62e64795-8b56-4230-8bce-e3cbe3886c33

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

Ductile Fracture

The ductile fracture is a more complex process and its description requires a proper
account of dislocation nucleation and motion in the neighborhood of the crack tip.

From: Handbook of Silicon Based MEMS Materials and Technologies, 2010

Related terms:

Alloys, Nucleation, Brittle Fracture, Plastic Deformation, Crack Tips, Toughness

View all Topics

Learn more about Ductile Fracture

Damage Evolution and Ductile Fracture


Heng Li, Mingwang Fu, in Deformation-Based Processing of Materials, 2019

3.2 Mechanism of Damage Evolution and Ductile Fracture


Ductile fracture of metallic materials is a very complex phenomenon significantly
influenced by many factors, such as material state, workpiece geometry, strain path,
working temperature, and strain rate. The mechanism of damage evolution and
ductile fracture is sophisticated and has garnered much attention. As is known,
under tension-dominated deformation conditions, ductile fracture results from the
nucleation, growth, and coalescence of microvoids. It is noteworthy that the ductile
fracture mechanism under shearing-dominated deformation still has a dispute. One
accepted interpretation is that damage and fracture mainly result from shear linkup
of the voids. For deformation-based materials processing, ductile fracture mainly
originates from a mixture of tension- and shear-induced damage. In this section,
the common phenomena and mechanisms of damage evolution and ductile fracture
during deformation-based materials processing are introduced and discussed.

3.2.1 Tension-Induced Void Initiation, Growth, and Coalescence


Ductile fracture in metals and metallic alloys often originates from the initiation,
growth, and coalescence of microscopic voids during plastic deformation [1–6].
The nucleation of voids usually takes place at the interfaces of inclusions and sec-
ond-phase particles. The dissociation of these interfaces is regarded as the dominant
mechanism of void nucleation. Once the voids nucleate, further plastic deformation
enlarges the size of voids and distorts the shape, which is often called void growth
[7]. As voids enlarge and distort substantially with plastic deformation, adjacent voids
ultimately link up or coalesce with each other due to the localization of plastic strain
in the intervoid matrix, forming the final fracture surface. Void coalescence can be
observed directly by SEM analysis from fracture surfaces, since void coalescence
is the final stage in ductile fracture [8]. Void initiation, growth, and coalescence
mainly occur along the direction of maximum tension stress, as shown in Fig. 3.1.
Under the uniaxial tensile load, damage evolution and ductile fracture of steel at
room temperature (shown in Fig. 3.2 [9]) confirm this, as the fracture surface is
perpendicular to the maximum tensile stress.

Figure 3.1. Void evolution under maximum tension stress.

Figure 3.2. Diagram of void initiation, growth, and coalescence [9].

From Fig. 3.2 necking, which is a mode of tensile deformation where relatively large
amounts of strain localization appear in a small region of the material, occurs before
ductile fracture. In general, necking can be classified as diffuse necking and localized
necking. The first one is the case where the uniform reduction of thickness in a
relatively large range occurs, while the second is the case where the thinning of
material concentrates in a localized region. During deformation-based materials
processing, fracture caused by necking is an irrecoverable failure. Localized necking,
rather than diffuse necking, is an important factor that determines the amount of
useful deformation, so the point in which localized necking first occurs is regarded
as an ideal critical point.

It should be noted that fracture may occur without apparent necking for materials
with low plasticity. For this phenomenon, the mechanism of damage evolution
and ductile fracture is determined based on fractography analysis. For different
materials, the mechanisms are different. Al-alloy contains a large-volume fraction
of various intermetallic particles. In its microstructure, two types of inclusions and
particles in brittle phases are dispersed in the Al-matrix: Fe-based intermetallics and
Mg2Si intermetallics [4]. Due to the incompatibility of the Al-matrix and intermetallic
particles with different properties, the voids are initiated or nucleated primarily
at the grain boundaries once plastic deformation occurs. These characteristics are
convenient for exploring the mechanisms of damage evolution and ductile fracture.
As can be seen from situ tensile tests of the Al-alloy, dimple-dominant fracture
usually occurs under the tension-stress-dominant condition for ductile materials,
as shown in Fig. 3.3. Under tension-dominant deformation, a higher triaxiality
accelerates void initiation and propagation, which results in a large number of
dimples and cracks perpendicular to the maximum tensile stress.

Figure 3.3. Dimple-dominant fracture caused by the maximum tension stress [10].

