You are on page 1of 14

Theoretical and Applied Fracture Mechanics 108 (2020) 102636

Contents lists available at ScienceDirect

Theoretical and Applied Fracture Mechanics


journal homepage: www.elsevier.com/locate/tafmec

Predicting ductile fracture in ferrous materials during cylinder upsetting T


using an ellipsoidal void model
Kazutake Komori
Department of Mechanical Systems Engineering, School of Engineering, Daido University, 10-3 Takiharu-town, Minami-ward, Nagoya-city, Aichi-prefecture 457-8530,
Japan

ARTICLE INFO ABSTRACT

Keywords: The ductile fracture in ferrous materials during cylinder upsetting was predicted using an ellipsoidal void model
Ductile fracture to show the validity of the proposed model. The simulation and experiment of cylinder upsetting were performed
Void model using five types of ferrous bars. Seven kinds of prestrains different in magnitude were introduced in the bars by
Steels drawing, and special attention was paid to the effect of the prestrain on ductile fracture during cylinder up-
Cylinder upsetting
setting, which was scarcely paid attention to in earlier studies. A specimen for the plane-strain tensile test
Prestrain
obtained from a bar was proposed to obtain a material constant on ductile fracture. Both in the simulation and in
the experiment, with increasing the prestrain, the axial strain at fracture initiation monotonically increased,
whereas the circumferential strain at fracture initiation monotonically decreased. The effect of the prestrain on
the circumferential and axial strains at fracture initiation calculated using an ellipsoidal void model substantially
agrees with that obtained experimentally, whereas the effect of the prestrain on the circumferential and axial
strains at fracture initiation calculated using conventional ductile fracture criteria considerably differs from that
obtained experimentally. Hence, the validity of the ellipsoidal void model was confirmed.

1. Introduction meaning microscopically.


A great number of studies have been performed on the ductile
Ductile fracture, which occurs when a material is subjected to large fracture of materials using a representative volume element [12,13],
plastic deformation, is undesirable during metal-forming processes. which is composed of a void and numbers of finite elements that sur-
Numerous ductile fracture criteria have been proposed for various round the void. Although the void coalescence is evaluated using a
materials. However, a ductile fracture criterion which is applicable to criterion which has a definite physical meaning microscopically, the
all metal-forming processes is yet to be developed [1,2]. number of voids is usually only one in the representative volume ele-
Ductile fracture is a microscopic phenomenon because ductile ment. A material which is subjected to metal-forming operations gen-
fracture occurs through nucleation, growth, and coalescence of voids erally contains a large number of voids. Therefore, it is unrealistic to
[3]. However, the ductile fracture criteria which are widely used for predict the ductile fracture of a material which has a large number of
metal-forming processes [4–7] are derived from a macroscopic view- voids using a large number of representative volume elements due to
point. Hence, it is challenging to improve the prediction accuracy for a the required computational time.
microscopic ductile fracture phenomenon using a macroscopic ductile In metal-forming processes, it is well known that localized de-
fracture criterion. formation occurs and that plane-strain deformation appears irrelevant
A large number of studies have been performed on the ductile to the prior deformation mode, immediately before ductile fracture
fracture of materials using the Gurson yield function [8]. Although occurs. For instance, during the uniaxial tensile test of a sheet metal, the
these studies simulate the nucleation and growth of voids, these studies angle between the length direction of the fractured surface and the
are not intrinsically able to simulate the coalescence of voids. Hence, tensile direction is approximately equal to 55 degrees when the sheet
the coalescence of voids is generally assumed to occur when the void metal is isotropic [14]. Hill [15] performed the analysis on the localized
volume fraction reaches a critical value [9]. However, this assumption necking of a biaxially stretched sheet metal, in which the sheet metal in
is shown to be inappropriate because the critical void volume fraction the region of the localized necking was assumed to neither elongate nor
depends on the stress state [10,11]. Therefore, the coalescence of voids contract in the length direction of the region of the localized necking. In
should be evaluated using a model which has a definite physical other words, plane-strain deformation in the plane perpendicular to the

E-mail address: komori@daido-it.ac.jp.

https://doi.org/10.1016/j.tafmec.2020.102636
Received 14 February 2020; Received in revised form 29 April 2020; Accepted 29 April 2020
Available online 12 May 2020
0167-8442/ © 2020 Elsevier Ltd. All rights reserved.
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

Nomenclature rb radius of bar before single pass bar drawing


rf radius of bar after single pass bar drawing
a, b major diameter of void, minor diameter of void v , v' material velocity in y'-direction, material velocity in
A0 material constant defined in Eqs. (3) and (5) x'-direction
A1 material constant defined in Eqs. (4) and (5) v1 , v 2 amounts of velocity discontinuity
B1 material constant defined in Eqs. (4) and (5) Z, z initial coordinate in axial direction, deformed coordinate
C material constant defined in Eq. (15) in axial direction
c carbon content of material in percentage die semiangle
C2 material constant defined in Eq. (18) ¯ equivalent strain rate
C3 material constant defined in Eq. (19) 0 material constant defined in Eq. (15)
di diameter of bar without prestrain before multipass bar M strain of matrix
drawing , initial coordinate in circumferential direction, deformed
do diameter of bar at cylinder upsetting after multipass bar coordinate in circumferential direction
drawing 1, 2 angles between direction of maximum principal stress and
E ratio of energy dissipation rate of internal necking to en- velocity discontinuity lines
ergy dissipation rate of homogeneous deformation ¯ equivalent stress
f void volume fraction of material 0 material constant defined in Eq. (15)
hb height of specimen before upsetting M tensile yield stress of matrix
hf height of specimen after upsetting , '
original Gurson yield function, approximated Gurson yield
k shear yield stress of matrix function
L , L' length of rectangle in x'-direction, length of rectangle in angle between coordinate axis in upsetting direction and
y'-direction line connecting two neighboring voids
2l1, 2l2 lengths of velocity discontinuity lines d range of in which ratio E for two neighboring voids after
n material constant defined in Eq. (15) forming is less than unity
p distance between two neighboring voids before forming
R (x ) ramp function

length direction of the region of the localized necking was assumed. element simulation is performed using the representative volume ele-
During the uniaxial tensile test of a sheet metal, the angle between the ment, it is easy to assume an ellipsoidal void and the plane-strain de-
length direction of the region of the localized necking and the tensile formation of a matrix. However, it is inappropriate to use the numerical
direction was analyzed to be approximately equal to 55 degrees. Hence, model for void coalescence in the simulation of ductile fracture in
it is appropriate to assume that, in the uniaxial tensile test of a sheet metal-forming processes, in which a large number of voids should be
metal, plane-strain deformation occurs in the plane perpendicular to assumed, because of the required computational time. In the analytical
the length direction of the fractured surface. During the upsetting of a models for void coalescence [22–25], in which an axisymmetric void
cylinder, Kuhn and Lee [16] experimentally demonstrated that plane- and the axisymmetric deformation of a matrix are assumed, it is im-
strain deformation appeared immediately before ductile fracture oc- possible to assume an ellipsoidal void and the plane-strain deformation
curred. of a matrix because of the assumption of the analytical models for void
A number of studies, which are introduced in several review papers coalescence.
[17–20], have been performed recently on the simulation of ductile Therefore, it is not possible to predict ductile fracture in metal-
fracture from a microscopic viewpoint. However, while void growth forming processes, in which void nucleation as well as void growth and
and void coalescence are dealt with in these studies, void nucleation is void coalescence occurs, and in which plane-strain deformation appears
scarcely dealt with. Void nucleation is the primary phenomenon in immediately before ductile fracture occurs, from a microscopic view-
ductile fracture [3], and the following points on void nucleation should point using these analytical or numerical models [21–25].
be assumed to perform the simulation of ductile fracture from a mi- Recently, the author has attempted to predict ductile fracture during
croscopic viewpoint. hole expansion test [26,27], bar drawing [28] and tensile test using
* the criteria for the void nucleation notched specimens [29,30] from a microscopic viewpoint using a model
* the void shape and void configuration at the void nucleation which has a definite physical meaning: an ellipsoidal void model. The
The simulation of ductile fracture in metal forming processes from a author’s proposed model of void coalescence is based on the two-di-
microscopic viewpoint cannot be performed if the aforementioned mensional void model proposed by Thomason [31] and the two-di-
points on void nucleation are not assumed. To the best of the author’s mensional void model proposed by Melander and Ståhlberg [32], which
knowledge, the aforementioned points on void nucleation as well as are also derived from a microscopic viewpoint and have a definite
certain points on void growth and void coalescence have not been as- physical meaning. In the Thomason model, the Melander and Ståhlberg
sumed in the research other than the research conducted by the current model, and the author’s model, plane-strain deformation is assumed to
author. occur at void coalescence, and void coalescence is assumed to occur
While a number of analytical or numerical models for void coales- when the energy dissipated in the internal necking deformation mode is
cence have been proposed from a microscopic viewpoint, an axisym- smaller than the energy dissipated in the homogeneous deformation
metric void and the axisymmetric deformation of a matrix have been mode.
assumed in most analytical or numerical models [21–25]. In other Both the Thomason model [31] and the Melander and Ståhlberg
words, an ellipsoidal void and the plane-strain deformation of a matrix model [32] assume that the void is rectangular and that the direction of
have not been assumed in the most analytical or numerical models. the major axis of the void coincides with the direction of the maximum
However, the plane-strain deformation of the material appears irrele- principal stress. However, the void is not rectangular in reality and the
vant to the prior deformation mode, immediately before ductile frac- direction of the major axis of the void in general does not coincide with
ture occurs. the direction of the maximum principal stress in metal-forming pro-
In the numerical model for void coalescence [21], in which finite- cesses. Hence, the author’s proposed model assumes that the void is

