Experimental and Numerical Analysis of The Tensile Test Using Sheet Specimens

Available online at www.sciencedirect.
com
Finite Elements in Analysis and Design 40 (2004) 555 – 575

www.elsevier.com/locate/nel
Experimental and numerical analysis of the

tensile test using sheet specimens
Eduardo E. Cabezas, Diego J. Celentano∗
Departamento de Ingeniera Mecanica, Universidad de Santiago de Chile, Av. Bdo. O’Higgins 3363, Santiago, Chile
Received 1 September 2002; received in revised form 28 March 2003; accepted 20 April 2003
Abstract
This paper presents an experimental and numerical study of the mechanical behaviour of SAE 1045 steel
sheet specimens during the conventional tensile test. Due to the complex stress state that develops at the
neck for high levels of axial deformation, an experimental-numerical methodology is proposed in order to
derive the elastic and hardening parameters which characterize the material response. Such methodology is
essentially an extension to sheet samples of the well established procedure used for cylindrical ones. The
simulation of the deformation process during the whole test is performed with a nite element large strain
elastoplasticity-based formulation. Finally, the experimental validation of the obtained numerical results allows
to assess the performance of the proposed methodology for the 3D analysis of sheet specimens and, in addition,
to discuss the range of applicability of plane stress conditions.
? 2003 Elsevier B.V. All rights reserved.
Keywords: Tensile test; Mechanical characterization; Sheet specimens; large strains; Plasticity; Finite element formulations
1. Introduction
The tensile test is an important standard engineering procedure useful to characterize some relevant
elastic and plastic variables related to the mechanical behaviour of materials. Due to the non-uniform
stress and strain distributions existing at the neck for high levels of axial deformation, it has been
long recognized that signicant changes in the geometric conguration of the specimen have to be
∗
Corresponding author. Fax: +56-2-681-7522.
E-mail address: dcelenta@lauca.usach.cl (D.J. Celentano).
0168-874X/$ - see front matter ? 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0168-874X(03)00096-9
556 E.E. Cabezas, D.J. Celentano / Finite Elements in Analysis and Design 40 (2004) 555 – 575
considered in order to properly describe the material response during the whole deformation process
up to the fracture stage.
Although in many engineering applications the design of structural parts is restricted to the elas-
tic response of the materials involved, the knowledge of their behaviour beyond the elastic limit is
relevant since plastic e>ects with usually large deformations take place in the previous manufacturing
procedures such as metal forming (traditionally classied into (a) bulk-forming operations: forging,
drawing, extrusion and rolling and (b) sheet forming operations: deep drawing, magnetic-pulse form-
ing, tube enlarging and bulging [1]). Other important applications of elastoplastic models for metals
are: crashworthiness, impact problems, inelastic buckling of thin-walled structures, superplastic form-
ing, etc. (see, e.g. [2–4]).
The di>use necking process of both cylindrical and sheet samples used in the tensile test has been
extensively studied (see, e.g. [2,5]). In particular, Bridgman [5] derived, based on some geometric
considerations of the deformation pattern, analytical expressions for the stress distribution at the neck
written in terms of the ratio =a at the necking zone ( is the curvature radius of the neck and a is
either its diameter D or its width w for the cylindrical or sheet samples, respectively; see Fig. 1). As
is an extremely diCcult variable to be measured in practice, alternative equations depending only
on a have been also proposed exclusively for cylindrical samples [5]. Therefore, the applicability
of this methodology to Eat bars is rather limited. However, it should be noted that the use of sheet
specimens is the sole possibility to test di>erent strip-manufactured materials as those produced by
means of rolling processes.
In recent years, several nite element large strain formulations usually dened within the plas-
ticity framework have been developed and applied to the analysis of this test under isothermal
and non-isothermal conditions (see, e.g. [4,6–10]). Moreover, some of such formulations have been
validated, generally under isothermal conditions, with experimental data considering cylindrical spec-
imens of di>erent materials. In contrast, there have been only few studies focused on the necking
phenomenon of strips (see, e.g. [11,12]).
The aim of this paper is to present an experimental analysis and a numerical simulation of the
mechanical behaviour during the tensile test experienced by sheet specimens of SAE 1045 steel. The
experimental procedure undertaken to characterize some specic features of the material response
is described in Section 2. Moreover, the governing equations together with the constitutive model
Fig. 1. Analysis of an SAE 1045 steel tension specimen: schematic representation of the necking zone for both cylindrical
and sheet specimens.
E.E. Cabezas, D.J. Celentano / Finite Elements in Analysis and Design 40 (2004) 555 – 575 557
proposed to simulate the deformation process that takes place during the test are presented in Sec-
tion 3. This large strain isotropic elastoplasticity-based formulation includes the denitions of a
specic free energy function and plastic evolution equations which are the basis to derive the
stress–strain relationship and a thermodynamically consistent expression for the internal dissipation.
The corresponding nite element model is brieEy presented in Section 4 where a particular treatment
of the incompressible plastic Eow in order to overcome the well-known volumetric locking in the
numerical behaviour is discussed. It should be mentioned that this nite element formulation is an
alternative approach to existing methodologies dealing with large plastic deformations such as the
standard mixed or the enhanced strain methods (see, e.g. [4,7]).
The experimental hardening characterization and the numerical simulation of the tensile test ap-
plied to both cylindrical and sheet specimens of SAE 1045 steel are performed in Section 5. In
particular, details on the derivation of the parameters involved in the assumed exponential plastic
hardening law are given. The results obtained with the proposed formulation are validated with the
corresponding experimental measurements. Aside from the engineering stress–strain curve, di>erent
results at the section undergoing extreme necking are specically analyzed: ratio of current to initial
diameter in terms of the elongation and both load and mean true axial stress versus logarithmic strain.
Furthermore, computed non-uniform stress components and e>ective plastic deformation contours at
the fracture stage are also shown conrming the strong inEuence of the di>use necking formation
in the material response.
2. Experimental procedure
The experimental procedure adopted in this work to characterize the mechanical behaviour of a
material consisted in the following steps:
(1) Selection of the material and the specimens to be tested according to the ASTM standards
[13]. The chosen material is SAE 1045 steel considering cylindrical and sheet specimens as sketched
in Fig. 2. The distance between the two black markers denotes the initial extensometer length taken
as 50 mm for both cases. A nearly linear gradual reduction in diameter and width (for the cylindrical
and sheet specimens, respectively) is considered in order to trigger the necking development which
has to take place approximately at the middle of the extensometer length. This tapered prole ts
the ASTM standards since the di>erence between the maximum and minimum diameter or width
values existing in the extensometer length is lower than 1%.
Fig. 2. Analysis of an SAE 1045 steel tension specimen: geometric congurations for (a) cylindrical and (b) sheet samples
(dimensions in mm, initial extensometer length = 50 mm and load cell speed of 2:5 mm=min for both cases).
Table 1
Analysis of a SAE 1045 steel tension specimen: average chemical composition (% in weight)
C Si Mn P S
0.447 0.213 0.756 0.0148 0.03272
Cr Mo Ni Al Cu
0.0635 0.0139 0.0914 ¡ 0:00106 0.277
Nb Ti V W Pb
¡ 0:0050 0.00219 0.00917 ¡ 0:01 ¡ 0:005
Sn B Fe Co
0.0159 ¡ 0:00051 98.02 0.0172
900 900
Engineering stress P/Ao [Mpa]
800 800
Engineering stress P/Ao [MPa]
700 700
600 600
500 500
400 400
300 300
200 200
100 100
0 0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24
(a) Engineering deformation (L-Lo)/Lo (b) Engineering deformation (L-Lo)/Lo
Fig. 3. Analysis of an SAE 1045 steel tension specimen: average experimental stress–strain curve for (a) cylindrical and
(b) sheet samples.
(2) Chemical characterization to check an adequate composition according to the selected ma-
terial. This routine task is carried out by means of an optical spectrometer. The average chemical
composition for the studied steel is shown in Table 1.
(3) Mechanical tensile test. The engineering stress–strain curves obtained with ve cylindrical and
sheet specimens considering a load cell speed of 2:5 mm=min (value within the range specied by
the ASTM standards) are plotted in Fig. 3. As usual, the engineering stress is dened as P=A0 ,
where P is the axial load and A0 is the initial transversal area while the engineering strain or
elongation is computed as (L − L0 )=L0 , with L and L0 being the current and initial extensometer
lengths, respectively. At the beginning of the deformation process the material behaves elastically.
After the yield strength is reached, the plastic hardening begins and the load increases up to a
maximum value for a specic elongation. Then, the load decreases since the e>ect of the reduction
Table 2
Analysis of a SAE 1045 steel tension specimen: average experimentally measured values
Sample Yield strength Maximum Maximum Elongation at the

