Optics and Lasers in Engineering 110 (2018) 122–131

Contents lists available at ScienceDirect

Optics and Lasers in Engineering

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

Virtual field method for identifying elastic-plastic constitutive parameters

of aluminum alloy laser welding considering kinematic hardening
Ruixiang Bai a, Hao Jiang a, Zhenkun Lei a,∗, Da Liu a, Yang Chen a, Cheng Yan b, Wang Tao c,
Qiaoling Chu d
a
State Key Laboratory of Structural Analysis for Industrial Equipment, Dalian University of Technology, Dalian 116024, China
b
School of Chemistry, Physics and Mechanical Engineering, Queensland University of Technology, Brisbane 4001, Australia
c
State Key Laboratory of Advanced Welding and Joining, Harbin Institute of Technology, Harbin 150001, China
d
College of Materials Science and Engineering, Xi’an University of Technology, Xi’an 710048, China

a r t i c l e i n f o a b s t r a c t

Keywords: The local high temperature of laser welding affects the mechanical properties of aluminum alloys. In this paper, a
Elastic-plastic constitutive parameters three-dimensional digital image correlation system was used to measure the displacement field of the aluminum
Kinematic hardening alloy-laser welding joint under a uniaxial tensile load. By taking into consideration the metal properties of welded
Virtual field method
joints, an estimation correction scheme for updating stress was designed for anisotropic materials based on the
3D-DIC
kinematic hardening process, and the proposed method was then applied to the nonlinear virtual field method, the
Laser welding
validity of which was verified by a finite element simulation. The elastic-plastic constitutive parameters of various
local regions of the welding joint (fusion zone, heat-affected zone, transition base zone, and base zone) were then
retrieved from the measured field data, and the stress–strain relationship of each sub-region was established.
The results show that the mechanical properties of the laser-welded joint are weakened while the Poisson’s ratio
decreases with the increase of the weld distance. Furthermore, the elastic modulus of the heat-affected zone is
evidently lower than that of the other sub-regions, while it also has the lowest yield strength and the highest
plastic strain rate. All these results are consistent with the trend of the hardness measurements.

1. Introduction
Laser welding technology [1], as an important manufacturing pro-
cess of an integral plate, can greatly reduce the structural weight and
production costs, as well as improve welding efficiency, thus leading to
the development of integral plate production. An advanced laser weld-
ing technology is a good solution to the problem of the excessive-heat-
affected area, and it could produce a heat source with a high energy
density such that the welding material has a very narrow heat-affected
area and thus minimize the deformation and strength loss of the heat-
affected zone. However, the laser welding process still has an impor-
tant effect on the connection quality between the beam and skin, while
the thermal effect of the welding degrades the local material proper-
ties, resulting in inevitable thermal deformation and residual thermal
stress in the welded joint [2–4]. During the welding process, the high-
temperature gradient changes the organization of the weld area, which
leads to a change in the material properties. At the microscopic level,
this change in the properties of the material is caused by the transforma-
tion of the microscopic crystal structure of the metal under the action
of the laser but the mechanical parameters of the metal from the micro-
scopic crystal structure transform, which is a very difficult and cumber-
some process. Therefore, it is necessary to develop an appropriate means
of obtaining measurements at the macroscopic level in order to charac-
terize the local mechanical properties of the weld joint. This study aids
in finding the fragile areas of the overall panels while providing reliable
allowable-load values. Furthermore, it helps to optimize the selection of
the welding parameters and manufacturing processes in order to reduce
the thermal deformation caused by the welding and provide accurate
material parameters for the simulation of the mechanical behavior of
the welding parts.
From the microstructure point of view, the welding parts can be di-
vided into three main areas: the fusion zone, heat-affected zone, and
parent metal zone. However, the mechanical parameters of the material
still have large differences even in the same area. A classic and common
method is to use hardness as an indicator for characterizing the degree of
weakening of the mechanical properties in various positions of the ma-
terial. However, this method cannot be used to accurately establish the
relationship between the hardness value and Young’s modulus or Pois-
son’s ratio [5]. Song et al. [6] used the ultrasonic to estimate the Young’s
modulus and Poisson’s ratio around the welding joint of a titanium-alloy

∗
Corresponding author.
E-mail address: leizk@dlut.edu.cn (Z. Lei).