2
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

ellipsoidal and that the direction of the major axis of the void does not
coincide with the direction of the maximum principle stress. Therefore,
although both the Thomason model and the Melander and Ståhlberg
model cannot be used for simulating metal-forming processes, the au-
thor’s model can be used for simulating metal-forming processes.
In the author’s proposed model, as well as the Thomason model [31]
and the Melander and Ståhlberg model [32], the deformation gradient
of the matrix is implicitly assumed to be identical to the deformation
gradient of the void. Hence, the effect of prestrain on void coalescence
is naturally incorporated into the author’s model without any addi-
tional assumptions. This natural incorporation is one of the character- Fig. 1. Overview of entire simulation procedure at each time step.
istics of the author’s model.
This study investigates the impact of non-proportional loading on
inclusions is assumed in Komori [28] and in this study.
fracture from a microscopic viewpoint, whereas the role of non-pro-
portional loading has also been investigated from the perspectives of
formability in forming limit curves [33] as well as phenomenological 2.1. Outline
characterization of fracture [34].
Although a number of numerical studies are performed on the effect Fig. 1 shows the overview of the entire simulation procedure at each
of prestrain on ductile fracture, most of these studies [33–37] are per- time step. Macroscopic simulation and microscopic simulation are al-
formed from a macroscopic viewpoint, using variables such as stress ternately performed. In the macroscopic simulation, the deformation of
and strain. Although a few numerical studies [38,39] are performed the material is simulated by the conventional rigid-plastic finite-ele-
from a microscopic viewpoint on the effect of prestrain on void coa- ment method. The displacement gradient rate and the void volume
lescence, the application of these studies to the simulations of metal- fraction rate of the material in each finite element, which are calculated
forming processes is not straightforward, because these studies use fi- in the macroscopic simulation, are used in the subsequent microscopic
nite elements to discretize voids. simulation. The fracture of the material in each finite element is eval-
Surface cracking in cylinder upsetting is a representative ductile uated by the ellipsoidal void model in the microscopic simulation. The
fracture phenomenon in metal-forming processes [40]. A number of information whether the material fractures in each finite element is
studies have been performed on the surface cracking in cylinder up- used in the macroscopic simulation of the next time step.
setting experimentally [41,42,43,16,44,45,46] and numerically The macroscopic simulation is performed to obtain the material
[1,47–55]. deformation, whereas the microscopic simulation is performed to
Kudo and Aoi [41] performed the cylinder upsetting of an annealed evaluate the material fracture. As described in the introduction, plane-
medium carbon steel and obtained the following three types of cracks: strain deformation appears irrelevant to the prior deformation mode,
the longitudinal crack, the oblique crack, and the mixed crack, which is immediately before ductile fracture occurs. Hence, although three-di-
a mixture of the longitudinal crack and the oblique crack. Kuhn and Lee mensional deformation can be assumed in the macroscopic simulation,
[16] performed the cylinder upsetting of an as-drawn medium carbon plane-strain deformation should be assumed in the microscopic simu-
steel and an annealed medium carbon steel and observed localized lation.
strain instability prior to ductile fracture, during which plane-strain The macroscopic simulation and the microscopic simulation are
deformation appeared. Because plane-strain deformation is assumed to uncoupled, because the information transferred from the macroscopic
occur at void coalescence in the author’s proposed model, the author’s simulation scheme to the microscopic simulation scheme differs from
model is appropriate to predict surface cracking in cylinder upsetting. the information transferred from the microscopic simulation scheme to
In several studies [48,51,55], models for the surface cracking in cy- the macroscopic simulation scheme. As explained in Section 2.2, the
linder upsetting were proposed. However, in most of these numerical yield function proposed by Gurson [8] is used in the macroscopic si-
studies, simulations were performed using the finite-element method mulation, whereas the conventional Mises yield function is used in the
and various ductile fracture criteria [4,5,6,7,56] were evaluated by the microscopic simulation. Although a spherical void is assumed in the
comparisons of the experimental results and the simulation results. Gurson yield function, the assumption of an ellipsoidal void in the
In these numerical studies, the effect of prestrain on the surface microscopic simulation is justified since the macroscopic simulation
cracking in cylinder upsetting was not addressed. Although there was a and the microscopic simulation are uncoupled.
study which dealt with the effect of prestrain on the surface cracking in The material has to be rigid-plastic in the macroscopic simulation
cylinder upsetting [57], the maximum prestrain was smaller than 0.5, for consistency, because the material is assumed to be rigid-plastic in
which was not sufficiently large. the ellipsoidal void model used in the microscopic simulation. The
In this study, surface cracking in cylinder upsetting is predicted macroscopic simulation, as well as the microscopic simulation, is per-
using the author’s proposed void model. First, a specimen for the plane- formed using an in-house computer software, which is developed with
strain tensile test obtained from a bar is proposed to obtain a material reference to a specialized book [58].
constant on ductile fracture. Next, the experiment and simulation of
multipass bar drawing of five types of carbon steels are performed, and 2.2. Outline of macroscopic simulation
the effect of prestrain on the surface cracking in cylinder upsetting is
demonstrated. The deformation of the material is simulated using the conventional
rigid-plastic finite-element method [58]. Axisymmetric state is assumed
2. Simulation method during the simulation of the cylinder upsetting of a bar and during the
simulation of the uniaxial tensile test of a bar, whereas plane-strain
The simulation method used in this study, which is the same as the state is assumed during the simulation of the plane-strain tensile test of
simulation method used in the author’s previous studies [28–30] and is a bar.
explained in detail in Komori [40], is described only briefly. Voids The yield function proposed by Gurson [8] is adopted:
nucleate due to either the separation of inclusions from the matrix or
the cracking of inclusions [3]. The separation of inclusions from the 3 ij ij kk
= · + 2f cosh 1 f 2 = 0,
(1)
2
matrix is assumed in Komori [29,30], whereas the cracking of 2 M 2 M

3
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

where M is the tensile yield stress of the matrix, and f is the void Maximum
volume fraction of the material. Although the Gurson yield function shear stress
was modified by Tvergaard [13], this modification was not made the-
oretically but made practically; several parameters were introduced in
the Gurson yield fuction so that in a bifurcation simulation the simu-
lation results calculated using the modified Gurson yield function do
not differ considerably from the simulation results calculated using the
representative volume element. Moreover, the relationship between the
initial void volume fraction and the nominal fracture strain calculated
using the ellipsoidal void model incorporated into the Gurson yield
fuction agrees well with the relationship calculated using the ellipsoidal Before forming
void model incorporated into the modified Gurson yield function [59].
Pearlite nodule
Hence, the original Gurson yield function is used.
Because the yield function is not a function of the second power of Ferrite matrix After forming
stress, the rigid-plastic simulation is not easily performed using Eq. (1). Fig. 2. Void configuration and void shape.
Hence, coshx is approximated to be 1 + x 2 /2 [60]. Therefore, the ap-
proximated yield function ' used in this study is
the void configuration and the void shape are calculated. Third, as
2 explained in Section 2.5, the ratio of the energy dissipation rate of in-
' 3 ij ij f kk
= · 2
+ · 2
(1 f ) 2 = 0. ternal necking to the energy dissipation rate of homogeneous de-
2 M 4 M (2) formation E is calculated, and whether the material fractures is de-
In metal-forming processes, the maximum stress triaxiality in the termined.
material without a crack is low, whereas the maximum stress triaxiality Whether the material fractures is determined using the void volume
in the material with a crack is high. Here, the simulation of crack in- fraction f and the deformation gradient x / X , which are non-dimen-
itiation in the material without a crack is performed, and the simulation sional values. Hence, the void size is not specified. In other words, the
of crack propagation in the material with a crack is not performed. size of a void is assumed to be much smaller than the size of a finite
Hence, the accuracy of the approximation of hyperbolic cosine by the element such that the voids are distributed uniformly throughout the
Maclaurin expansion is high due to the low maximum stress triaxiality. finite element.
The following three types of evolution equations, which denote the
change in the void volume fraction, are assumed: 2.4. Void configuration and void shape

