West Conshohocken, PA, 281–302.
Barsom, J. M., and Rolfe, S. T. (1999). Fracture and fatigue control in
structures: Applications of fracture mechanics, 3rd Ed., ASTM, West
Conshohocken, PA.
Basu, S., and Benzerga, A. (2015). “On the path dependence of the fracture
locus in ductile materials: Experiments.” Int. J. Solids Struct., 71(2015),
79–90.
Beremin, F. M. (1983). “A local criterion for cleavage fracture of a nuclear
pressure vessel steel.” Metall. Trans., 14A(11), 2277–2287.
Besson, J., Moinereau, D., and Steglich, D., eds. (2006). “Local approach
to fracture.” Euromech-Mecamat 2006, 9th European Mechanics of
Materials Conf., European Mechanics Society, Moret Sur Loing,
France.
J. Struct. Eng.
 Bridgman, P. W. (1964). Studies in large plastic flow and fracture, Harvard
University Press, Cambridge, MA.
Chi, W. M., Kanvinde, A. M., and Deierlein, G. (2006). “Prediction of duc-
tile fracture in welded connections using the SMCS criterion.” J. Struct.
Eng., 10.1061/(ASCE)0733-9445(2006)132:2(171), 171–181.
Cooke, R. J. (2015). “Micromechanical simulation of ductile fracture
processes in structural steel.” Ph.D. dissertation, Dept. of Civil and
Environmental Engineering, Univ. of California, Davis, CA.
Cooke, R. J., and Kanvinde, A. M. (2015). “Constitutive parameter cali-
bration for structural steel: Non-uniqueness and loss of accuracy.”
J. Constr. Steel Res., 114, 394–404.
Dodds, R., et al. (2012). “WARP3D-release 17.6.0: 3-D nonlinear fracture
analysis of solids using parallel computers.” Structural Research Series,
No. 607, Univ. of Illinois at Urbana–Champaign, Urbana, IL.
Elices, M., Guinea, G. V., Gomez, J., and Planas, J. (2002). “The cohesive
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
zone model: Advantages, limitations, and challenges.” Eng. Fract.
Mech., 69(2), 137–163.
Enakoutsa, K., Leblond, J. B., and Perrin, G. (2007). “Numerical imple-
mentation and assessment of a phenomenological nonlocal model of
ductile rupture.” Comput. Methods Appl. Mech. Eng., 196(13–16),
1946–1957.
Faleskog, J., Gao, X., and Shih, C. (1998). “Cell model for nonlinear frac-
ture analysis—I. Micromechanics calibration.” Int. J. Fract., 89(4),
355–373.
Fell, B. V., Kanvinde, A. M., Deierlein, G. G., and Myers, A. T. (2009).
“Experimental investigation of inelastic cyclic buckling and fracture of
steel braces.” J. Struct. Eng., 10.1061/(ASCE)0733-9445(2009)135:
1(19), 19–32.
FEMA (Federal Emergency Management Agency). (2000). “Recom-
mended design criteria for new steel moment-frame buildings.”
FEMA-350, Washington, DC.
FEMA (Federal Emergency Management Agency). (2009). “Quantification
of building seismic performance factors.” FEMA-P695, Washington,
DC.
Griffith, A. A. (1921). “The phenomena of rupture and flow in solids.”
Philos. Trans. R. Soc. London, A, 221(582–593), 163–198.
Gurson, A. L. (1977). “Continuum theory of ductile rupture by void
nucleation and growth: Part I—Yield criteria and flow rules for porous
ductile media.” J. Eng. Mater. Technol., 99(1), 2–15.
Hancock, J. W., and Mackenzie, A. C. (1976). “On the mechanics of ductile
failure in high-strength steel subjected to multi-axial stress-states.”
J. Mech. Phys. Solids, 24(2-3), 147–160.
Hutchinson, J. W. (1968). “Singular behavior at the end of a tensile crack in
a hardening material.” J. Mech. Phys. Solids, 16(1), 13–31.
Johnson, G. R., and Cook, W. H. (1983). “A constitutive model and data for
metals subjected to large strain, high strain rate and high temperature.”
7th Int. Symp. on Ballistics, American Defense Preparedness
Association, Arlington, VA, 541–547.