To describe the damage evolution during plastic deformation and predict ductile
fracture, several indicators related to voids have been studied and used to model
fracture. Void-volume fraction is widely used [1,3,11–13], especially the Gurson-type
fracture model. As shown in the X-ray microtomography in Fig. 3.4, the evolution of
voids and the void-volume fraction during deformation can be obtained, as shown
in Fig. 3.5. The volume fraction usually slowly increases first, before rapidly increas-
ing, which can be attributed to the coalescence of voids under large plastic strain
[12,13]. Rapid increase of the void-volume fraction signifies that necking has already
appeared. Recently, the evolution of void shape and orientation during deformation
has also been considered to improve the prediction accuracy of ductile failure under
shear-dominated loadings [14,15]. Those have greatly promoted understanding of
the effects of void evolution on ductile fracture.

Figure 3.4. Schematic view of laminography setup [16].

Figure 3.5. Evolution of (A) three chosen voids and (B) the void-volume fraction
during deformation [13,17].

3.2.2 Shear-Stress-Induced Damage and Ductile Fracture


Shear deformation often occurs during deformation-based materials processing.
Nevertheless, the mechanism of shear-stress-induced damage and ductile fracture
is still controversial. One of the main points is that void linking leads to damage and
fracture under shearing-dominated loadings [8,18,19]. SEM analysis of fractured
surfaces for DP980 steel and AA7075 alloy indicates that fractured voids tend to
elongate along the direction of the maximum shear stress in various stress states
[8], as shown in Fig. 3.6.

Figure 3.6. Schematic illustration of shear linkup of voids.

In situ tensile tests of Al-6061 alloy also reveal that shear linkup of voids is the main
cause of void coalescence [10]. As the voids nucleate and grow with a lower growth
rate, shear linkup of voids is the main mechanism of fracture. A shear band with
minor voids was observed, as depicted in Fig. 3.7. The tests confirmed that void
linkup had significant effects on damage evolution. The voids’ deformation mode
is very sensitive to stress triaxiality [20]. Void linkup is strongly affected by numerous
factors: void shape, relative void spacing, the nucleation of secondary voids, and the
surface contact of flattened voids at low or negative stress triaxiality [8].

Figure 3.7. Shear-deformation-induced ductile fracture [10].

3.2.3 Damage Caused by a Mixture of Tension and Shear Stress


Under deformation-based processing, the materials are usually subjected to a
mixture of tensile and shear stresses. The effects of the maximum tensile stress
and shear stress on void evolution are different, which may lead to difference in
macrofracture mode [10], as shown in Figs. 3.3 and 3.7. Under this mixed stress
state, the voids elongate along the direction of the maximum tensile stress and
shear linkup occurs simultaneously, as shown in Fig. 3.8. In situ tensile tests of
Al-6061 alloy, as shown in Fig. 3.9, illustrate the presence of dimples with shear
bands in the center region of the fractured surface. It should be noted that the
mixture of tension- and shear-induced damage is the main mechanism of ductile
fracture during deformation-based materials processing.

Figure 3.8. Void evolution caused by a mixture of tension and shear stress.
Figure 3.9. Ductile fracture caused by a mixture of tension and shear stress [10].

> Read full chapter

Fracture processes of aerospace materi-


als
In Introduction to Aerospace Materials, 2012

Crack tip stress and plastic zone


Ductile fracture is the most common failure mode in aerospace metal alloys and
polymers (including structural adhesives). A ductile crack in a metal usually starts
at an existing flaw, such as a brittle inclusion within a grain, a precipitate at a grain
boundary, or a void. The stress condition ahead of a crack in a ductile material loaded
in tension is shown schematically in Fig. 18.3. The stress ahead of the crack is not
distributed evenly. Instead, the stress in the region immediately ahead of the crack
tip is much higher than the nominal (applied) stress. Figure 18.3 shows that the
closer one approaches the crack tip, the higher the local stress becomes until, at some
distance ry from the crack the stress reaches the yield strength y of the material. The
material over a distance between the crack tip (r = 0) and ry is plastically deformed,
and this region is called the plastic zone.
18.3. Stress field ahead of the main crack front in a ductile material, showing the
plastic zone (shaded).

The size of the plastic zone is dependent on the yield strength of the material, the
applied stress level, and the load conditions (e.g. tension, shear). The plastic zone
can range in size from a few tens of micrometres in high-strength metals to many
millimetres in soft plastics. The material outside the plastic zone is not stressed
above the yield strength and, therefore, does not plastically deform. Owing to the
formation of the plastic zone, which absorbs energy and thereby resists crack growth,
the stress needed to initiate a crack in ductile materials is lower than the stress
needed to grow the crack. In other words, it is easier to form a crack than to get
the crack to grow and, therefore, the applied stress needed to cause crack growth
increases with the length. This behaviour provides ductile materials with an intrinsic
amount of damage tolerance because, as the crack becomes larger, it also becomes
more difficult to grow to the critical size necessary to cause complete failure.