f = (1 f) kk + A0 ¯, (3) Fig. 2 shows the void configuration and the void shape. The black
part denotes the pearlite nodule, the gray part denotes the ferrite ma-
kk
f = (1 f) kk + A1 R B1 ¯, trix, and the white part denotes the void. Pearlite nodules are assumed
3¯ (4)
to be uniformly distributed in the material. Before forming, the pearlite
and nodule configuration is assumed to be the centroid of a hexagonal grid,
and the pearlite nodule shape is assumed to be a circle. When the area
f = (1 f) kk + A0 ¯ + A1 R kk
B1 ¯, of a hexagonal grid is assumed to be unity, the area of a pearlite nodule
3¯ (5)
is equal to (c 0.02)/0.75 [62], where c denotes the carbon content of
where ¯ is the equivalent stress, ¯ is the equivalent strain rate, and A0 , the material in percentage, whereas the area of a void is equal to the
A1, and B1 are the material constants. The function R (x ) in Eqs. (4) and void volume fraction f .
(5) denotes the ramp function, that is, R (x ) = x when x is positive and For carbon steels, the direction of the major axis of the crack at
R (x ) = 0 when x is negative. The first terms on the right-hand sides of crack initiation in the pearlite nodule is known to coincide with the
Eqs. (3), (4), and (5) denote void growth, whereas the second and third direction of the maximum shear stress of the material [63]. Although a
terms on the right-hand sides of Eqs. (3), (4), and (5) denote void nu- crack appears in the pearlite nodule, an ellipsoidal void is assumed in
cleation, which is caused by the cracking of perlites. the ellipsoidal void model. Hence, the following assumptions are made
Eq. (5) is derived by combining Eqs. (3) and (4). When the material on the void shape before forming.
is specified, any one of Eqs. (3), (4), and (5) can be used. The choice of *The void volume fraction before forming is specified to be the in-
either one of Eqs. (3), (4), or (5) is discussed in Section 4.1. itial void volume fraction.
The physical meanings of the evolution equations of Eqs. (3), (4), *The void shape before forming is an ellipsoid whose flattening is
and (5) are described in detail in the author’s previous study [30]. equal to 0.9.
The evolution equations of Eqs. (3), (4), and (5) are assumed using *The direction of the major axis of the void before forming is
macroscopic stress and strain. It is desirable that the evolution equa- identical to the direction of the maximum shear stress of the material.
tions are assumed using microscopic stress and strain such as the stress A hexagonal grid embedded in the material before forming becomes
and strain inside the inclusion and the stress and strain on the interface a deformed hexagonal grid after forming. The deformation gradient
between the matrix and the inclusion. However, the evolution equa- ( x / X ) material before forming is defined as the deformation gradient of the
tions are assumed using macroscopic stress and strain, because it is not deformed hexagonal grid, and is calculated in the macroscopic simu-
easy to predict microscopic stress and strain [61] in the macroscopic lation. The deformation gradients of the pearlite nodule configuration,
simulation. the pearlite nodule shape, the void configuration, and the void shape,
are used in the microscopic simulation, are assumed in the following
2.3. Outline of microscopic simulation relationships.
x x
In each time step, the following microscopic simulation is per- =
X pearlite nodule configuration X pearlite nodule shape
formed in each finite element. First, the void volume fraction f and the
x
deformation gradient x / X are calculated in the macroscopic rigid- =
plastic finite-element simulation. Second, as explained in Section 2.4,
X material before forming (6)

4
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

x x x l1sin 2 + l2sin 1
= = k· v1·2l2 + k · v2·2l1 = 4kv .
X void configuration X void shape X material before forming (7) sin( 1 + 2 ) (10)

With reference to the Thomason model [31] and the Melander and
The void shape is basically determined by Eq. (7). However, because
Ståhlberg model [32], void coalescence is assumed to occur when the
a crack appears in the pearlite nodule in reality, the void shape is
energy dissipation rate of internal necking is less than the energy dis-
modified; the major diameter of the void and the minor diameter of the
sipation rate of homogeneous deformation. Hence, using Eqs. (9) and
void are modified in order that the flattening of the ellipsoidal void is
(10), the criterion of void coalescence is expressed by
equal to 0.9, under the condition that the void volume fraction does not
change due to the modification. Hence, the major diameter of void a l1sin 2 + l2sin 1
(1 f )L .
and the minor diameter of void b are calculated from sin( 1 + 2 ) (11)

10f f The ratio of the energy dissipation rate of internal necking to the
a= , and b =
(8) energy dissipation rate of homogeneous deformation E is defined as
10
l1sin 2 + l2 sin 1
For carbon steels, a void remains in the pearlite nodule and does not E= .
(1 f ) L ·sin( 1 + 2 ) (12)
penetrate into the ferrite matrix until the internal necking between two
neighboring voids appears [64]. However, in the microscopic simula- The criterion of void coalescence is satisfied when E is less than
tion, because insufficient restrictions are made on the void shape, the unity.
void may break the inclusion into two parts and penetrate into the When l2 is equal to l1 and 2 is equal to 1, the following inequality,
matrix. Hence, the following restriction is made on the void shape. which is the criterion of void coalescence, is obtained from Eq. (11).
Fig. 3 shows the void shape in the hexagonal grid. When the void l1
penetrates into the matrix, the void shape is modified in order that the (1 f )L
cos 1 (13)
void does not penetrate into the matrix, under the condition that the
void volume fraction does not change due to the modification. Hence, Fig. 5 shows the schematic representation of the criterion of void
the flattening of the ellipsoidal void becomes smaller than 0.9 after the coalescence. Since the initial shape of an ellipsoidal void is a circle, a
modification. After the inclusion is broken into two parts, a void is not circular void is shown for convenience. Eq. (13) indicates that the cri-
assumed to nucleate because the inclusion does not fracture further- terion of void coalescence is satisfied when the length of (1 f ) L is
more, in other words, the material constants A0 and A1 in Eqs. (3), (4), larger than the length of l1/cos 1 in Fig. 5.
and (5) are assumed to be zero. Although the criterion of void coalescence is microscopically sa-
tisfied for a specified angle between the coordinate axis in the upsetting
direction and the line connecting two neighboring voids, the material
2.5. Determination of material fracture does not necessarily fracture macroscopically. The criterion of micro-
scopic void coalescence is related to the macroscopic material fracture
Fig. 4 shows the velocity field. First, the material is assumed to in the followings.
deform homogeneously and the void coalescence is not assumed to Fig. 6 shows the two neighboring voids and the void coalescence
occur. Fig. 4(a) shows the velocity field for homogeneous deformation. region. The non-dimensional distance between two neighboring voids
The y'-direction is made to coincide with the direction of the maximum before forming is p , which is equal to 4 4/3 [59]. Variable is defined
principal stress, and the x'-direction is made to coincide with the di- as the angle between the coordinate axis in the upsetting direction and
rection of the minimum principal stress. the line connecting two neighboring voids. is an arbitrary value be-
The rectangle indicated by the dotted lines is considered. The length tween 0 and 2 , and the probability density function of becomes
of the rectangle in the x'-direction is L , which is assumed to be the constant regardless of the value of , since the void configuration is
distance between two neighboring voids, and the length of the rec- assumed to be macroscopically isotropic before forming. When is
tangle in the y'-direction is L'. The material velocity in the y'-direction is specified, the ratio E can be calculated for two neighboring voids after
v , and the material velocity in the x'-direction is v', which is equal to forming. d is defined as the range of in which the ratio E for two
(L / L') v because of the incompressibility condition, since no void neighboring voids after forming is less than unity. The void coalescence
growth is assumed at each time step in the microscopic simulation. region is defined as a sector whose central angle is d .
Hence, considering the conventional Mises yield criteria and the Six voids surround the central void, as shown in Fig. 2. When d is
void volume fraction f , the energy dissipation rate in the material due
to homogeneous deformation is expressed by Pearlite nodule
(1 f )(2vL max + 2v'L' min ) = (1 f )4kvL (9) Ferrite matrix
where k denotes the shear yield stress of the matrix.
Next, the material between two neighboring voids is assumed to
neck and the void coalescence is assumed to occur. Fig. 4(b) shows the
velocity field for internal necking. The velocity discontinuity line,
which is indicated by dotted lines, is not necessarily assumed to coin-
cide with the tangent of the two neighboring voids, because the location
of the velocity discontinuity line is optimized so that the energy dis-
sipation rate on the velocity discontinuity line due to internal necking
defined below is minimized. In Fig. 4(b), 1 and 2 denote the angles
between the direction of the maximum principal stress and the velocity
discontinuity lines, whereas 2l1 and 2l2 denote the lengths of the velocity
discontinuity lines. The velocity field is shown as a hodograph. More- Before After
over, v1 and v2 denote the amounts of velocity discontinuity. modification modification
Using the law of sines, the energy dissipation rate on the velocity
discontinuity line due to internal necking is expressed by Fig. 3. Void shape in hexagonal grid.

5
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

Fig. 4. Velocity field.