(elongation: 0.2%) load engineering stress fracture stage
(MPa) (kN) (MPa) (%)
Cylindrical 450.4 46.6 749.2 18.5

Sheet 451.6 56.5 762.0 20.0
of the transversal area at the necking zone is stronger than that of the hardening mechanism. Di>use
necking has been developed for both types of samples (localized necking is only observable for very
thin strips; see Tvergaard [11] and Ling [12]). The e>ect of plastic deformation without hardening
developed immediately after the elastic response (usually known as LMuders’ band formation which
is a typical phenomenon for steels with low to moderate carbon contents; see Dieter [14]) can also
be observed in these curves. It should be noted, however, that this fact is almost imperceptible for
the test with sheet samples due to, presumably, the distorting e>ects produced by the machining
operations necessary to make the specimens. The average experimentally measured values for the
yield strength, maximum load, maximum engineering stress and elongation at the fracture stage are
summarized in Table 2.
(4) Characterization of the plastic behaviour. At high levels of elongation, the stress and strain
distributions are no longer uniform along the specimen due to the necking formation that takes place
for both cylindrical and sheet samples. Therefore, the stress–strain curve of Fig. 3 cannot provide a
proper description of the physical phenomena involved in the test. Following the procedure proposed
by Bridgman [5], the mechanical response can be adequately described by an alternative stress–
strain curve dened in terms of the mean equivalent stress O eq versus an equivalent deformation eq
(composed of an elastic and plastic contributions) respectively given by O eq =fB P=A and eq = O eq =E+
p , where fB (p ) 6 1 is an assumed known correction factor applied to the mean true axial stress
P=A, A is the current transversal area at the necking zone (A = D2 =4 or A = wt for the cylindrical
and sheet samples, respectively, where D, w and t are the current diameter, width and thickness of
the neck), E is the Young’s modulus and p = ln(A0 =A) is the true (logarithmic) deformation. A
preliminary step to obtain such stress–strain relationship (in which, as can be seen, D, w and t are
the additional variables to be measured), consists of deriving the P=A − p curve shown in Fig. 4
using four specimens for each type of samples. Moreover, Fig. 5 shows the geometric congurations
of broken cylindrical and sheet specimens at the end of the test.
Details of the application of this procedure are described in Sections 5.1 and 5.3 for cylindrical
and sheet specimens, respectively.
3. Governing equations and constitutive model
The local governing equations describing the evolution of an assumed quasi-static isothermal
process (i.e., that with negligible inertia e>ects and identically fullled energy balance) can be
expressed by the continuity equation, the equation of motion and the dissipation inequality (all of
1200 1200
1000 1000
800 800
P/A [MPa]
P/A [MPa]
600 600
400 400
200 200
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6
(a) True deformation -2ln(D/Do) (b) True deformation ln(Ao/A)
Fig. 4. Analysis of an SAE 1045 steel tension specimen: mean true axial stress versus true (logarithmic) deformation for
(a) cylindrical and (b) sheet samples.
them valid in × , where is the spatial conguration of a body and denotes the time interval
of interest with t ∈ ) respectively written in a Lagrangian description as [15]:
J = 0 ; (1)
∇ · + bf = 0; (2)
Dint ¿ 0; (3)
together with appropriate boundary conditions and an adequate constitutive relation for the Cauchy
stress tensor (which is symmetric for the non-polar case adopted in this work). In these equations,
is the density, J is the determinant of the deformation gradient tensor F(F−1 =1−∇u, where 1 is the
unity tensor, ∇ is the spatial gradient operator and u is the displacement vector), the subscript 0 ap-
plied to a variable denotes its value at the initial conguration 0 , bf is the specic body force vector
and Dint is the internal dissipation which imposes restrictions over the constitutive model denition.
In this framework, a specic Helmholtz free energy function , assumed to describe the material
behaviour during the deformation process, can be dened in terms of some thermodynamic state vari-
ables chosen in this work as the Almansi strain tensor e (e =1=2(1−F−T ·F−1 ), where T is the trans-
pose symbol) and a set of nint phenomenological internal variables k (usually governed by rate equa-
tions with k = 1; : : : ; nint ) accounting for the non-reversible e>ects [9,10,16]. This free energy deni-
tion, based on the Doyle–Ericksen’s approach [17], is only valid for small elastic strains and isotropic
material response, both assumptions being normally accepted for metals and other materials.
Invoking the Coleman’s method [18], the following relationships are obtained: = ( 99e ) and Dint =
k ) where q =− ( 9 ) are the conjugate variables of and, according to the nature of each
qk ∗( DDt k 0 9 k k
internal variable, the symbols ∗ and D(·)=Dt appearing in the previous expressions respectively
indicate an appropriate multiplication and a time derivative satisfying the principle of material frame-
indi>erence [15].
It is seen that the denitions of = (e; k ) and Dk =Dt are crucial features of the model in
order to derive the constitutive equations presented above.
(a)
(b)
Fig. 5. Analysis of an SAE 1045 steel tension specimen: fracture stage for (a) cylindrical and (b) sheet samples.
The internal variables and their corresponding evolution equations are dened in this work within
the associate rate-independent plasticity theory context [3,4]. A possible choice is given by the
plastic Almansi strain tensor ep and the e>ective plastic deformation eO p related to the isotropic strain
hardening e>ect [9,10,16]. The evolution equations for such plastic variables are written as
• 9F • •
Lv (ep ) = ; eO p = ; (4)
9
where is the Kirchho> stress tensor (=J ), Lv is the well-known Lie (frame-indi>erent) derivative,
•
is the rate (or increment in this context) of the plastic consistency parameter computed according
to classical concepts of the plasticity theory [4] and F = F(; eO p ) is the yield function governing
the plastic behaviour of the solid such that no plastic evolutions occur when F ¡ 0. A Von Mises
yield function is adopted:

F = 3J2 − C p ; (5)
where J2 is the second invariant of the deviatoric part of and C p is the plastic hardening function
given by
p
C p = Ap (eOp0 + eO p )n ; (6)
where eOp0 is an assumed initial value of eO p such that C th = Ap eOp0 n with C th being the yield strength
p
dening the material elastic bound. The hardening material parameters Ap and np appearing in the
isotropic strain hardening law (6) are assumed to characterize the material behaviour in the plastic
range (strain rates e>ects are neglected due to, as described in Section 2, the very low load cell
velocity considered in the tensile tests). Their derivation from the experimental-based correlation
of the mean equivalent stress versus equivalent deformation curve will be described in Section 5.
•
Furthermore, in this context the e>ective plastic deformation rate can be also computed as eO p =