> Read full chapter

Practical Failure Assessment Methods


Y. Takahashi, in Comprehensive Structural Integrity, 2003

7.10.4.3.2 Unstable fracture


Ductile fracture from nonpenetrating inner-surface cracks has received much atten-
tion for evaluating the possibility of break-before-penetration. Many tests have been
conducted for pipes with an initial notch in the inner surface in various organizations
(e.g., Roos et al. 1989; Kurihara et al., 1988; Krishnaswamy, 1995). A bending moment
with or without superposed internal pressure was applied to the test pipes in many
cases.

It should be noted that there are two definitions for the limit load for plates and
cylinders with a surface crack. These can be designated as “global” and “local” limit
loads, which correspond to collapse of the whole body and attainment of a fully
yielded state only along the ligament at the deepest point of the crack, respectively.
A lower bound solution for the global limit load can be obtained relatively easily
by considering the equilibrium of axial and bending loads and achievement of a
fully yielded state at the cracked section (e.g., Zahoor, 1989). The solution for the
constant-depth crack shown in Figure 20 can be written as

Figure 20. Geometry of constant-depth surface crack.

(115)

with

(116)

This limit load solution and its modification by Kurihara et al. (1988) by considering
a difference in the flow stress at the radial and circumferential ligaments, were
used in the development of J-estimation methods for finite-length constant-depth
cracks (Krishnaswamy, 1995). The predicted maximum loads were compared with
over 40 test results and the conservatism of the J-integral approach was verified. The
net-section collapse criterion with the global limit loads of Equations (115) and (116)
tended to overpredict the maximum loads under combined moment and internal
pressure loading, whereas Kurihara's equation gave better agreement with the test
results. Another failure criterion based on an elastic stress distribution was also
proposed and termed the moment method (Roos et al., 1989). It was reported that
for ferritic steels with relatively low toughness this method gave better predictions
for the global limit loads than the conventional net-section collapse criterion of
Equations (115) and (116). Based on a comparison of the fracture strength of pipes
with through-wall cracks and those with part-through cracks, LBB diagrams were
developed (Stadmüller and Sturm, 1997). The LBB diagram inherently assumes that
part-through cracks extend predominantly in the radial direction before penetration
and start to grow in the circumferential direction afterwards. Figure 21 shows one
example of such a diagram for a ferritic steel. Although a large break region seems
to exist for shallower cracks, the actual applied loads may be smaller than the loads
required for fracture in such cases. For example, at a bending moment of 500 kNm
in Figure 21, a break is expected to occur for cracks larger than a depth of 0.7t and
an angle of 150°. The presence of a hoop stress generated by the internal pressure is
expected to have some influence on the limit load. Desquines et al. (1995) evaluated
this effect and it was shown that a hoop stress exceeding the yield stress has a
nonnegligible effect and reduces the limit load, especially for shallow cracks. This
effect can be neglected when the pressure is low or the crack is deeper than 0.5t.

Figure 21. LBB assessment diagram. (Source: Stadmüller and Sturm, 1997.)

The J-integral can also be applied for ductile fracture evaluation of pipes with a
surface crack, in principle. However, as of early 2000s, available solutions are very
limited. For semi-elliptical cracks in a plate subjected to global tensile and bending
load, fully plastic solutions for the J-integral were obtained and interpolated by
polynomial equations of a/t and a/c (Yagawa and Ueda, 1989). Fully plastic solutions
were also obtained for a constant-depth inner-surface crack in a cylinder (Kumar
and German, 1988), but they are very limited. A simplified approach, such as the
reference stress method, can also be applied by using the stress intensity factor
and the limit load solution. In this case, the local limit loads are considered more
appropriate because the J-integral is a local parameter changing along the crack
front. Though solutions for the local limit load are very limited, one such solution
was obtained by Sattari-Far (1994) using the FE method for semi-elliptical cracks in
a plate under tensile and bending load. However, later studies showed that the use
of a global limit load solution by Goodall and Webster (2001) brings about better
estimates for the J-integral than the local limit load solution by Sattari-Far (1994) for
plates with a semi-elliptical crack under tension, bending, and combined load (Lei,
2001a, 2001b, 2002).

> Read full chapter

Creep Damage Mechanics


In High Temperature Deformation and Fracture of Materials, 2010

20.3.3 Ductile-Brittle creep fracture


Pure ductile fracture and pure brittle fracture are two extreme situations. The real
creep fracture of engineering materials lies between these two cases. The damages
by the reduction of cross-sectional area caused by creep deformation and by the
reduction of effective load-supporting area caused by the internal damage of the
material both develop simultaneously during creep.