13 mm, were used because of these steel bars’ availability and wide
usage: JIS SS400, JIS S15C, JIS S55C, JIS SCM415, and JIS SCM435. JIS
SS400 is a rolled steel for general structures. JIS S15C and JIS S55C are
carbon steels for machine structural use and are equivalent to ISO
C15E4, and ISO C55E4, respectively. JIS SCM415 and JIS SCM435 are
l1
low-alloyed steels for machine structural use and are equivalent to ISO
cosθ1 18CrMo4, and ISO 34CrMo4, respectively. Table 1 shows the chemical
composition of steel bars.
θ1 θ1 We used as-rolled bars of SS400 and S55C which were not annealed,
because these as-rolled bars are assumed to have no prestrain. We used
cold drawn bars of S15C, SCM415, and SCM435 which were annealed
l1 L l1 at 1153 K (880 °C), 1153 K (880 °C), and 1103 K (830 °C), respec-
tively, for an hour and then cooled in a vacuum chamber to remove
prestrain, because these cold drawn bars are assumed to have large
Fig. 5. Schematic representation of criterion of void coalescence.
prestrain.
Fig. 7 shows the stress–strain relationships of the steel bars. The
Void coalescence region steel bars had yield point elongations. A uniaxial tensile test, in which
the specimen was a bar, was performed, and the stress–strain re-
lationship of the bar was obtained and subsequently approximated
d using the following equation:
p
M = max( 0, C ·( M 0)
n),
(15)

where M is the strain of the matrix, and 0 , C , 0 , and n are the material
constants. Five specimens were used for each steel bar in the uniaxial
Before forming After forming tensile test, and the averaged experimental results were demonstrated.
Fig. 6. Two neighboring voids and void coalescence region. Table 2 shows the material constants in the stress–strain relationship.
These material constants were used in the simulation in Section 4. The
stress–strain relationships of the steel bars calculated using Eq. (15) are
/6(=(2 /6)/2) , the probability that the ratio E is less than unity for a
also shown in Fig. 7, and agree with the stress–strain relationships of
specified is half. Hence, the macroscopic probability of void coales-
the steel bars obtained experimentally. To obtain the material constants
cence is half, which is assumed to be a reasonable criterion for mac-
in the evolution equations, the diameter of the specimen at rupture in
roscopic material fracture. Therefore, when the following equation is
the uniaxial tensile test was measured. The initial diameter of the
satisfied, the material is assumed to fracture macroscopically:
Table 1
d = .
6 (14) Chemical composition of steel bars (unit: mass%).
C Si Mn P S Cr Mo Cu Ni

3. Experimental results SS400 0.14 0.20 0.50 0.014 0.023 0.10


S15C 0.16 0.19 0.40 0.018 0.021 0.05 0.01 0.02
S55C 0.54 0.18 0.70 0.011 0.023 0.09 0.10 0.05
3.1. Tensile test experiments
SCM415 0.16 0.21 0.69 0.017 0.009 0.95 0.15 0.16 0.09
SCM435 0.36 0.24 0.78 0.016 0.009 1.10 0.16 0.18 0.16
The following five types of steel bars, whose initial diameters were

6
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

SS400 (Exp.) SS400 (Sim.) observed and the size of the void was much smaller than the size of the
S15C (Exp.) S15C (Sim.) specimen, were shown in the author’s previous study [30], only the
S55C (Exp.) S55C (Sim.)
SCM415 (Exp.) SCM415 (Sim.) micrographs of SCM415 and SCM435 were shown in this study. The
SCM435 (Exp.) SCM435 (Sim.) micrographs of SCM415 and SCM435 indicated that dimples that were
1000 yielded from voids were observed and that the size of the void was
much smaller than the size of the specimen. Hence, the simulation
Actual stress (MPa)

800 method described in Section 2 was confirmed to be applicable to all the


materials.
600

400 3.2. Experiments of drawing and upsetting

200
Prior to cylinder upsetting, multipass drawing of a bar was per-
0 formed to introduce prestrain in the bar. The prestrain is defined as
0.0 0.1 0.2 0.3 2ln(di /d o) , where di denotes the diameter of the bar without prestrain
Logalithmic strain before multipass bar drawing, and do denotes the diameter of the bar at
Fig. 7. Stress–strain relationships of steel bars.
cylinder upsetting after multipass bar drawing.
Fig. 11 shows the coordinates and the notations used in single pass
bar drawing. rb denotes the radius of the bar before single pass bar
Table 2 drawing, rf denotes the radius of the bar after single pass bar drawing,
Material constants in stress–strain relationship.
and denotes the die semiangle. The diameter of the bar without
0 (GPa) C (GPa) 0 n prestrain before multipass bar drawing was equal to 13 mm. For SS400,
S15C, S55C, and SCM415, the reduction in area in each die (rb2 r f2)/ rb2
SS400 0.35 0.78 0.02 0.18 was equal to 15%, the die angle 2 was equal to 15 degrees, and the
S15C 0.27 0.67 0.015 0.21
S55C 0.54 1.14 0.02 0.10
cylinder upsetting was performed after each two-pass bar drawing until
SCM415 0.34 0.81 0.015 0.19 the diameter of the bar became smaller than 5 mm. For SCM435, the
SCM435 0.40 0.97 0.02 0.14 reduction in area in each die (rb2 r f2)/ rb2 was equal to 20%, the die
angle 2 was equal to 15 degrees, and the cylinder upsetting was per-
formed after each pass bar drawing until the diameter of the bar be-
specimen was equal to 13 mm. came smaller than the critical diameter at which the inner fracture
Plane-strain tensile test, as well as the uniaxial tensile test, was defect appeared in the bar. Conventional machine oil was used between
performed using a bar, because multipass drawing and cylinder upset- the die and the bar for lubrication.
ting were performed using a bar. Since there were no specimens for the In cylinder upsetting, a pair of flat dies were used and no lubricants
plane-strain tensile test obtained from a bar, the following specimen for were used between the die and the specimen. Five kinds of specimens,
the plane-strain tensile test, which was obtained from a bar, was pro- whose initial height/diameter ratios were 1.0, 1.25, 1.5, 2.0, and 2.5,
posed with reference to the specimen for the plane-strain tensile test were used. First, square grids, in which the distance between two
obtained from a plate [65]. neighboring lines was equal to 147 µm, were embedded on the side
Fig. 8 shows the specimen for the plane-strain tensile test. Because surface of the specimen. Herein, two video cameras connected to two
the bar parts of the specimen whose initial diameters were equal to monitors were used to observe the side surface of the specimen. Next,
13 mm scarcely deform in the plane-strain tensile test, plane-strain the upsetting of the specimen was performed until the side surface of
deformation is obtained approximately at the center in the width di- the specimen fractured. The coordinates of the vertices of a deformed
rection of the plate part of the specimen whose initial width, initial square grid that was next to a fractured square grid were then measured
length, and initial thickness were equal to 13 mm, 5 mm, and 2 mm, using an optical microscope, and the deformation gradient of the de-
respectively. Since, according to the observation using a video camera, formed square grid was subsequently calculated.
fracture initiation occurred at the center in the width direction of the Given that sufficient preliminary experiments were performed in
plate part of the specimen, the thickness of the plate part of the spe- bar drawing and cylinder upsetting, one specimen was used for each
cimen at rupture in the plane-strain tensile test was measured at the experimental condition in cylinder upsetting. However, for each ex-
center in the width direction. Five specimens were used for each steel perimental condition, the coordinates of the vertices of a total of five
bar in the plane-strain tensile test, and the averaged experimental re- deformed square grids that were next to a total of five fractured square
sults were demonstrated. Table 3 shows the diameter of the specimen at grids were measured using an optical microscope, and the averaged
rupture in the uniaxial tensile test and the thickness of the specimen at experimental results were demonstrated.
rupture in the plane-strain tensile test. Fig. 12 shows the side surfaces of the specimens fractured for S55C.
Fig. 9 shows the micrographs on the longitudinal sections of the The figures in the caption indicate the diameters of the bars at cylinder
specimens without prestrain. Because the micrographs of SS400, S15C, upsetting after multipass bar drawing, which are related to prestrain.
and S55C, in which ferrite-pearlite microstructure was observed and Longitudinal cracks were observed in these specimens.
pearlite was uniformly distributed, were shown in the author’s previous Since the non-diagonal component of the deformation gradient of
study [30], only the micrographs of SCM415 and SCM435 were shown the square grid next to the fractured square grid was negligibly small,
in this study. The horizontal direction coincides with the axial direction the non-diagonal component of the deformation gradient of the square
of the specimen. The specimens of SCM415 and SCM435 were etched grid was assumed to be zero. Because the deformation gradient at
using nital. As expected [66], ferrite-pearlite microstructure was ob-
served. Although ferrite-pearlite banding [67] was slightly observed in 5
the micrographs of SCM415 and SCM435, pearlite was rather uniformly
distributed. Hence, all the materials were used to evaluate the validity 13
of the simulation method described in Section 2.
Fig. 10 shows the micrographs of the fractured surfaces of the 2
specimens without prestrain. Because the micrographs of SS400, S15C,
and S55C, in which dimples that were yielded from voids were Fig. 8. Specimen for plane-strain tensile test.