Kanvinde, A. M., and Deierlein, G. G. (2006). “The void growth model and
the stress modified critical strain model to predict ductile fracture in
structural steels.” J. Struct. Eng., 10.1061/(ASCE)0733-9445(2006)
132:12(1907), 1907–1918.
Kanvinde, A. M., and Deierlein, G. G. (2007). “A cyclic void growth model
to assess ductile fracture in structural steels due to ultra low cycle
fatigue.” J. Eng. Mech., 10.1061/(ASCE)0733-9399(2007)133:6(701),
701–712.
Kanvinde, A. M., Fell, B. V., Gomez, I. R., and Roberts, M. (2008).
“Predicting fracture in structural fillet welds using traditional and
micromechanics-based models.” Eng. Struct., 30(11), 3325–3335.
Kiran, R., and Khandelwal, K. (2014). “Fast-to-computer weakly coupled
ductile fracture model for structural steels.” J. Struct. Eng., 10.1061/
(ASCE)ST.1943-541X.0001025, 04014018.
Kiran, R., and Khandelwal, K. (2015). “A coupled microvoid elongation
and dilation based ductile fracture model for structural steels.” Eng.
Fract. Mech., 145, 15–42.
Kumar, V., German, M. D., Wilkening, W. W., Andrews, W. R., deLorenzi,
H. G., and Mowbray, D. F. (1984). “Advances in elastic-plastic fracture
analysis.” EPRI Final Rep. RP 1237-1, NP-3607, Electric Power
Research Institute, Palo Alto, CA.
© ASCE 03116001-14
J. Struct. Eng., 03116001
Malley, J. O. (1998). “SAC steel project: Summary of Phase 1 testing
investigation results.” Eng. Struct., 20(4–6), 300–309.
Matos, C. G., and Dodds, R. H. (2001). “Probabilistic modelling of weld
fracture in steel frame connections. Part I: Quasi-static loading.” Eng.
Struct., 23(8), 1011–1030.
McClintock, F. A. (1968). “A criterion for ductile fracture by the growth of
holes.” J. Appl. Mech., 35(2), 363–371.
McMeeking, R. M., and Parks, D. M. (1979). “On criteria for J-dominance
of crack-tip fields in large scale yielding.” Elastic Plastic Fracture,
ASTM STP 668, West Conshohocken, PA, 175–194.
Miehe, C., Teichtmeister, S., and Aldakheel, F. (2016). “Phase-field mod-
eling of ductile fracture: A variational gradient-extended plasticity-
damage theory and its micromorphic regularization.” Philos. Trans.
R. Soc. A., 374(2066), 20150170.
Myers, A. T., Deierlein, G. G., and Kanvinde, A. M. (2009). “Testing and
probabilistic simulation of ductile fracture initiation in structural steel
components and weldments.” Technical Rep. 170, John A. Blume
Earthquake Engineering Center, Stanford Univ., Stanford, CA.
Myers, A. T., Kanvinde, A. M., and Deierlein, G. G. (2010). “Calibration of
the SMCS criterion for ductile fracture in steels: Specimen size depend-
ence and parameter assessment.” J. Eng. Mech., 10.1061/(ASCE)EM
.1943-7889.0000178, 1401–1410.
Myers, A. T., Kanvinde, A. M., Deierlein, G. G., and Baker, J. W. (2014).
“A probabilistic formulation of the cyclic void growth model to predict
ultra low cycle fatigue in structural steel.” J. Eng. Mech., 10.1061/
(ASCE)EM.1943-7889.0000728, 04014028.
Nahshon, K., and Hutchinson, J. (2008). “Modification of the Gurson
model for shear failure.” Eur. J. Mech., 27(1), 1–17.
Panontin, T. L., and Sheppard, S. D. (1995). “The relationship between
constraint and ductile fracture initiation as defined by micromechanical
analyses.” Fracture Mechanics, Vol. 26, W. Reuter, J. Underwood, and
J. Newman, eds., ASTM STP16379S, West Conshohocken, PA, 54–85.
Pardoen, T., and Hutchinson, J. W. (2000). “An extended model for void
growth and coalescence.” J. Mech. Phys. Solids, 48(12), 2467–2512.
Paris, P., and Erdogan, F. (1963). “A critical analysis of crack propagation
laws.” J. Basic Eng., 85(4), 528–534.