Kachanov assumed that the creep is independent of damage evolution. This means
that the creep constitutive relation still agrees with the power law equation , where
a is an instantaneous stress which is related to the reduction of the cross-sectional
area caused by creep deformation but without regard for the increase of the effective
stress resulting from the internal damages, i. e.

(20.22)

Substituting Eq. (20.16) into Eq. (20.22) gives an expression of . Inserting the
expression into Eq. (20.18), we can obtain the ductile-brittle rupture time tr3 as

(20.23)

Since Eq. (20.16) is valid at t≤ trl , Eq. (20.23) is also valid at tr3≤ trl , thus

(20.24)
Figure 20.1 shows schematically the curves of stress-rupture life based on the three
failure modes mentioned above. The two straight lines represent the pure ductile
fracture (line A) defined by Eq. (20.17) and pure brittle fracture (line B) defined by
Eq. (20.21), and curve C represents the ductile-brittle fracture. Many experimental
results suggest that the stress-rupture time curves are similar to curve C (see Figs.
21.6 and 21.8 in Chapter 21). The experimental results also show that at high stresses
the creep rupture elongation r is high, but decreases with the decrease of the stress,
indicating that the failure mode is close to ductile fracture at higher stresses and to
brittle fracture at lower stresses. These experimental results are consistent with the
ductile-brittle creep fracture theory mentioned above.

20.1. The curves of stress-rupture life based on the ductile, brittle and ductile-brittle
failure modes.

The fact that curve C is upper convex indicates n> m. For the special case of n= m,

(20.25)

where constant k is given by

(20.26)

and the rupture strain (elongation) is a constant

(20.27)

> Read full chapter

MODELS OF FIBRE FRACTURE


M. Elices, J. Llorca, in Fiber Fracture, 2002

Ductile Behaviour
Modelling fibre ductile fracture is an involved problem, even for homogeneous and
macro-defect-free fibres. Under tensile loading, fibres eventually reach an instability
point, where strain hardening cannot keep pace with loss in cross-sectional area,
and a necked region forms beyond the maximum load. A central crack is nucleated,
spreads radially and finally, when approaching the fibre surface, propagates along
localised shear planes, at roughly 45° to the axis, to form the ‘cone’ part of the
fracture. This is the typical cup-and-cone fracture of a ductile failure after a tensile
test.

Exceptions to this form of ductile fracture appear in microstructurally clean high-pu-


rity metal fibres, which sometimes neck down to a point. On the contrary, in ductile
fibres containing large defects, or deep surface notches, necking is absent and a
brittle fracture occurs. Ruptures under tensile loading and strong transversal forces,
as in drawing fibres through dies, are also different (see the paper by Künzi, section
‘Drawing Defects, Nonhomogeneous Microstructure and Texture’, and the paper by
Yoshida, both in this volume). See, for example Elices (1985), for an overview of
fracture of steel wires under different loadings.

The physics of ductile fracture exhibit the following stages: formation of a free
surface at an inclusion, or second-phase particle, by either interface decohesion or
particle cracking, growth of the void around the particle by means of plastic strain
and hydrostatic stress, and coalescence of the growing void with adjacent voids,
forming a microcrack. When inclusions and second-phase particles are strongly
bonded to the matrix, void nucleation is often the critical step and brittle fracture
occurs after void formation. When void nucleation is easy, the fracture behaviour
is controlled by the growth and coalescence of voids: growing voids reach a critical
size, relative to their spacing, and a local plastic instability develops between voids
forming a macroscopic flaw, which leads to fracture. These three steps, nucleation,
growth and coalescence of voids, occur in highly stressed regions of the fibre: in
the necking zone, in smooth and perfect fibres, (Fig. 6a) or in stress concentrators,
near some cracks or notches (Fig. 6b); the triaxiality ahead of the crack tip provides
sufficient stress elevation for void nucleation, so growth and coalescence of
microvoids are usually the critical steps in ductile crack growth.
Fig. 6. Nucleation, growth and coalescence of voids: (a) in the necking zone; (b) near
cracks and notches.