7
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

Table 3 thickness direction in the plane-strain tensile test were assumed to be


Diameter of specimen at rupture in uniaxial tensile test and thickness of spe- equal to ten.
cimen at rupture in plane-strain tensile test. A cup and cone fracture was observed in the experiment of the
Diameter (mm) Thickness (mm) uniaxial tensile test, whereas a flat and slant fracture [68] was observed
in the experiment of the plane-strain tensile test. Hence, line symmetry
SS400 6.8 1.0 with respect to the central cross section was observed in neither the
S15C 7.4 1.1
experiment of the uniaxial tensile test nor the experiment of the plane-
S55C 9.6 1.5
SCM415 7.6 1.1 strain tensile test. However, it is not easy to perform simulations of the
SCM435 9.6 1.5 cup and cone fracture and the flat and slant fracture when the direction
of the crack propagation is not determined beforehand. Moreover, the
cup and cone fracture and the flat and slant fracture exhibited some
fracture initiation was equal to the product of the deformation gradient degree of line symmetry. Thus, for simplicity, line symmetry with re-
of the square grid and the deformation gradient due to prestrain, the spect to the central cross section was assumed in both the simulation of
axial strain at fracture initiation, ln(( z / Z ) fracture) , and the circumfer- the uniaxial tensile test and the simulation of the plane-strain tensile
ential strain at fracture initiation, ln(( / ) fracture) , were calculated as test.
follows. The simulation was performed using the node separation method
until the specimen ruptures [69]. When one or more finite elements
z z z
ln = ln + ln fractures in the nth time step, the displacement of the tool was modified
Z fracture Z squaregrid Z prestrain (16) such that only one finite element fractures in the nth time step. Then,
the restriction of the displacement of the node in the axial direction was
released such that the finite element fractures. The initial notched
ln = ln + ln
fracture squaregrid prestrain (17) depth of the specimen was assumed to be 10 µm so that necking occurs
on the axis of line symmetry with respect to the central cross section.
Fig. 13 shows the relationship between the axial strain at fracture The initial void volume fraction was assumed to be 0.001 with re-
initiation and the circumferential strain at fracture initiation for all the ference to the initial void volume fraction of aluminum-killed steel in
materials. The experimental results are plotted using solid symbols or Schmitt and Jalinier [70].
bold symbols. The figures in the legend indicate the diameters of the The simulation was performed using any one of Eqs. (3), (4), and
bars at cylinder upsetting after multipass bar drawing, which are re- (5). For simplicity, B1 = 0 was assumed in Eqs. (4) and (5), whereas
lated to prestrain. Because the effect of the initial height/diameter ratio A0 = A1 was assumed in Eq. (5). In other words, in this study, Eqs. (4),
of the specimen on the axial strain at fracture initiation and the cir- and (5) were replaced with Eqs. (4)′, and (5)′, respectively:
cumferential strain at fracture initiation was not large, the initial
height/diameter ratio of the specimen was not shown in the legend. For
f = (1 f ) kk + A1 R 3kk¯ ¯,( ) (4)′
and
all the materials, with decreasing the diameter of the bar at cylinder
upsetting after multipass bar drawing, that is, with increasing the
f = (1 f) kk (
+ A0 1 + R ( ) ) ¯.
kk

(5)′
prestrain, the axial strain at fracture initiation monotonically increases, Hence, only one material constant in an evolution equation, that is,
whereas the circumferential strain at fracture initiation monotonically A0 in Eq. (3), A1 in Eq. (4)′, and A0 in Eq. (5)′, is required to be iden-
decreases. tified. The material constants A0 and A1 were determined such that the
diameter of the specimen at rupture in the uniaxial tensile test or the
thickness of the specimen at rupture in the plane-strain tensile test
4. Simulation results
calculated from the simulation agrees with the diameter of the spe-
cimen at rupture or the thickness of the specimen at rupture obtained in
4.1. Simulations of tensile test
the experiment, respectively.
Table 4 shows the material constants in the evolution equations
The simulation of the uniaxial tensile test of a bar and the simula-
obtained in the simulation of the uniaxial tensile test of a bar or in the
tion of the plane-strain tensile test using a bar were performed using the
simulation of the plane-strain tensile test using a bar. When the material
conventional rigid-plastic finite-element method [58]. In the simulation
is specified, the material constant in the plane-strain tensile test is
of the uniaxial tensile test, the linear rectangular element was used and
larger than the material constant in the uniaxial tensile test. It is a well-
axisymmetric deformation was assumed, whereas in the simulation of
known experimental fact [2] that the strain to fracture decreases with
the plane-strain tensile test, the linear rectangular element was used
increasing the mean normal stress and that the strain to fracture de-
and plane-strain deformation was assumed. The number of finite ele-
creases with decreasing the magnitude of the Lode parameter. The
ments both in the radial direction in the uniaxial tensile test and in the

100μm 100μm

(a) SCM415 (b) SCM435


Fig. 9. Micrographs on longitudinal sections of specimens without prestrain.

8
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

(a) SCM415 (b) SCM435


Fig. 10. Micrographs of fractured surfaces of specimens without prestrain.

r φ13.0mm (Exp.)
φ11.1mm (Exp.)
φ13.0mm (Simu.)
φ11.1mm (Simu.)
φ9.5mm (Exp.) φ9.5mm (Simu.)
φ8.1mm (Exp.) φ8.1mm (Simu.)
φ6.9mm (Exp.) φ6.9mm (Simu.)
φ5.9mm (Exp.) φ5.9mm (Simu.)
φ5.0mm (Exp.) φ5.0mm (Simu.)
rb rf axial strain
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0
1.0
z
0

circumferential strain
0.5
Fig. 11. Coordinates and notations used in single pass bar drawing. 0.0

-0.5

-1.0

-1.5
(a) SS400

φ13.0mm (Exp.) φ13.0mm (Simu.) φ13.0mm (Exp.) φ13.0mm (Simu.)


φ11.1mm (Exp.) φ11.1mm (Simu.) φ11.1mm (Exp.) φ11.1mm (Simu.)
φ9.5mm (Exp.) φ9.5mm (Simu.) φ9.5mm (Exp.) φ9.5mm (Simu.)
φ8.1mm (Exp.) φ8.1mm (Simu.) φ8.1mm (Exp.) φ8.1mm (Simu.)
φ6.9mm (Exp.) φ6.9mm (Simu.) φ6.9mm (Exp.) φ6.9mm (Simu.)
φ5.9mm (Exp.) φ5.9mm (Simu.) φ5.9mm (Exp.) φ5.9mm (Simu.)
φ5.0mm (Exp.) φ5.0mm (Simu.) φ5.0mm (Exp.) φ5.0mm (Simu.)
axial strain axial strain
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
1.0 1.0
circumferential strain
circumferential strain

0.5 0.5

0.0 0.0

(a) φ=13.0mm, initial height/diameter ratio=1 -0.5 -0.5

-1.0 -1.0
(b) S15C (c) S55C

(b) φ=8.1mm, initial height/diameter ratio=1


Fig. 12. Side surfaces of specimens fractured for S55C.

Fig. 13. Relationship between axial strain at fracture initiation and cir-
simulation result that the material constant in the plane-strain tensile cumferential strain at fracture initiation for all materials.
test is larger than the material constant in the uniaxial tensile test is
inferred from the well-known experimental fact.

9
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

Table 4 In practical terms, the material constant A0 or A1 should be obtained


Material constants in evolution equations. in the simulation of the uniaxial tensile test or the plane-strain tensile
A0 A 0 (=A1)
test and the obtained material constant should then be used in the si-
mulations of multipass bar drawing and cylinder upsetting. However, in
Uniaxial Plane- Upsetting Uniaxial Plane- Upsetting this study, the material constant A0 or A1 was obtained in the simula-
strain strain tion of cylinder upsetting and the obtained material constant was
SS400 0.016 0.046 0.13 – – –
subsequently used in the simulations of multipass bar drawing and
S15C 0.022 0.066 0.13 – – – cylinder upsetting. The method for obtaining the material constant A0
S55C – – – 0.039 0.052 0.06 or A1 should therefore be improved in the future research.
SCM415 0.024 0.063 0.10 – – – Table 4 shows the material constants in the evolution equations
SCM435 – – – 0.040 0.053 0.045
obtained in the simulation of cylinder upsetting using the specimen
without prestrain. The material constants obtained in the simulation of
cylinder upsetting substantially agreed with the material constants
In the author’s previous study [30], by the comparison of the si-
obtained in the simulation of plane-strain tensile test for S55C and
mulation results and the experimental results, Eq. (3) was shown to be
SCM435. However, the material constants obtained in the simulation of
more appropriate than Eq. (4) in low carbon steels, whereas Eq. (4) was
cylinder upsetting were approximately twice the material constants
shown to be more appropriate than Eq. (3) in medium carbon steels.
obtained in the simulation of plane-strain tensile test for SS400, S15C,
Hence, Eq. (3) was used for SS400, S15C, and SCM415, whereas Eq. (4)’
and SCM415.
was used for S55C and SCM435. However, Eq. (4)’ was shown to be
The material constants obtained in the simulation of cylinder up-
inappropriate for S55C, because the void volume fraction on the axis of
setting did not considerably differ from the material constants obtained
the bar became larger than 0.5 after multipass bar drawing which was
in the simulation of uniaxial tensile test for S55C and SCM435.
performed in Section 4.2. Hence, Eq. (5)’, which is the combination of
However, the material constants obtained in the simulation of cylinder
Eqs. (3) and (4)’, was used not only for S55C but also for SCM435,
upsetting were approximately six times the material constants obtained
because the carbon content of SCM435 is smaller than the carbon
in the simulation of uniaxial tensile test for SS400, S15C, and SCM415.
content of S55C.
Hence, for all the materials, the differences between the material
constants obtained in the simulation of cylinder upsetting and the
4.2. Simulations of drawing and upsetting material constants obtained in the simulation of plane-strain tensile test
were much smaller than the differences between the material constants
The simulation of multipass bar drawing and the simulation of cy- obtained in the simulation of cylinder upsetting and the material con-
linder upsetting were performed using the conventional rigid-plastic stants obtained in the simulation of uniaxial tensile test. Therefore,
finite-element method [58]. In the simulations of both multipass bar obtaining material constants by the simulation of plane-strain tensile
drawing and cylinder upsetting, the linear rectangular element was test was proved to be more appropriate than obtaining material con-
used and axisymmetric deformation was assumed. Line symmetry with stants by the simulation of uniaxial tensile test, to use the material
respect to the central cross section was assumed in the simulation of constants in the simulation of cylinder upsetting.
cylinder upsetting. Because conventional machine oil was used between Fig. 14 shows the initial finite-element meshes and finite-element
the die and the bar for lubrication, the friction between the die and the meshes at fracture initiation for SS400 without prestrain. Simulation
bar was assumed to obey the Coulomb’s law, and the coefficient of results hardly differed from the simulation results calculated using the
friction in multipass bar drawing was assumed to be 0.1. Since no lu- finite elements the number of which was larger than the number of
bricants were used between the die and the specimen, the friction be-
tween the specimen and the die was assumed to obey the Coulomb’s
law, and the coefficient of friction in cylinder upsetting was assumed to
be 0.3 with reference to Kobayashi and Lee [47].
The number of finite elements in the radial direction in cylinder
upsetting was made to coincide with the number of finite elements in
the radial direction in multipass bar drawing to easily transfer in- Fracture
formation such as the void volume fraction from the simulation scheme initiation
in multipass bar drawing to the simulation scheme in cylinder upset- (a) Initial height/diameter ratio=1
ting.
When the simulation of cylinder upsetting using the specimen
without prestrain was performed using the material constant A0 or A1
obtained in the simulation of the uniaxial tensile test or the plane-strain
tensile test, the circumferential and axial strains at fracture initiation
calculated from the simulation did not necessarily agree with the cir-
cumferential and axial strains at fracture initiation obtained experi-
mentally. When the simulations of multipass bar drawing and cylinder
upsetting are performed using the material constant A0 or A1 obtained
in the simulation of the uniaxial tensile test or the plane-strain tensile
test, it is not clear whether the effect of the prestrain on the cir-
cumferential and axial strains at fracture initiation would be evaluated
correctly. Hence, the material constant A0 or A1 was determined such
that the circumferential and axial strains at fracture initiation for the Fracture
specimen without prestrain calculated from the simulation agree with initiation
the circumferential and axial strains at fracture initiation for the spe- (b) Initial height/diameter ratio=2
cimen without prestrain obtained experimentally, to accurately eval-
uate the effect of the prestrain on the circumferential and axial strains Fig. 14. Initial finite-element meshes and finite-element meshes at fracture
at fracture initiation. initiation for SS400 without prestrain.