2=3dp : dp , where dp = Lv (ep ) is the plastic contribution of the rate-of-deformation tensor d given
•
by d = 12 (∇v + v∇) where v = u is the velocity vector. A consequence of this model is that tr(dp ) = 0
(tr is the trace operator) which reEects the incompressibility nature of the plastic Eow that has been
physically observed in many metals at moderate pressure levels [5].
The following specic free energy function = (e − ep ; eO p ) is proposed:
1 1
= (e − ep ) : C : (e − ep ) + (e − ep ) : C : 0
20 0
1 p 1 th p
− Ap (eOp0 + eO p )n +1 + C eO + 0; (7)
(np + 1)0 0
where C is the isotropic elastic constitutive tensor. This last equation is a partially coupled form
of dening which considers the density at the initial conguration according to the simplication
of the Doyle–Ericksen’s approach [17]. However, the elastic/plastic decomposition of can be
considered nowadays well established since di>erent versions of it have successfully been used in
many engineering applications (see, e.g. [3,4,6–10,16,19]). Moreover, the additive decomposition of
the Almansi strain tensor is recovered in this context through the multiplicative decomposition of
the deformation gradient into elastic and plastic contributions [9]. It should also be mentioned that
the description of the damage and fracture phenomena are not included in the proposed specic free
energy function given by Eq. (7).
The previous denitions allow to derive the stress–strain law (secant or hyperelastic form for the
Cauchy stress tensor) and the expression of the internal dissipation which are respectively given by
1
= [C : (e − ep − eth ) + 0 ]; (8)
J
•
Dint = : Lv (ep ) + (C p − C th )eO p ¿ 0: (9)
The tangent form of the stress–strain law stated by Eq. (8) can be obtained applying standard
procedures of the plasticity theory [4]. Although this rate expression is not strictly needed within
the present hyperelastic context, its derivation is particularly relevant in the computation of the
sti>ness matrix appearing in the nite element formulation described in Section 4. Finally, it is
worth noting that the internal dissipation inequality (9) is e>ectively fullled (i.e., the terms : Lv (ep )
•
and (C p − C th )eO p are separately non-negative for every thermodynamic state) owing to the adopted
denitions for the plastic evolution equations (4).
4. Finite element formulation
The nite element equation derived from the model presented above is brieEy described in this
Section together with some important features of the numerical strategy used to solve the resulting
system of discrete equations.
Following the standard procedures within the nite element framework [20], the global discretized
equilibrium equation including mass conservation can be written in matrix form for a certain time t
(or load level for the present quasi-static case) as
RU ≡ FU − F = 0; (10)
where RU is the residual vector, FU is the external force vector and F denotes the internal force
vector. The element expressions of these vectors are shown in Box 1. Note that RU is computed
at the initial conguration using the well-known total Lagrangian approach [21]. In this context,
all the variables involved in RU have to be transformed to the initial conguration. Moreover,
a unconditionally stable generalized mid-point rule algorithm has been used to integrate the plastic
rate equations presented above via a return-mapping procedure [4,21]. On the other hand, the Jacobian
matrix needed in the Newton–Raphson iterative process is shown in Box 2 where, owing to the strong
non-linearities inherent in the formulation, an approximated but numerically accurate expressions for
JUU is considered where KU is the so-called sti>ness matrix. In this total Lagrangian approach, the
sti>ness matrix consists of two terms usually denoted as the material and geometric contributions
respectively related to the elastoplastic constitutive behaviour and the non-linear e>ects of the adopted
strain measure [21].
Box 1. Element vectors in the discretized equilibrium equation:

ncU
FU(e) = (e)
0
NUT bF0 d0 + (e)
0
NUT tO0 d 0 + Fc(e)
Uj ;
j=1

F(e) = (e) BT S d0 ;
0
where NU is the shape function matrix for displacements, bF0 the body force vector at initial
conguration 0 , tO0 the traction vector at the boundary 0 ⊂ 0 ( 0 = 90 ),
Fc(e)
U the point force vector at element (e) with ncU loaded nodes.
BO the strain–displacement matrix for large strains to avoid numerical locking due to incompress-
ible plastic deformation; see Celentano [16].
S = J F−1 · · F−T : Second Piola–Kirchho> stress tensor.
Superscript T: transpose symbol.
Box 2. Jacobian matrix for the Newton–Raphson iterative procedure:
9RU ∼
JUU = − = KU
9U
with

KU(e) = O T 9S B
B (e) O d0 + (e) HO T SH d0 ;
0 9E 0
where U is the nodal displacement vector, 9S=9E the tangent elastoplastic constitutive tensor at
initial conguration 0 (E: Green–Lagrange strain tensor) and HO the strain–displacement matrix
O
for large strains derived by linearization of B.
Although classical spatial interpolations for the displacement eld have been considered in
Eq. (10), an improved strain–displacement matrix B, O previously proposed by Celentano [16] and