Modelling ductile fibre failure, using the continuum approach, should consider all
these facts. A number of models for estimating void nucleation stress have been
published; among them, those of Argon et al. (1975) and Beremin (1981) are often
used. The most widely referenced models for growth and coalescence of voids were
published by Rice and Tracey (1969) and Gurson (1977). Rice and Tracey considered
a single void in an infinite solid with a rigid-plastic and a linear strain hardening
behaviour. Gurson analysed plastic flow in a porous medium assuming that the
material behaves as a continuum and the effect of voids is averaged. The main
difference between this model and standard plasticity is that the yield surface in
the Gurson model exhibits a weak hydrostatic stress dependence. Ductile fractures
are assumed to occur as a result of a plastic instability that produces a band of
localised deformation. The Gurson model, and later improvements (Tvergaard and
Needleman, 1984 for example), characterise plastic flow quite well in the early
stages of the ductile fracture process, but do not provide a good description of
the events that lead to final failure, because of not containing discrete voids. These
shortcomings are intended to be surmounted by the model of Thomason (1990),
where holes are explicitly considered.

Once the crack is nucleated, crack growth can be modelled using the above-men-
tioned models, or from a more macroscopic point of view by using the techniques
of elasto-plastic fracture mechanics (EPFM) (see, for example, Anderson, 1995, and
Broberg, 1999). To the authors’ knowledge, few results are available of ductile fibre
failure based on EPFM, most probably because the small size of fibre diameters
invalidates some of the hypotheses on which this theory is based.

Fracture of thin metallic wires during cold drawing is also worth mentioning here.
Wire breakage in the industrial practice of wire drawing is often due to the presence
of non-metallic inclusions and a severe plastic deformation; the wire is pulled
through the dies and highly stressed, withstanding large plastic deformations. Voids,
nucleated at the interfaces between inclusions and matrix, generate cracks that
eventually lead to fracture. A physical insight into void initiation during wire drawing
of pearlitic steels, an important part of commercial metallic fibres, appears, for
example, in Nam and Bae (1995). At high strains, globular cementite particles whose
size is much larger than the thickness of cementite lamella, provide sites for void
formation due to the enhanced stress concentration. These observations may be
helpful when modelling fibre fracture at the mesolevel. Numerical predictions of
the rupture stress and initiation sites can be obtained using finite element methods
(FEM), where elasto-plastic behaviour of elements is complemented by a fracture
criterion. In general, these fracture criteria have the following functional form:

(5)

where is the effective strain, f the effective fracture strain and C is a parameter,
usually known as a damage parameter, obtained experimentally. Different forms of
the integrand are summarised in Doege et al. (2000). Yoshida, in his paper in this
volume, summarises the FE modelling of a superfine wire with a cylindrical inclusion
placed on the wire axis.

> Read full chapter

Weld metal ductility and its influence


on formability of tailor welded blanks
A.A. Zadpoor, J. Sinke, in Failure Mechanisms of Advanced Welding Processes, 2010

Physical models of ductile fracture


Physical models of ductile fracture work directly with nucleation, growth and coa-
lescence of voids. They are also known as Gurson type models (Gurson, 1977). The
von Mises yield function is modified to account for the presence and evolution of
voids. The modified yield function can be stated as follows (Tvergaard, 1987):

[10.19]

where eq, H, and y are, respectively, the effective Mises stress, hydrostatic pressure
and yield stress of the fully dense matrix material. The parameter fvoid is the ratio of
the volume of voids to the total volume of the material and is called the void volume
fraction. Three remaining parameters, namely and , are experimentally determined
material parameters. It is assumed that there is an initial population of voids in
the material that is represented by the initial void volume fraction, fvoid,0. As the
material deforms, voids nucleate, grow and coalesce, resulting in contraction of the
yield locus. The evolution, of the void volume fraction is due to two contributing
phenomena, namely void growth and void nucleation. Therefore, the time derivative
of the void volume fraction can be calculated as:

[10.20]

where subscripts ‘gr’ and ‘nucl’ stand for void growth and void nucleation, respec-
tively. Based on the mass conservation law, the change in the void volume fraction
due to void growth can be described as:

[10.21]

where I is the identity matrix. The void nucleation dynamics is governed by the
relationship:

[10.22]

Where

[10.23]

In Equation [10.23], it is assumed that the nucleation strain is normally distributed


within the material with a mean value of N and a standard deviation of SN. The
volume fraction of nucleated voids is denoted by fvoid,N. Voids are assumed to
nucleate only in tension.