10
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

finite elements in Fig. 14. Hence, the number of finite elements in evolution equation is required to be identified in Section 4.1, the
Fig. 14 was considered to be sufficient. The diameter was largest at the number of material constants on ductile fracture is only one in Section
center in the height direction regardless of the initial height/diameter 4.2. Therefore, ductile fracture criteria in which the number of material
ratio of the specimen because of the friction between the specimen and constants on ductile fracture is only one should be used for comparison
the die. Fracture initiation occurred at the surface of the center in the so that the number of material constants on ductile fracture coincides
height direction, at which the mean normal stress was largest, regard- with the number of material constants on ductile fracture in Section 4.2.
less of the initial height/diameter ratio of the specimen. Hence, the ductile fracture criterion proposed by Cockcroft and Latham
Fig. 13 shows the relationship between the axial strain at fracture [5] and the ductile fracture criterion proposed by Brozzo et al. [6],
initiation and the circumferential strain at fracture initiation for all the which were used in the author’s previous studies [29,30], were also
materials. The simulation results are plotted using open symbols or fine used in this study.
symbols. The figures in the legend indicate the diameters of the bars at The non-dimensional Cockcroft and Latham fracture criterion [5]
cylinder upsetting after multipass bar drawing, which are related to can be expressed by the following equation
prestrain. Although the axial strain at fracture initiation decreases with max
increasing the initial height/diameter ratio of the specimen, the initial ¯dt C2
¯ (18)
height/diameter ratio of the specimen is not shown in the legend to
reduce the complexity of the legend. For all the materials, with de- where C2 is the material constant. The Brozzo et al. fracture criterion
creasing the diameter of the bar at cylinder upsetting after multipass [6] can be expressed by the following equation
bar drawing, that is, with increasing the prestrain, the axial strain at 2 max
¯dt C3
fracture initiation monotonically increases, whereas the circumferential 3( max kk /3) (19)
strain at fracture initiation monotonically decreases. The circumfer-
ential and axial strains at fracture initiation calculated from the simu- where C3 is the material constant.
lation are substantially coincident with the circumferential and axial Table 5 shows the material constants used in the conventional
strains at fracture initiation obtained in the experiment. ductile fracture criteria. The material constants C2 and C3 were de-
Fig. 15 shows the relationship between the axial strain and the termined such that the circumferential and axial strains at fracture in-
circumferential strain calculated from the simulation from the onset of itiation for the specimen without prestrain calculated from the simu-
deformation to fracture initiation for SCM415. The initial height/dia- lation agree with the circumferential and axial strains at fracture
meter ratio of the specimen is equal to one. The figures in the legend initiation for the specimen without prestrain obtained in the experi-
indicate the diameters of the bars at cylinder upsetting after multipass ment.
bar drawing, which are related to prestrain. The axial strain increases Since it is not necessarily required to compare the simulation results
and the circumferential strain decreases in multipass bar drawing, calculated using the conventional ductile fracture criteria with the ex-
whereas the axial strain decreases and the circumferential strain in- perimental results obtained in the experiment for all the materials, the
creases in cylinder upsetting. comparison was made for the following two materials: SS400 and S55C.
Fig. 16 shows the relationship between the axial strain at fracture SS400 is a low carbon steel, in which the evolution equation of Eq. (3) is
initiation and the circumferential strain at fracture initiation in various used, whereas S55C is a medium carbon steel, in which the evolution
evolution equations for SCM435. The simulation results are plotted equation of Eq. (5)’ is used.
using open symbols or fine symbols, whereas the experimental results Figs. 17 and 18 show the relationships between the axial strain at
are plotted using solid symbols or bold symbols. The figures in the le- fracture initiation and the circumferential strain at fracture initiation
gend indicate the diameters of the bars at cylinder upsetting after for the two materials calculated using the Cockcroft and Latham frac-
multipass bar drawing, which are related to prestrain. A0 was assumed ture criterion [5], and calculated using the Brozzo et al. fracture cri-
to be 0.045 when Eq. (3) was used , whereas A1 was assumed to be 0.8 terion [6], respectively. The simulation results are plotted using open
when Eq. (4)’ was used. With reference to Fig. 13(e) and 16, when both symbols or fine symbols, whereas the experimental results are plotted
the prestrain and the initial height/diameter ratio of the specimen were using solid symbols or bold symbols. The figures in the legend indicate
specified, with increasing A1, both the circumferential strain at fracture the diameters of the bars at cylinder upsetting after multipass bar
initiation and the axial strain at fracture initiation slightly increase for drawing, which are related to prestrain. For both the ductile fracture
large prestrain. However, the circumferential and axial strains at frac- criteria and for both the materials, with decreasing the diameter of the
ture initiation calculated from the simulation are almost coincident bar at cylinder upsetting after multipass bar drawing, that is, with in-
with the circumferential and axial strains at fracture initiation obtained creasing the prestrain, the axial strain at fracture initiation mono-
in the experiment, that is, the circumferential and axial strains at tonically increases, whereas the circumferential strain at fracture in-
fracture initiation calculated from the simulation scarcely depend on itiation monotonically decreases. However, for both the ductile fracture
the type of the evolution equation used.
φ13.0mm (Simu.) φ11.1mm (Simu.)
φ9.5mm (Simu.) φ8.1mm (Simu.)
4.3. Simulations using conventional ductile fracture criteria φ6.9mm (Simu.) φ5.9mm (Simu.)
φ5.0mm (Simu.)
axial strain
The simulations were performed using conventional ductile fracture -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
criteria. The conventional Mises yield function is used in this section in 1.0
place of the approximated yield function ' . The conventional rigid-
circumferential strain

plastic finite-element method [58] was also used, in which the in- 0.5
compressibility condition was satisfied by the penalty method.
It is generally recognized that the simulation result on ductile 0.0
fracture approaches the experimental result on ductile fracture with
increasing the number of material constants on ductile fracture. Hence, -0.5
it is meaningless to compare the simulation result on ductile fracture
using the simulation method in which the number of material constants -1.0
on ductile fracture is one with the simulation result on ductile fracture
using the simulation method in which the number of material constants Fig. 15. Relationship between axial strain and circumferential strain calculated
on ductile fracture is two. Because only one material constant in an from simulation from onset of deformation to fracture initiation for SCM415.

11
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

φ13.0mm (Exp.) φ13.0mm (Simu.) φ13.0mm (Exp.) φ13.0mm (Simu.)