checked in problems involving moderate deformations, is also employed in this work in order to
overcome the volumetric locking e>ect on the numerical solution when incompressible plastic Eows
are studied. The performance of this methodology is now tested in a large strain situation like the
necking process of a cylindrical tension specimen described below. Based on the deformation gradi-
ent standard decomposition into deviatoric and volumetric parts, and assuming a selective numerical
integration for the volumetric part of F, the B O matrix is obtained by linearization of the Green–
Lagrange strain tensor. The expressions of this matrix for the 2D, axisymmetric and 3D cases can
be found in the mentioned reference. It should noted that the B O matrix does not have a sparse
structure. Nevertheless, the additional computations required at element level were found not to sig-
nicantly increase the CPU times in comparison with the standard sparse strain-displacement matrix.
This methodology is an alternative approach to the assumed strain mixed nite element method de-
veloped by Simo and Armero [7] in which a sparse gradient operator is obtained with the drawback
of computing and storing, at element level, enhanced strain parameters dened in such context. Note
that these last operations are not needed in the present B O algorithm and, hence, a simple computa-
O
tional implementation may be attained. Moreover, this B approach is also di>erent from the widely
used constant pressure formulation [6,10] which is a classical mixed method dened within the con-
text of a three-eld variational formulation of the Hu-Washizu type (see Simo [4] for a complete
discussion on this subject). The main advantage of this last formulation is apparent when dealing
with incompressible plastic Eow problems since the resulting general weak form can be simplied
in this case by treating the pressure and deviatoric responses in an uncoupled form.
5. Experimental characterization and numerical simulation
The main objective of the present analysis is to validate the predictions of the proposed formulation
with the experimental data obtained in the tensile tests of SAE 1045 steel in order to achieve an
adequate mechanical characterization of this material when either cylindrical or sheet specimens
are considered. To this end, the experimental procedure described in Section 2 is rstly applied to
cylindrical specimens. The nite element formulation presented above is used later to simulate the
material behaviour during the tension deformation process (Section 5.1). Using the same material
properties derived from the experiments with cylindrical specimens, the simulation of the tensile
test using sheet specimens is subsequently performed (Section 5.2). An experimental validation of
the numerical predictions is included for both simulations. Afterwards, a correction factor obtained
from the numerical analysis with sheet specimens allows to carry out the experimental hardening
characterization applied to this type of samples for the sake of checking the appropriateness of such
procedure (Section 5.3). Finally, this analysis also includes an assessment on the validity range of
plane stress conditions for sheet specimens (Section 5.4).
5.1. Experimental hardening characterization and simulation using cylindrical specimens
The experimental procedure detailed in Section 2 is applied to SAE 1045 steel cylindrical spec-
imens to characterize its plastic hardening behaviour. The correction factor fB (p ) proposed by
Bridgman [5] for cylindrical specimens is plotted in Fig. 6 (p = −2 ln(D=D0 ) for this case).
np
Fig. 7 shows the O eq − eq average experimental data and the potential correlation O eq = Ap eq derived
from it where Ap and np are, as mentioned in Section 3, the hardening parameters. These hardening
parameters are the basic data for the numerical simulation and experimental validation presented
below.
The nite element formulation described in Section 4 is used to simulate the material behaviour
during the tensile test. According to the experimental measurements performed during these tests,
the material properties considered in the numerical analysis for SAE 1045 steel are shown in
Table 3. The spatially non-uniform nite element mesh shown in Fig. 8 have been chosen in order
to correctly describe the large stress and deformation gradients expected in the necking zone. As-
suming axisymmetry, a fourth of the specimen is discretized with a height of 25 mm (half of the
initial extensometer length) and a linear radius variation along the bar according to the geometry
specications depicted in Fig. 2a where Utop denotes the axial displacement imposed at the top
boundary up to a value which corresponds to the fracture elongation (see Table 2). Moreover, all
the simulations are performed with bf = 0 and 0 = 0.
0.95
Correction factor
0.9
0.85
Bridgman's factor
0.8
0.75
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
True deformation -2ln(D/Do)
Fig 6. Analysis of an SAE 1045 steel tension specimen: Bridgman [5] correction factor as a function of true deformation
for a cylindrical sample.
1200
1000
Equivalent stress [MPa]