Prinz, G. S., and Nussbaumer, A. (2012). “On the low-cycle fatigue capac-
ity of unanchored steel liquid storage tank shell-to-base connections.”
Bull. Earthquake Eng., 10(6), 1943–1958.
Raina, A., and Miehe, C. (2016). “A phase field model for fracture in bio-
logical tissues.” Biomech. Model. Mechanobiol., 15(3), 479–496.
Rice, J. R. (1968). “A path independent integral and the approximate analy-
sis of strain concentration for notches and cracks.” J. Appl. Mech.,
35(2), 379–386.
Rice, J. R., and Rosengren, G. F. (1968). “Plane strain deformation near
crack tip in power-law hardening material.” J. Mech. Phys. Solids,
16(1), 1–12.
Rice, J. R., and Tracey, D. M, (1969). “On the ductile enlargement
of voids in triaxial stress fields.” J. Mech. Phys. Solids, 17(3),
201–217.
Ritchie, R. O., Knott, J. F., and Rice, J. R. (1973). “On the relationship
between critical tensile stress and fracture toughness in mild steel.”
J. Mech. Phys. Solids, 21(6), 395–410.
Rooke, D. P., and Cartwright, D. J. (1976). “Compendium of stress inten-
sity factors.” Her Majesty’s Stationery Office (HMSO), London.
Roth, C. C., and Mohr, D. (2016). “Ductile fracture experiments with
locally proportional loading histories.” Int. J. Plast., 79, 328–354.
Rousselier, G. (1987). “Ductile fracture models and their potential in local
approach of fracture.” Nucl. Eng. Des., 105(1), 113–120.
Ruggieri, C., Panontin, T. L., and Dodds, R. H., Jr. (1996). “Numerical
modeling of ductile crack growth in 3-D using computational cell
elements.” Int. J. Fract., 79(4), 309–340.
Saxena, A. (1997). Nonlinear fracture mechanics for engineers, CRC
Press, Boca Raton, FL.
Sheldon, L. (2010). “Structural failures: Testing lessons learned.” Test Eng.
Manage., 71(6), 10–12.
Smith, C. M., Deierlein, G. G., and Kanvinde, A. M. (2014). “A stress-
weighted damage model for ductile fracture initiation in structural steel
under cyclic loading and generalized stress states.” TR 187, Blume
Earthquake Engineering Center, Stanford Univ., Stanford, CA.
J. Struct. Eng.
 Srivastava, A., and Needleman, A. (2012). “Porosity evolution in a
creeping single crystal.” Model. Simul. Mater. Sci. Eng., 20(3),
035010.
Stillmaker, K., Kanvinde, A. M., and Galasso, C. (2016). “Fracture me-
chanics based design of column splices with partial joint penetration
welds.” J. Struct. Eng., 10.1061/(ASCE)ST.1943-541X.0001380,
04015115.
Suresh, S. (1998). Fatigue of materials, 2nd Ed., Cambridge University
Press, Cambridge, MA.
Wallin, K. (1998). “Master curve analysis of ductile to brittle transition
region fracture toughness round robin data.” Technical Research Center
of Finland, Espoo, Finland.
Downloaded from ascelibrary.org by University of Waterloo on 11/12/16. Copyright ASCE. For personal use only; all rights reserved.
Wen, H., and Mahmoud, H. (2015). “Prediction of block shear fracture
in bolted connections.” 8th Int. Conf. on Advances in Steel Structures,
Hong Kong Institute of Engineers, Lisbon, Portugal.
Wen, H., and Mahmoud, H. (2016a). “New model for ductile fracture of
metal alloys. I: Monotonic loading.” J. Eng. Mech., 10.1061/(ASCE)
EM.1943-7889.0001009, 04015088.
Wen, H., and Mahmoud, H. (2016b). “New model for ductile fracture of
metal alloys. II: Reverse loading.” J. Eng. Mech., 10.1061/(ASCE)EM
.1943-7889.0001010, 04015089.
Xue, L. (2007). “Damage accumulation and fracture initiation in uncracked
ductile solids subject to triaxial loading.” Int. J. Solids Struct., 44(16),
5163–5181.

© ASCE 03116001-15 J. Struct. Eng.

J. Struct. Eng., 03116001