φ11.7mm (Exp.) φ11.7mm (Simu.) φ11.7mm (Exp.) φ11.7mm (Simu.)
φ10.5mm (Exp.) φ10.5mm (Simu.) φ10.5mm (Exp.) φ10.5mm (Simu.)
φ9.5mm (Exp.) φ9.5mm (Simu.) φ9.5mm (Exp.) φ9.5mm (Simu.)
φ8.5mm (Exp.) φ8.5mm (Simu.) φ8.5mm (Exp.) φ8.5mm (Simu.)
axial strain axial strain
-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0
1.0 1.0
circumferential strain

circumferential strain
0.5 0.5

0.0 0.0

-0.5 -0.5 Fig. 19. Relationship between prestrain and reduction in height at fracture
(a) Equation (3) (b) Equation (4)
initiation for two materials.
Fig. 16. Relationship between axial strain at fracture initiation and cir-
cumferential strain at fracture initiation in various evolution equations for criteria and for both the materials, the axial and circumferential strains
SCM435. at fracture initiation calculated from the simulation considerably differ
from the axial and circumferential strains at fracture initiation obtained
Table 5 in the experiment, especially for large prestrain.
Material constants used in conventional ductile fracture criteria. Fig. 19 shows the relationship between the prestrain and the re-
C2 C3
duction in height at fracture initiation for the two materials. The si-
mulation results calculated in the simulation in Section 4.2 and the
SS400 0.25 0.3 experimental results obtained in the experiment in Section 3.2, as well
S55C 0.25 0.3 as the simulation results calculated using the conventional ductile
fracture criteria, are shown. The simulation results are plotted using
φ13.0mm (Exp.) φ13.0mm (Cock.) φ13.0mm (Exp.) φ13.0mm (Cock.)
open symbols, whereas the experimental results are plotted using solid
φ11.1mm (Exp.) φ11.1mm (Cock.) φ11.1mm (Exp.) φ11.1mm (Cock.) symbols. The initial height/diameter ratio of the specimen is equal to
φ9.5mm (Exp.) φ9.5mm (Cock.) φ9.5mm (Exp.) φ9.5mm (Cock.)
φ8.1mm (Exp.) φ8.1mm (Cock.) φ8.1mm (Exp.) φ8.1mm (Cock.) 1.5. The reduction in height is defined as (hb hf )/ hb , where hb denotes
φ6.9mm (Exp.) φ6.9mm (Cock.) φ6.9mm (Exp.) φ6.9mm (Cock.) the height of the specimen before cylinder upsetting, and hf denotes the
φ5.9mm (Exp.) φ5.9mm (Cock.) φ5.9mm (Exp.) φ5.9mm (Cock.)
φ5.0mm (Exp.) φ5.0mm (Cock.) φ5.0mm (Exp.) φ5.0mm (Cock.) height of the specimen after cylinder upsetting.
axial strain axial strain With increasing the reduction in height in cylinder upsetting, a
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
damage value at the surface of the center in the height direction, at
1.0 1.0
which the specimen fractured, increased monotonically [49], in other
circumferential strain

circumferential strain

0.5
0.5 words, residual ductility at the surface of the center in the height di-
0.0 rection decreased monotonically. Hence, the reduction in height in
0.0 cylinder upsetting could be used as a measure for evaluating residual
-0.5 ductility. For both the materials, with increasing the prestrain, the re-
-0.5 ductions in height at fracture initiation both calculated in the simula-
-1.0
tion in Section 4.2 and obtained in the experiment in Section 3.2 de-
-1.5 -1.0
(a) SS400 (b) S55C crease, whereas the reductions in height at fracture initiation calculated
using the conventional ductile fracture criteria increase. Hence, for both
Fig. 17. Relationship between axial strain at fracture initiation and cir- the materials, with increasing the prestrain, the residual ductility both
cumferential strain at fracture initiation for two materials calculated using
calculated using the ellipsoidal void model and obtained experimen-
Cockcroft and Latham fracture criterion.
tally decreased, whereas the residual ductility calculated using the
conventional ductile fracture criteria increased.
φ13.0mm (Exp.) φ13.0mm (Brozzo) φ13.0mm (Exp.) φ13.0mm (Brozzo) The maximum prestrain in a preceding study on the effect of the
φ11.1mm (Exp.) φ11.1mm (Brozzo) φ11.1mm (Exp.) φ11.1mm (Brozzo)
φ9.5mm (Exp.) φ9.5mm (Brozzo) φ9.5mm (Exp.) φ9.5mm (Brozzo) prestrain on the reduction in height at fracture initiation in cylinder
φ8.1mm (Exp.) φ8.1mm (Brozzo) φ8.1mm (Exp.) φ8.1mm (Brozzo) upsetting [57] was smaller than 0.5, whereas the maximum prestrain in
φ6.9mm (Exp.) φ6.9mm (Brozzo) φ6.9mm (Exp.) φ6.9mm (Brozzo)
φ5.9mm (Exp.) φ5.9mm (Brozzo) φ5.9mm (Exp.) φ5.9mm (Brozzo) this study is approximately equal to two. Because the effect of the
φ5.0mm (Exp.) φ5.0mm (Brozzo) φ5.0mm (Exp.) φ5.0mm (Brozzo)
axial strain axial strain prestrain on the reduction in height at fracture initiation is not ne-
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 cessarily clear in Fig. 19 when the prestrain is below 0.5, the study on
1.0 1.0 the effect of the prestrain on the reduction in height at fracture initia-
tion in the preceding study [57] was confirmed to be insufficient.
circumferential strain

circumferential strain

0.5
0.5
0.0
0.0 5. Conclusions
-0.5

-1.0
-0.5 Surface cracking in cylinder upsetting for five types of ferrous bars
was predicted using the ellipsoidal void model proposed by the author,
-1.5 -1.0 which is able to evaluate ductile fracture in the simulation of metal-
(a) SS400 (b) S55C
forming processes. The following results were obtained:
Fig. 18. Relationship between axial strain at fracture initiation and cir-
cumferential strain at fracture initiation for two materials calculated using (1) A specimen for the plane-strain tensile test obtained from a bar was
Brozzo et al. fracture criterion.
proposed to obtain a material constant on ductile fracture.
(2) Both in the simulation and in the experiment, with increasing the
prestrain, the axial strain at fracture initiation monotonically in-
creased, whereas the circumferential strain at fracture initiation