800
600 σ eq = 1047.7 ε eq0.1206

400 R2 = 0.9763
200
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Equivalent deformation
Fig 7. Analysis of an SAE 1045 steel cylindrical tension specimen: mean equivalent stress versus equivalent deformation.
Table 3
Analysis of a SAE 1045 steel tension specimen: material prop-
erties considered in the numerical simulations
Young’s modulus E 222000 (MPa)

Poisson’s ratio # 0.30
Yield strength C th 450 (MPa)
Hardening coeCcient Ap 1047.7 (MPa)
Hardening exponent np 0.1206
Fig 8. Analysis of an SAE 1045 steel cylindrical tension specimen: axisymmetric nite element mesh used in the simulation
(mesh composed of 360 four-noded isoparametric elements).
Fig. 9 shows the engineering stress–strain relationship and some results at the section undergoing
extreme necking: the radii relation versus the elongation in the necking zone together with the load
and mean true axial stress both against the logarithmic deformation. An overall good agreement
between the numerical predictions and the average experimental values can be observed in these
1
900
800 0.95
700 0.9
600 0.85
D/Do
500
0.8
400
0.75
300
simulation 0.7 simulation
200
experimental experimental
100 0.65
0 0.6
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24
50 1200
45
1000
40
35 800
Load P [kN]
P/A [MPa]
30
25 600
20
400
15
10 simulation simulation
200
5 experimental experimental
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
(c) True deformation-2ln(D/D0) (d) True deformation -2ln(D/Do)
Fig 9. Analysis of an SAE 1045 steel cylindrical tension specimen: (a) Engineering stress–strain relationship. Results at
the section undergoing extreme necking: (b) ratio of current to initial diameter versus axial elongation, (c) load versus
true deformation and (d) mean true axial stress versus true deformation.
curves. The discrepancies appearing in the engineering stress–strain curve can be mainly attributed
to the inaccuracy of the potential correlation at the beginning of the plastic region where no hardening
is produced.
The experimentally measured load decreases from an elongation of 11.5% or, equivalently, to
a logarithmic deformation p of 11.2% onwards (the corresponding deformations provided by the
simulation are 12.4% and 11.7%, respectively). However, the mean true axial stress continues in-
creasing until the fracture stage where a big amount of plastic hardening can be appreciated. This
fact conrms that a geometrical instability occurs (instead of a constitutive instability) since, as
already commented in Section 2, the e>ect on the stress caused by the reduction of the transversal
area at the necking zone predominates over the material hardening. At high level of deformations,
the regions of the specimen outside the necking zone are being elastically unloaded. Moreover, note
that the well-known simplied relationship [14], stating that the related logarithmic deformation at
the point of maximum load has to be equal to the hardening exponent, is approximately veried
(see Table 3 and Fig. 7).
0.95
Correction factor
0.9
0.85
Bridgman's factor
simulation
0.8
0.75
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
True deformation -2ln(D/Do)
Fig 10. Analysis of an SAE 1045 steel tension specimen: simulated correction factor as a function of true deformation
for a cylindrical sample.
The numerical predictions for the ratio of current to initial diameter in terms of the elongation starts
with a linear relationship, reEecting uniform distributions of stresses and strains, which presents an
approximate slope of 0.5 due to the incompressibility nature of the plastic Eow. The same situation is
kept up to an elongation of 12.5% which corresponds, as mentioned above, to the point of maximum
load. Afterwards, a sudden reduction of the diameter takes place causing the necking formation and,
hence, non-homogeneous stress and strain distributions along the specimen. As can be seen, the
numerical results t reasonably well the experimental ones during the whole test even though the
inherent diCculty associated with the measurement of the diameter at the neck.
Fig. 10 compares the correction factor in terms of true deformation proposed by Bridgman [5] with
that obtained through the numerical simulation as the quotient between the average equivalent stress
at the neck and the mean true axial stress P=A. It is observed that both curves present practically
the same response (the maximum discrepancy existing within the range of deformation 0 –57%
experienced by the material as shown in Fig. 9 is less than 1%). A uniform stress distribution,
expressed by the condition fB ≈ 1, is obtained for p 6 0:1. For larger deformations, triaxial stresses
are developed and, therefore, the e>ect of the correction factor on P=A becomes relevant.
Di>erent stress components (hoop, radial, axial and shear√ components of together with the
pressure p=1=3 tr() and the equivalent stress eq given by 3J2 ) and e>ective plastic deformation
contours at the end of the analysis (fracture stage) can be found in Fig. 11. Non-uniform distributions
are clearly obtained due to the complex deformation pattern of the neck. As expected, the maximum
values of eq and consequently of eO p are concentrated in the neck. Note that values around 0.10
of e>ective plastic deformation mainly found at the rear part of the specimen indicate the level of
uniform deformation experienced until the maximum load is reached (see Fig. 9). Moreover, the
maximum of eO p nearly coincides with the value of true deformation −2 ln(D=D0 ) at fracture. The
neck formation even induces the development of low pressures at the center of the bar (this fact has
also been pointed out by Armero and Simo [8] and Goicolea et al. [10] in the tension analysis of
other materials). Furthermore, some assumptions considered in the analytical study of Bridgman [5]
at the neck are ratied by the simulation, e.g., eq and eO p are approximately constant, rr and %%
present a strong variation but the condition rr ≈ %% is fulllled, zz ≈ eq + rr ¿ eq that explains
Fig 11. Analysis of an SAE 1045 steel cylindrical tension specimen: stress (MPa) and e>ective plastic deformation contours
at the end of the simulation corresponding to the fracture stage for an elongation of 18.5%.
Fig 12. Analysis of an SAE 1045 steel sheet tension specimen: 3D nite element mesh used in the simulation (mesh
composed of 3440 eight-noded isoparametric elements).
the need to correct the stress distribution, p ≈ eq =3 + rr , ezz ≈ −2err up to the maximum load and
err ≈ e%% during the whole test.
5.2. Simulation using sheet specimens
The previously described nite element formulation is also used to simulate the material response
during the tensile test using sheet specimens considering the same properties derived from the ex-
periments with cylindrical samples (see Table 3). The 3D nite element mesh is shown in Fig. 12
where, due to symmetry, only one eighth is studied (note that this mesh presents a similar element
densication as that shown in Fig. 8). Fitting the specications of Fig. 2b, the neck in the middle of
the specimen is triggered by assuming a linear width variation along its length. As in the previous
analysis, Utop denotes the prescribed axial displacement imposed at the top boundary up to fracture.
The engineering stress–strain relationship together with some results at the necking zone (width
and thickness ratios versus the elongation and the load and mean true axial stress both against the
logarithmic deformation) are plotted in Fig. 13. Once more, an overall good agreement between
900 1
800
Engineering stress P/Ao [Mpa]
0.95
700
Width and thickness ratios