The constitutive equations discussed above are used for modeling TWB during the
forming process. A failure criterion is also needed to detect the failure during the
forming process. The yield function needs to be modified such that the effects of
void evolution and the resulting loss of the stress carrying capacity are taken into
account. The yield function can be then rewritten as:

[10.24]

where the effective void volume fraction, fvoid , is defined as a function of the void
volume fraction, fvoid, as follows:

[10.25]

Where

[10.26]
The critical void volume fraction, fvoid,c, is the void volume fraction at which the loss
of stress carrying capacity starts. The material continues to lose its stress carrying
capacity until the void volume fraction reaches the failure void volume fraction,
fvoid,f, at which point the material fails. The Gurson-type constitutive equations are
often combined with Marciniak–Kuczynski or bifurcation models to predict the
bifurcation localization which is often assumed to be responsible for failure. Bayley
and Pilkey combined the Gurson-type constitutive equations with a bifurcation
criterion to study the influence of welding effects on the localization behavior and
predict weld line failures of an aluminum TWB made from AA5754-O sheets (2.1
and 1.6 mm) (Bayley and Pilkey, 2005, 2006). The sheets were butt welded together
by non-vacuum electron beam welding. Chien et al. (2003) combined Gurson-type
constitutive equations with Marciniak–Kuc-zynski theory to predict the failure of a
laser-welded aluminum (AA5754) TWB.

> Read full chapter

FAILURES OF MATERIALS
HUY DUONG BUI, ... N. STALIN-MULLER, in Handbook of Materials Behavior
Models, 2001

7.3.9 DUCTILE FRACTURE


The first approach to ductile fracture, that is, fracture of materials undergoing
considerable plastic deformation prior to failure, was based on Rice's J-integral.
Indeed, the property of invariance of J with respect to the integration path remains
true in nonlinear elasticity, which made it tempting to apply it to problems involving
plasticity (unloading effects being disregarded). In this approach, propagation was
assumed to occur when some critical value of J was reached, this critical value being
allowed to depend upon the crack length. This theory was later named the global
approach to ductile rupture, since it did not rely on any detailed micromechanical
analysis of the mechanism of ductile rupture, that is, nucleation, growth, and finally
coalescence of voids, through breaking of the inclusions-matrix interfaces or the
inclusions themselves and subsequent plastic flow of the matrix. In contrast, the
more modern theory of ductile rupture, named the local approach, is based on such
an analysis. As years passed, the superiority of the latter approach has become clear,
although the older one is still widely used in practical problems. A first, a seminal
contribution was made by Rice and Tracey [16]; it consisted of an approximate
analysis of the growth of a void in an inifinite plastic matrix loaded arbitrarily
at infinity. Later, approximate criteria for porous plastic solids were proposed by
Rousselier [14], who extended Rice and Tracey's void model, and Gurson [9], who
proposed a model based on an approximate limit analysis of a typical elementary cell
in a porous medium (hollow sphere). This model has become very popular. It was
recently extended by Gologanu et al. [8] to incoporate void shape effects, which were
neglected in Gurson's model of spherical voids. Void growth in plastic solids is thus
now fairly understood and described by suitable models. Therefore, the theoretical
analysis of coalescence has now become the major challenge in the local approach
to ductile rupture.

This phenomenon is very complex and is influenced by numerous factors, such as


void shape, inhomogeneities in the distribution of cavities, the presence of a second
population of secondary, smaller voids, etc. Significant contributions have already
been made in this direction, but much remains to be done.

> Read full chapter

Ductile Fracture: Micromechanisms


V. Tvergaard, in Encyclopedia of Materials: Science and Technology, 2001

In metals, the most important ductile fracture mechanisms depend on the growth
of small voids until failure occurs by void coalescence. Some porosity may be present
before loading is applied, but most voids nucleate at second phase particles, by
decohesion of the particle–matrix interface or by particle fracture. In the material
around the voids, large plastic straining develops during the growth process and the
final coalescence by necking of the ligaments between voids. Early micromechanical
studies considered the growth of a single void in an infinite elastic–plastic solid,
and used the results to obtain simple estimates of the critical strain for coalescence.
Subsequently, most work has also incorporated the effect of interaction between
neighboring voids.

> Read full chapter

Numerical and Computational Meth-


ods
G.E. Stavroulakis, ... L. Ziemianski, in Comprehensive Structural Integrity, 2003

3.13.4.2.2 Parameter identification for ductile materials and


fracture
Widely used isotropic elastoplastic models for simulating ductile fracture and shear
localization processes in various metals are derived from Gurson's proposal (Gurson,
1977) (see Chapter 2.03), which describes the progressive degradation of the material
strength in the process zone by microvoid growth up to coalescence. Main improve-
ments have been introduced by Tvergaard (1990) and Tvergaard and Needelman
(1995). A comprehensive review of alternative formulations for poroplastic materials,
including porosity in geomaterials, can be found in Mahnken (2002).

Direct identification of elastoplastic material models like these, based on local


stress–strain input–output, has been also proposed (Furukawa et al., 2002), but in
most cases recourse should be made to indirect techniques requiring the modeling
of the experimental test. Main difficulties in this respect are related to the broad
anelastic phenomena and to the geometrical nonlinearity which are basic aspects
of this structural problem. The computational burden associated to the recursive
solution of the direct problem and evaluation of the sensitivity matrix in the inverse
procedure is then particularly heavy.