12
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

monotonically decreased. [25] L. Morin, J.B. Leblond, A.A. Benzerga, Coalescence of voids by internal necking:
(3) The effect of prestrain on the circumferential and axial strains at Thoretical estimates and numerical results, J. Mech. Phys. Solids 75 (2015)
140–158.
fracture initiation calculated using the ellipsoidal void model sub- [26] K. Komori, An ellipsoidal void model for the evaluation of ductile fracture in sheet
stantially agrees with that obtained experimentally. metal forming, Comput. Mater. Sci. 67 (2013) 408–416.
(4) The effect of prestrain on the circumferential and axial strains at [27] K. Komori, Evaluation of ductile fracture in sheet metal forming using the ellip-
soidal void model, Mech. Mater. 77 (2014) 67–79.
fracture initiation calculated using conventional ductile fracture [28] K. Komori, Proposal and use of a void model for inclusion cracking for simulating
criteria considerably differs from that obtained experimentally. inner fracture defects in drawing of ferrite-pearlite steels, Mech. Mater. 77 (2014)
98–109.
[29] K. Komori, Predicting ductile fracture in pure metals and alloys using notched
Hence, the validity of the ellipsoidal void model was confirmed. tensile specimens by an ellipsoidal void model, Eng. Fract. Mech. 151 (2016)
This study did not receive any specific grant from funding agencies 51–69.
in the public, commercial, or not-for-profit sectors. [30] K. Komori, Predicting ductile fracture in ferrous materials during tensile tests using
an ellipsoidal void model, Mech. Mater. 113 (2017) 24–43.
[31] P.F. Thomason, A theory for ductile fracture by internal necking of cavities, J. Inst.
CRediT authorship contribution statement Met. 68 (1968) 360–365.
[32] A. Melander, U. Ståhlberg, The effect of void size and distribution on ductile frac-
Kazutake Komori: Conceptualization, Data curation, Formal ana- ture, Int. J. Fract. 16 (5) (1980) 431–439.
[33] T.B. Stoughton, A general forming limit criterion for sheet metal forming, Int. J.
lysis, Funding acquisition, Investigation, Methodology, Project admin- Mech. Sci. 42 (2000) 1–27.
istration, Resources, Software, Supervision, Validation, Visualization, [34] A. Abedini, C. Butcher, M.J. Worswick, Experimental fracture characterisation of an
Writing - original draft, Writing - review & editing. anisotropic magnesium alloy sheet in proportional and non-proportional loading
conditions, Int. J. Solids Struct. 144–145 (2018) 1–19.
[35] A. Barata da Rocha, J.M. Jalinier, Plastic instability of sheet metals under simple
Declaration of Competing Interest and complex strain paths, Trans. Iron Steel Inst. Japan 24 (1984) 132–140.
[36] K. Enami, The effects of compressive and tensile prestrain on ductile fracture in-
itiation in steels, Eng. Fract. Mech. 72 (2005) 1089–1105.
The authors declare that they have no known competing financial [37] Y. Bai, T. Wierzbicki, Forming severity concept for predicting sheet necking under
interests or personal relationships that could have appeared to influ- complex loading histories, Int. J. Mech. Sci. 50 (2008) 1012–1022.
[38] D. Chae, J.P. Bandstra, D.A. Koss, The effect of pre-strain and strain-path changes
ence the work reported in this paper.
on ductile fracture: experiment and computational modeling, Mater. Sci. Eng., A
285 (2000) 165–171.
References [39] Y. Alinaghian, M. Asadi, A. Weck, Effect of pre-strain and work hardening rate on
void growth and coalescence in AA5052, Int. J. Plast. 53 (2014) 193–205.
[40] K. Komori, Ductile Fracture in Metal Forming – Modeling and Simulation, Academic
[1] S.E. Clift, P. Hartley, C.E.N. Sturgess, G.W. Rowe, Fracture prediction in plastic Press, London, 2019.
deformation processes, Int. J. Mech. Sci. 32 (1) (1990) 1–17. [41] H. Kudo, K. Aoi, Effect of compression test condition upon fracturing of a medium
[2] T. Wierzbicki, Y. Bao, Y.W. Lee, Y. Bai, Calibration and evaluation of seven fracture carbon steel, J. Japan Soc. Technol. Plasticity 8 (72) (1967) 17–27 (in Japanese).
models, Int. J. Mech. Sci. 47 (4–5) (2005) 719–743. [42] P.F. Thomason, The free surface ductility of uniaxial compression specimens with
[3] B. Dodd, Y. Bai, Ductile Fracture and Ductility, Academic Press, London, 1987. longitudinal surface defects, Int. J. Mech. Sci. 11 (1) (1969) 65–73.
[4] A.M. Freudenthal, The Inelastic Behavior of Engineering Materials and Structures, [43] S. Kobayashi, Deformation characteristics and ductile fracture of 1040 steel in
John Wiley & Sons, New York, 1950. simple upsetting of solid cylinders and rings, Trans. ASME J. Eng. Indus. 92 (2)
[5] M.G. Cockcroft, D.J. Latham, Ductility and the workability of metals, J. Inst. Met. (1970) 391–399.
96 (1968) 33–39. [44] P.W. Lee, H.A. Kuhn, Fracture in cold upset forging - A criterion and model, Metall.
[6] P. Brozzo, B. De Luca, R. Rendina, A new method for the prediction of the form- Trans. 4 (4) (1973) 969–974.
ability limits of metal sheets, A new method for the prediction of the formability [45] S.I. Oh, S. Kobayashi, Workability of aluminum alloy 7075–T6 in upsetting and
limits of metal sheets, Research Group, 1972. rolling, Trans. ASME J. Eng. Indus. 98 (3) (1976) 800–806.
[7] M. Oyane, Criteria of ductile fracture strain, Bull. JSME 15 (90) (1972) 1507–1513. [46] R. Sowerby, I. O’Reilly, N. Chandrasekaran, N.L. Dung, Materials testing for cold
[8] A.L. Gurson, Continuum theory of ductile rupture by void nucleation and growth: forging, Trans. ASME J. Eng. Mater. Technol. 106 (1) (1984) 101–106.
Part I-Yield criteria and flow rules for porous ductile media, Trans. ASME J. Eng. [47] S. Kobayashi, C.H. Lee, Deformation mechanics and workability in upsetting solid
Mater. Technol. 99 (1) (1977) 2–15. circular cylinders, in: Proceedings of the First North American Metalworking
[9] V. Tvergaard, A. Needleman, Analysis of the cup-cone fracture in a round tensile Research Conference, (1973) pp. 185–204.
bar, Acta Metall. 32 (1) (1984) 157–169. [48] A. Jenner, Y. Bai, B. Dodd, A thermo-plastic shear instability criterion applied to
[10] Z.L. Zhang, E. Niemi, Studies on the ductility predictions by different local failure surface cracking in upsetting and related processes, J. Strain Anal. 16 (3) (1981)
criteria, Eng. Fract. Mech. 48 (4) (1994) 529–540. 159–164.
[11] Z.L. Zhang, E. Niemi, Analyzing ductile fracture using dual dilational constitutive [49] R. Sowerby, N. Chandrasekaran, N.L. Dung, O. Mahrenholtz, The prediction of
equations, Fatigue Fract. Eng. Mater. Struct. 17 (6) (1994) 695–707. damage accumulation during upsetting tests based on McClintock's model, Forsch.
[12] A. Needleman, A continuum model for void nucleation by inclusion debonding, Ingenieurwes. 51 (5) (1985) 147–150.
Trans. ASME J. Appl. Mech. 54 (3) (1987) 525–531. [50] N. Bontcheva, R. Iankov, Numerical investigation of the damage process in metal
[13] V. Tvergaard, Influence of voids on shear band instabilities under plane strain forming, Eng. Fract. Mech. 40 (2) (1991) 387–393.
conditions, Int. J. Fract. 17 (4) (1981) 389–407. [51] H. Moritoki, Free surface ductility in upsetting, Int. J. Plast. 9 (4) (1993) 507–523.
[14] V.K. Barnwal, S.Y. Lee, J.H. Kim, F. Barlat, Failure characteristics of advanced high [52] B.P.P.A. Gouveia, J.M.C. Rodrigues, P.A.F. Martins, Fracture predicting in bulk
strength steels at macro and micro scales, Mater. Sci. Eng., A 754 (2019) 411–427. metal forming, Int. J. Mech. Sci. 38 (4) (1996) 361–372.
[15] R. Hill, On discontinuous plastic states, with special reference to localized necking [53] H.S. Kim, Y.T. Im, M. Geiger, Prediction of ductile fracture in cold forging of alu-
in thin sheets, J. Mech. Phys. Solids 1 (1) (1952) 19–30. minum alloy, Trans. ASME J. Manuf. Sci. Eng. 121 (3) (1999) 336–344.
[16] H.A. Kuhn, P.W. Lee, Strain instability and fracture at the surface of upset cylinders, [54] H.P. Gänser, A.G. Atkins, O. Kolednik, F.D. Fischer, O. Richard, Upsetting of cy-
Metall. Trans. 2 (11) (1971) 3197–3202. linders: a comparison of two different damage indicators, Trans. ASME J. Eng.
[17] J. Besson, Continuum models of ductile fracture: a review, Int. J. Damage Mech. 19 Mater. Technol. 123 (1) (2001) 94–99.
(1) (2010) 3–52. [55] A.R. Ragab, Fracture limit curve in upset forging of cylinders, Mater. Sci. Eng., A
[18] A.A. Benzerga, J.B. Leblond, Ductile fracture by void growth to coalescence, Adv. 334 (1–2) (2002) 114–119.
Appl. Mech. 44 (2010) 169–305. [56] F.A. McClintock, A criterion for ductile fracture by the growth of holes, Trans.
[19] A.A. Benzerga, J.B. Leblond, A. Needleman, V. Tvergaard, Ductile failure modeling, ASME J. Appl. Mech. 35 (2) (1968) 363–371.
Int. J. Fract. 201 (1) (2016) 29–80. [57] Y. Neishi, S. Watanabe, Y. Haruhata, T. Kuboki, K. Kuroda, Influence of pre-
[20] A. Pineau, A.A. Benzerga, T. Pardoen, Failure of metals I: Brittle and ductile frac- deformation on the workability limit in cold upsetting, J. JSTP 43 (496) (2002)
ture, Acta Mater. 107 (2016) 424–483. 401–405 (in Japanese).
[21] T. Pardoen, J.W. Hutchinson, An extended model for void growth and coalescence, [58] S. Kobayashi, S.I. Oh, T. Altan, Metal Forming and the Finite-Element Method,
J. Mech. Phys. Solids 48 (12) (2000) 2467–2512. Oxford, New York (1989).
[22] M. Gologanu, J.B. Leblond, G. Perrin, J. Devaux, Theoretical models for void coa- [59] K. Komori, An ellipsoidal void model for simulating ductile fracture behavior,
lescence in porous ductile solids. I. Coalescence in layers, Int. J. Solids Struct. 38 Mech. Mater. 60 (2013) 36–54.
(2001) (32–33), 5581–5594. [60] Y. Tomita, Numerical Elasticity and Plasticity, Yokendo, Tokyo, 1990 (in Japanese).
[23] M. Gologanu, J.B. Leblond, J. Devaux, Theoretical models for void coalescence in [61] F.M. Beremin, Cavity formation from inclusions in ductile fracture of A508 steel,
porous ductile solids. II. Coalescence in columns, Int. J. Solids Struct. 38 (2001) Metall. Trans. A 12 (5) (1981) 723–731.
(32–33), 5595–5604. [62] W.D. Callister, Material Science and Engineering An Introduction, sixth edition,
[24] A.A. Benzerga, J.B. Leblond, Effective yield criterion accounting for microvoid John Wiley & Sons, Hoboken, 2003.
coalescence, Trans. ASME J. Appl. Mech. 81 (3) (2014) 031009. [63] L.E. Miller, G.C. Smith, Tensile fracture in carbon steels, J. Iron Steel Inst. 208 (11)

13
K. Komori Theoretical and Applied Fracture Mechanics 108 (2020) 102636

(1970) 998–1005. [67] L.E. Samuels, Optical Microscopy of Carbon Steels, American Society for Metals,
[64] T. Inoue, S. Kinoshita, Three stages of ductile fracture process and criteria of void Ohio, 1980.
inititation in spheroidized and ferrite/pearlite steels, Trans. ISIJ 17 (9) (1977) [68] J. Besson, D. Steglich, W. Brocks, Modeling of crack growth in round bars and plane
523–531. strain specimens, Int. J. Solids Struct. 38 (46–47) (2001) 8259–8284.
[65] D.P. Clausing, Effect of plastic strain state on ductility and toughness, Int. J. Fract. [69] K. Komori, Simulation of tensile test by node separation method, J. Mater. Process.
Mech. 6 (1) (1970) 71–85. Technol. 125–126 (2002) 608–612.
[66] T. Sato, Micrographs of Steels and Their Explanations, second edition, Maruzen, [70] J.H. Schmitt, J.M. Jalinier, Damage in sheet metal forming-I Physical behavior, Acta
Tokyo, 1968 (in Japanese). Metallurgica 30 (9) (1982) 1789–1798.

14

You might also like