0.9
600
0.85
500
0.8
400
0.75 w/wo: simulation
300
simulation w/wo: experimental
0.7 t/to: simulation
200
experimental t/to: experimental
100 0.65
0 0.6
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24
1200
60
1000
50
800 40
Load P [kN]
P/A [MPa]
600 30
400 20
simulation
simulation
experimental
200 10 experimental
0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.0 0.1 0.2 0.3 0.4 0.5 0.6
(c) True deformation ln(Ao/A) (d) True deformation ln (Ao/A)
Fig 13. Analysis of an SAE 1045 steel sheet tension specimen: (a) Engineering stress–strain relationship. Results at the
section undergoing extreme necking: (b) ratio of current to initial width and thickness (mean values along thickness and
width, respectively) versus axial elongation, (c) load versus true deformation and (d) mean true axial stress versus true
deformation.
the numerical predictions and the average experimental values can be observed in these curves. The
discrepancies appearing in the engineering stress–strain curve can be also attributed to the inaccuracy
of the potential correlation at the beginning of the plastic region where little hardening is produced.
Although some small discrepancies in the load and mean true axial stress appear at higher lev-
els of deformation, it should be noted that such di>erences are approximately bounded within the
experimental uncertainty range.
The experimentally measured load decreases from an elongation of 12.5% or, equivalently, to
a logarithmic deformation p of 11.8% onwards (the corresponding deformations provided by the
simulation are 13.0% and 12.2%, respectively). A geometrical instability caused by the necking
formation is also observed for this case since the mean true axial stress continues increasing until
the fracture stage.
The numerical predictions for the mean w=w0 and t=t0 ratios match each other at the beginning of
the test since a uniaxial stress is achieved for a low level of elongation in the range 0 –13.0%. Once
Fig 14. Analysis of an SAE 1045 steel sheet tension specimen: stress (MPa) and e>ective plastic deformation contours
at the end of the simulation corresponding to the fracture stage for an elongation of 20.0%.
the neck is formed, both curves progressively di>er where, as in the experiments, the reduction in
t=t0 is stronger than that of w=w0 .
The six components of , the pressure p, the equivalent stress eq and e>ective plastic deformation
contours at the fracture stage are plotted in Fig. 14. Once again, a non-uniform fully triaxial stress
0.98
correction factor
0.96
Correction factor
0.94
0.92
0.9
0.88
0.86
0.84
0.82
0.8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
True deformation ln (Ao/A)
Fig 15. Analysis of an SAE 1045 steel tension specimen: simulated correction factor as a function of true deformation
for a sheet sample.
Table 4
Analysis of a SAE 1045 steel tension specimen: correlation factor as a function of true deformation
ln(A0 =A) Correction factor
Cylindrical sample Sheet sample
0.10 1.000 1.000