Mahnken (1999) has addressed the problem by specializing to this context the
FE-oriented identification technique which exploits the algorithmic tangent matrix,
proposed by Mahnken and Stein (1996) within a least-squares minimization ap-
proach. The quite general underlying methodology has been applied also to different
related fields, such as gradient enhanced damage mechanics (Mahnken and Khul,
1999), thermoelastic damage (Mahnken, 2000), and fluid saturated porous media
(Mahnken and Steinmann, 2001).

Alternative mathematical model and identification procedure for elastoplastic solids


has been formulated as an optimization problem by Yoshida et al. (1998), using
noiseless computational information from bending test as main source of infor-
mation. The difficulties related to the complex nature of the experimental tests
to be simulated have been partially circumvented by introducing an approximate
response function, defined in the space of the sought parameters and obtained
through Lagrangian interpolation of a few direct (forward) analyses carried out for
fixed parameter values, chosen in a reasonable (in a Bayesian sense) rectangular
domain. The main drawback of this approach is that computing costs increase
exponentially with the number of the unknown parameters and with the degree of
the interpolation which has a strong influence on the accuracy of the estimation.

The same strategy has been used by Aoki et al. (1997) and by Corigliano et al. (2000),
in conjunction with Kalman filter methodology, to identify the parameters entering
the generalized Gurson model. The experimental information processed by the filter
was obtained from plate specimens with a central hole or with double-side round
notches under tension (see Figure 22) and from single notched specimens under
three-point bending. Results pointed out the central role in this context of the proper
experiment design and of the “expert” initialization of the identification procedure.

Figure 22. Plate specimens with a central hole or with double-side round notches
under tension used in material identification experiments.

Difficulties in converging toward the optimal solution due to noisy data and
inaccurate model equations suggested the use of robust optimization techniques
based on (zeroth order) evolutionary algorithms (GA, NN), even for parameter
identification procedures based on the direct knowledge of stress–strain responses
(Furukawa and Yagawa, 1997, 1998).

> Read full chapter

Inverse Analysis
G.E. Stavroulakis, ... L. Ziemianski, in Reference Module in Materials Science and
Materials Engineering, 2016

4.2.2 Parameter identification for ductile materials and fracture


Widely used isotropic elastoplastic models for simulating ductile fracture and shear
localization processes in various metals are derived from Gurson s proposal (Gur-
son, 1977), which describes the progressive degradation of the material strength
in the process zone by microvoid growth up to coalescence. Main improvements
have been introduced by Tvergaard (1990) and Tvergaard and Needelman (1995). A
comprehensive review of alternative formulations for poroplastic materials, includ-
ing porosity in geomaterials, can be found in Mahnken (2002).

Direct identification of elastoplastic material models like these, based on local


stress–strain input–output, has been also proposed (Furukawa et al., 2002), but in
most cases recourse should be made to indirect techniques requiring the modeling
of the experimental test. Main difficulties in this respect are related to the broad
inelastic phenomena and to the geometrical nonlinearity which are basic aspects
of this structural problem. The computational burden associated to the recursive
solution of the direct problem and evaluation of the sensitivity matrix in the inverse
procedure is then particularly heavy.

Mahnken (1999) has addressed the problem by specializing to this context the
FE-oriented identification technique which exploits the algorithmic tangent matrix,
proposed by Mahnken and Stein (1996) within a least-squares minimization ap-
proach. The quite general underlying methodology has been applied also to different
related fields, such as gradient enhanced damage mechanics (Mahnken and Khul,
1999), thermoelastic damage (Mahnken, 2000), and fluid saturated porous media
(Mahnken and Steinmann, 2001).

Alternative mathematical model and identification procedure for elastoplastic solids


has been formulated as an optimization problem by Yoshida et al. (1998), using
noiseless computational information from bending test as main source of infor-
mation. The difficulties related to the complex nature of the experimental tests
to be simulated have been partially circumvented by introducing an approximate
response function, defined in the space of the sought parameters and obtained
through Lagrangian interpolation of a few direct (forward) analyses carried out for
fixed parameter values, chosen in a reasonable (in a Bayesian sense) rectangular
domain. The main drawback of this approach is that computing costs increase
exponentially with the number of the unknown parameters and with the degree of
the interpolation which has a strong influence on the accuracy of the estimation.

The same strategy has been used by Aoki et al. (1997) and by Corigliano et al. (2000),
in conjunction with Kalman filter methodology, to identify the parameters entering
the generalized Gurson model. The experimental information processed by the filter
was obtained from plate specimens with a central hole or with double-side round
notches under tension (see Figure 23) and from single notched specimens under
three-point bending. Results pointed out the central role in this context of the proper
experiment design and of the ‘expert’ initialization of the identification procedure.
Figure 23. Plate specimens with a central hole or with double-side round notches
under tension used in material identification experiments.

Difficulties in converging toward the optimal solution due to noisy data and
inaccurate model equations suggested the use of robust optimization techniques
based on (zeroth order) evolutionary algorithms (GA, NN), even for parameter
identification procedures based on the direct knowledge of stress–strain responses
(Furukawa and Yagawa, 1997, 1998).

Significant computational gains can be achieved by the implementation of the


model reduction techniques introduced in Section 3.6. The effectiveness of this
approach has been tested by Bolzon and Talassi (2013) on the identification problem
of seven independent parameters entering the constitutive model of anisotropic
metals, on the basis of experimental information collected from an indentation test
schematized in Figure 24.

Figure 24. Finite element simulation model of a standard (Rockwell) indentation test.

The mechanical response of the considered material class is linear elastic, represent-
ed by the relationships:
within the domain:

defined by the equivalent stress and strain values, eq and eq respectively, where:

[E1]

An expression similar to [E1] defines eq.

The dimensionless parameters like Rx and Rxy represent the ratio of the yield limits
(under uniaxial loading along x) and (under pure shear in the plane x,y) and the
(immaterial) reference value Y, namely: ; .

The special case of transversal isotropy about the x–axis (in the assumed x−y−z
reference system, see Figure 24) was selected for demonstrative purpose. Thus, the
independent constitutive parameters are: the elastic moduli Ex, Ey=Ez, and Gxy=Gxz;
the yield limits , , and the hardening exponent n.

The lateral contraction ratios ( )yz and ( )xy=( )xz (being ( )ij/Ei=( )ji/Ej, i, j=x, y,
z) were supposed to be a priori known, due to their known poor influence on
indentation results (Bolzon et al., 2004), while the shear modulus and yield limit in
the transversal isotropy plane result: 2Gyz=Ey/(1+( )yz); .

The experiment was simulated by the finite element (FE) method, taking into account
material and geometry non–linearity, including the effects of large plastic defor-
mations and frictional contact. The relatively large 3D discretization included some
10 000 degrees of freedom and required 10–20 min computing time for a single
analysis.

The POD–RBF reduced model of the problem was trained on the results of 794
different combinations of parameter values randomly distributed within the range
defined in Table T1, defining almost isotropic as well as largely anisotropic respons-
es.

Table T1. Range of the investigated material properties

150≤Ex≤250 GPa 150≤Ey≤250 GPa 70≤Gxy≤110 GPa

0≤n≤0.477

The admissible yield limit combinations are represented by the dots in Figure 25.
Notice that the points do not span the whole investigated domain due to the physical
constraint of positive-definiteness of the quadratic form [E1].
Figure 25. POD-RBF training (dots) and verification (stars) yield limit combinations.

Further parameter combinations (indicated by stars in Figure 25) were generated for
verification purposes.

In the POD–RBF approximation, the indentation curves were reconstructed starting


from four basis vectors, the imprints were recovered by the combination of twenty
normalized elementary deformation modes. The accuracy associated to this selec-
tion can be appreciated in Figure 26, which visualizes the profile of a residual imprint
evaluated by FE simulation and the corresponding one resting on the parameters
identified through the POD–RBF–based inverse analysis procedure.

Figure 26. Profiles of the residual imprint obtained by FE simulation with input
values Ex=205 GPa, Ey=199 GPa, Gxy=96.4 GPa, , , , n=0.049 and the corresponding
reconstruction based on the parameter values identified by means of the POD-RBF
approximation.

It is worth noticing that the number of computations defining the POD–RBF ap-
proximation is comparable to the number of simulations required by a single inverse
analysis exercise performed by a first-order (gradient) optimization algorithm. In the
considered case, in fact, the average number of iterations was 15, 8 computations
were required at each iterations to evaluate the derivatives by a finite difference
scheme, and the optimization algorithm was initialized by 10 different trial vector
in the parameter space, in view of the possible convergence toward local minimum
points and of the possible existence of multiple solutions. It is also worth observing
that the cubic Lagrange interpolation of this system response would require more
than 16 000 preliminary analyses.

> Read full chapter

ScienceDirect is Elsevier’s leading information solution for researchers.


Copyright © 2018 Elsevier B.V. or its licensors or contributors. ScienceDirect ® is a registered trademark of Elsevier B.V. Terms and conditions apply.

You might also like