0.20 0.978 0.976
0.30 0.957 0.955
0.40 0.938 0.933
0.50 0.921 0.909
0.60 0.905 0.884
0.70 0.890 0.858
0.80 0.876 0.830
0.90 0.863 0.800
1.00 0.851 0.769
distribution is developed at the neck where, as expected, the maximum values of eq and eO p are found
in this zone. In particular, note that a very strong variation of the e>ective plastic deformation is
found at the neck. As reported by Ling [12], the maximum of eO p is, in contrast to the cylindrical case,
markedly greater than the value of true deformation ln(A0 =A) at fracture (see Fig. 13). Moreover,
only positive pressures are obtained for this thick strip sample.
Fig. 15 shows the correction factor in terms of true deformation obtained from the simulation.
This curve is also presented in Table 4 and compared to that corresponding to cylindrical specimens.
For a given value of logarithmic strain, note that the e>ect of the correction factor is larger for sheet
specimens than that for cylindrical ones. A similar trend has been found by Bridgman [5] for the
fB − =a relationship (see Fig. 1). However, the derived correction factor of Table 4 for obtaining
the adequate stress–strain curve for Eat samples is an alternative approach of practical interest since,
1000
900
800
Equivalent stress [MPa]

700
600
500 σ eq = 1024. 0 ε eq
0.1240
400 R2 = 0.9791
300
200
100
0
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45
Equivalent deformation
Fig 16. Analysis of an SAE 1045 steel sheet tension specimen: mean equivalent stress versus equivalent deformation.
as already commented, the experimental tasks needed to obtain the current radius of curvature at
the neck are relatively more complex than the measurement of the width w and thickness t at the
minimum cross section.
5.3. Experimental hardening characterization using sheet specimens
The consideration of the proposed correction factor of Fig. 15 allows the derivation of the ex-
perimental O eq − eq curve for sheet specimens. The average measured data together with the corre-
np
sponding potential correlation O eq = Ap eq are both shown in Fig. 16. As can be appreciated, these
hardening parameters are practically the same to those of Fig. 7 for cylindrical specimens. This
proposed methodology shows, therefore, that an adequate characterization of the material response
can be also achieved for sheet samples.
5.4. Assessment of plane stress conditions for sheet specimens
In order to evaluate the range of validity of plane stress (PS) conditions in sheet samples during
the tensile test, a comparative study between a full 3D and a simplied 2D (with zz = xz = yz = 0;
see Fig. 12) simulations is carried out. Fig. 17 shows the engineering stress–strain curve and the
correction factor in terms of true deformation for the two analyses. As can be seen, both predictions
are practically the same up to the instant at which the necking process starts (see Fig. 13b). From
this point onwards, the PS results provide an unrealistic response since, as observed in Fig. 18a, the
condition zz = 0 is no more fullled. This is due to the fact that the outward unit normal to the
specimen, whose direction coincides with the z axis at the initial conguration, rotates at high levels
of elongation for which, as can be appreciated in Fig. 18b, a strong thickness reduction at the neck
is developed. Finally, note that the typical plane stress relation exx = ezz = − 12 eyy is only veried
before the necking development.
800 1.00
0.95
700
0.90
600
Correction factor
0.85
500 0.80
400 0.75
0.70
300
simulation 3D 0.65
200
simulation PS 0.60 simulation 3D
100 simulation PS
0.55
0 0.50
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
(a) Engineering deformation (L-Lo)/Lo (b) True deformation ln Ao/A
Fig 17. Analysis of an SAE 1045 steel sheet tension specimen: comparison between 3D and plane stress simulations for
(a) engineering stress–strain relationship and (b) correction factor.
30
2
1.9
Thickness/Axial deformation ratio
edge: simulation 3D 1.8

25 1.7
center: simulation 3D 1.6
1.5 simulation 3D
Stress ZZ [MPa]
20 1.4 simulation PS
1.3
1.2
1.1
15 1
0.9
0.8
10 0.7
0.6
0.5
5 0.4
0.3
0.2
0.1
0 0
0 0.03 0.06 0.09 0.12 0.15 0.18 0.21 0.24 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24
Fig 18. Analysis of an SAE 1045 steel sheet tension specimen. Results at the necking zone: (a) Cauchy stress component
zz (mean values along thickness) versus elongation for a 3D simulation and (b) comparison between 3D and plane stress
simulations for the thickness/axial deformation ratio |ln(t=t0 )|=ln(L=L0 ) against elongation.
6. Conclusions
Experimental and numerical analyses of the mechanical behaviour occurring in cylindrical and
sheet specimens during the standard tensile test applied to SAE 1045 steel have been presented. A
characterization of the material response using cylindrical specimens has been rstly performed in
order to obtain the stress–strain curve and the diameter evolution at the neck which allow, in turn,
to derive the elastic and hardening parameters by means of a well-established methodology. Then, a
nite element large strain elastoplasticity-based formulation has been proposed and used to simulate
the tensile deformation process in such type of samples. Moreover, these material properties have
been considered in the subsequent numerical analysis of the tensile test using sheet specimens. The
results provided by both simulations have been satisfactorily validated with experimental data. In
particular, the necking formation, which was found to be di>use for both cases, has been properly
described. Afterwards, a successful experimental hardening characterization carried out for sheet
specimens using a correction factor deduced from the simulation has been performed. Finally, the
proposed methodology has enabled to achieve an adequate description of the mechanical response
of the material during the whole tensile test when using either cylindrical or sheet specimens.
Acknowledgements
The supports provided by the Chilean Council of Research and Technology CONICYT (FONDE-
CYT Project No. 1020026) and the Department of Technological and Scientic Research at the
University of Santiago de Chile (DICYT-USACH) are gratefully acknowledged. The authors wish
to express their appreciation to the Aeronautical Technical Academy at Santiago de Chile for the
provision of experimental facilities.
References
[1] S. Talbert, B. Avitzur, Elementary Mechanics of Plastic Flow in Metal Forming, Wiley, New York, 1996.
[2] A. NTadai, Theory of Flow and Fracture of Solids, McGraw-Hill Book Company, London, 1950.
[3] J. Lubliner, Plasticity Theory, Macmillan Publishing, New York, 1990.
[4] J. Simo, Topics on the Numerical Analysis and Simulation of Plasticity, in: Handbook of Numerical Analysis, Vol.
III, Elsevier Science Publishers, Amsterdam, 1995.
[5] P. Bridgman, Studies in Large Plastic and Fracture, McGraw-Hill Book Company, London, 1952.
[6] P. Wriggers, C. Miehe, M. Kleiber, J. Simo, On the coupled thermo-mechanical treatment of necking problems via
nite-element-method, Proceedings of COMPLAS II, 1989, pp. 527–542.
[7] J. Simo, F. Armero, Geometrically non-linear enhanced strain mixed methods and the method of incompatible modes,
Int. J. Numer. Methods Eng. 33 (1992) 1413–1449.
[8] F. Armero, J. Simo, A priori stability estimates and unconditionally stable product formula algorithms for non-linear
coupled thermoplasticity, Int. J. Plast. 9 (1993) 149–182.
[9] G.C. GarcTUa, J. Oliver, A numerical model for elastoplastic large strain problems. Fundamentals and applications,
Proceedings of COMPLAS III, Barcelona, 1993, pp. 117–122.
[10] J. Goicolea, F. GabaldTon, G.C. GarcTUa, Analysis of the tensile test using hypo and hyperlastic models, Proceedings
of the III Congress on Numerical Methods in Engineering, Zaragoza, 1996, pp. 875 –885 (in Spanish).
[11] V. Tvergaard, Necking in tensile bars with rectangular cross-section, Comput. Methods Appl. Mech. Eng. 103 (1993)
273–290.
[12] L. Ling, Uniaxial true stress–strain after necking, AMP J. Technol. 5 (1996) 37–48.
[13] ASTM 8 EM, “Standard test methods for tension testing of metallic materials”, Annual Book of Standards,
Section 3: Metals test methods and analytical procedures, Vol. 03.01, ASTM, Philadelphia, 2001.
[14] G. Dieter, Mechanical Metallurgy—SI Metric Edition, McGraw-Hill Book Company, London, 1988.
[15] L. Malvern, Introduction to the Mechanics of a Continuous Medium, Prentice-Hall, Englewood Cli>s, NJ, 1969.
[16] D. Celentano, A large strain thermoviscoplastic formulation for solidication of S.G. cast iron in a green sand mould,
Int. J. Plast. 17 (2001) 1623–1658.
[17] T. Doyle, J. Ericksen, Nonlinear elasticity, Advances in Applied Mechanics, Vol. 4, Academic Press, San Diego,
1956.
[18] B. Coleman, M. Gurtin, Thermodynamics with internal state variables, J. Chem. Phys. 47 (2) (1967) 597–613.
[19] D. Celentano, D. Gunasegaram, T. Nguyen, A thermomechanical model for the analysis of light alloy solidication
in a composite mould, Int. J. Solids Struct. 36 (1999) 2341–2378.
[20] O. Zienkiewicz, R. Taylor, The Finite Element Method, 4th Edition, Vols. 1 and 2, McGraw-Hill, London, 1989.
[21] M. Criseld, in: Non-linear Finite Element Analysis of Solids and Structures, Vols. 1 and 2, Wiley, Chichester,
1991.

Experimental and Numerical Analysis of The Tensile Test Using Sheet Specimens

Încărcat de

Informații document

Descriere originală:

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Experimental and Numerical Analysis of The Tensile Test Using Sheet Specimens

Încărcat de

Drepturi de autor:

Formate disponibile

Available online at www.sciencedirect.

Finite Elements in Analysis and Design 40 (2004) 555 – 575

Experimental and numerical analysis of the

(a) Engineering deformation (L-Lo)/Lo (b) Engineering deformation (L-Lo)/Lo

Sample Yield strength Maximum Maximum Elongation at the

Cylindrical 450.4 46.6 749.2 18.5

3. Governing equations and constitutive model

4. Finite element formulation

Box 1. Element vectors in the discretized equilibrium equation:

Box 2. Jacobian matrix for the Newton–Raphson iterative procedure:

5. Experimental characterization and numerical simulation

5.1. Experimental hardening characterization and simulation using cylindrical specimens

Equivalent stress [MPa]

600 σ eq = 1047.7 ε eq0.1206

Young’s modulus E 222000 (MPa)

5.2. Simulation using sheet specimens

Width and thickness ratios

(a) Engineering deformation (L-Lo)/Lo (b) Engineering deformation (L-Lo)/Lo

ln(A0 =A) Correction factor

Cylindrical sample Sheet sample

0.10 1.000 1.000

Equivalent stress [MPa]

5.3. Experimental hardening characterization using sheet specimens

5.4. Assessment of plane stress conditions for sheet specimens

edge: simulation 3D 1.8

S-ar putea să vă placă și