https://doi.org/10.1016/j.optlaseng.2018.05.017
Received 20 January 2018; Received in revised form 3 May 2018; Accepted 19 May 2018
0143-8166/© 2018 Elsevier Ltd. All rights reserved.
R. Bai et al.
test piece and found that as the distance from the fusion line increases,
the Young’s modulus gradually increases while the Poisson’s ratio de-
creases. Ambriz et al. [7] conducted an indentation experiment at the
fusion zone of the aluminum alloy material and found that the Young’s
modulus in the fusion zone was smaller than that in the basic zone.
In recent years, the methods of obtaining optical measurements, es-
pecially the digital image correlation method (DIC), by virtue of its non-
contact, low test-environment requirements, strong noise resistance,
and other advantages, have become an important means for welding-
deformation test measurement. Yoneyama et al. [8] measured the strain
field of the tensile aluminum-alloy specimen by using DIC techniques
and established stress–strain curves in accordance with the Ramberg–
Osgood exponential elasto-plastic constitutive model, and the J-integral
of the exponential hardening model was calculated from the displace-
ment field using path integration. Reynolds and Duvall [9] used DIC
to obtain the stress–strain curves at various locations of welding parts,
and it was found that when the welding nugget undergoes strain hard-
ening, the area away from the welding joint tends to yield slightly while
the plastic strain of the base material region remains unchanged. Tex-
ier et al. [10] measured the deformation of the welding joint using DIC
but did not obtain the constitutive parameters of the weld. In the elas-
tic range, Acar et al. [11] used DIC to study the vertical strain of the
welding test piece and found that the larger matching subset of DIC
tends to achieve more desirable measurement accuracy for weld-strain
measurement. Yan et al. [12] found that the stress at the welding core
was higher than that at the base zone. Dong et al. [13] used the inte-
grated DIC method (I-DIC) to determine the thermodynamic properties
of the weld. Sutton et al. [14] evaluated the inhomogeneous proper-
ties of the welds using the uniform stress method (USM) and the virtual
field method (VFM) after obtaining the strain field data of each zone of
the weld. Saranath and Ramji [15] used the field data measured using
DIC to invert the elasto-plastic mechanics parameters in various zones
of the weld and conducted tests using the results of microscopic imag-
ing, followed by the application of the USM to obtain some constitutive
parameters and their changing laws in the welding region of a titanium
alloy.
The DIC method provides the dense field of displacement and strain
data. Owing to the gradient changing material properties of the welds in
different regions, this field displacement and strain data are unevenly
distributed. Therefore, using this nonuniform field data to accurately
mine the mechanical properties of the laser welding zone is a challeng-
ing endeavor. Finite element model updating (FEMU) [16,17] involves
the establishment of a finite element model with the initial value of the
material constitutive parameters in order to calculate the displacement
information of each node, then the setting up of an evaluation function
of the obtained displacement with the measured field data, and the re-
vision of these constitutive parameters by gradual iteration to minimize
the difference. FEMU is widely used and very flexible; however, it has
four shortcomings. Firstly, this method relies heavily on the degree of fit-
ness between the numerical model and the actual experiment; however,
in the actual experiment, the boundary load distribution is very difficult
to determine. Secondly, each iteration of FEMU for updating the model
parameters is required to be re-calculated once using the finite element
method (FEM), and thus the calculation is very time-consuming. Thirdly,
the optimal iteration often cannot converge as it is highly dependent on
the initial settings of the deformation parameters. Finally, its data-noise
immunity is very poor because of the lack of statistical basis. In recent
years, an integrated digital image correlation (I-DIC) has been used to
measure coefficient of thermal expansion, multiple thermo-mechanical
parameters and residual stress [18,19], and it is an effective inversion
method.
Another method based on field data that can be used to solve the me-
chanical inverse problems is the VFM [20]. Compared with the FEMU,
the VFM is usually required to solve a small-scaled linear system with-
out any iterative process. VFM is based on the virtual work equation
and the relationship between the material parameters is established by
123
Optics and Lasers in Engineering 110 (2018) 122–131
Fig. 1. The tension specimen with a weld in the middle.
assuming a series of kinematically admissible (KA) virtual displacement
fields. Avril et al. [21] proposed a method of setting the KA virtual dis-
placement field such that it is immune to noise effects, while Grédiac
summed up the criteria for setting several types of virtual displacement
fields [22]. Avril et al. [23] proposed the basic VFM for the inversion
of the mechanical parameters of the elasto-plastic materials, and Notta-
Cuvier et al. [24] improved his method using coupling identification of
the damaged-material parameters. Valeri et al. [25] used the nonlinear
VFM to identify the general visco-plastic response of the Johnson-Cook
constitutive model. Pierron proposed a VFM scheme for calculating the
material parameters of composites [26] and a thick plate [27]. As an
interesting application, Cao and Xie [28] determined the principal axis
of the fused deposition modeling materials through characterizing an
orthotropic constitutive model of behavior by using the elastic VFM. In
order to identify the elasto-plastic mechanical parameters of welding
materials, Sutton et al. [14] used VFM to obtain the material param-
eters of the unevenly deformed zone around the weld joint. Louedec
et al. [29] studied the elastic-plastic behaviors of friction-stir-welding
test pieces at various strain rates. Their results show that the yield stress
and hardening modulus of the material at the welding joint are signifi-
cantly lower than those at the base zone. Saranath and Ramji [30] used
the USM and VFM to invert the elasto-plastic constitutive parameters
of titanium alloy welds and found that the fusion zone had the highest
yield stress and hardening modulus with the lowest Young’s modulus of
all the zones.
This study is divided into three parts. First, uniaxial tensile tests of an
aerospace aluminum 6061 standard part and a laser-welded part were
conducted to obtain the load–displacement curve. At the same time,
the strain field data of the laser-welded specimen was measured using
a 3D digital image correlation technique (3D-DIC). Along with the mi-
crostructure, the structure around the weld line is partially divided into
four sub-zones. Then, based on the kinematic elastic-plastic constitutive
model, a nonlinear VFM for the inversion of constitutive parameters is
established. The stress-tensor updating estimation correction algorithm
based on the constitutive model of kinematic hardening was proposed,
which was applied in the nonlinear VFM to identify these elasto-plastic
constitutive parameters. After being verified using the FEM, the method
proposed in this paper was used to inversely identify the distribution of
kinematic elastic-plastic constitutive parameters in each sub-region of
the 6061 laser-welded aluminum alloy.
2. Tensile mechanical behavior of laser-welded joints
2.1. Load–displacement curve
Laser-welded samples were fabricated at the State Key Laboratory
of Advanced Welding and Joining using an aluminum alloy 6061 plate
of a thickness of 2 mm. In the welding process, the welding rate was
maintained at 3 m/min with a welding power of 1900 W using ER4047
aluminum wire of a diameter of 1.2 mm as the welding filler. After the
welding work was completed, a dog bone tensile test specimen was pre-
pared from the aluminum alloy plate using wire cutting equipment while
the residual height was required to be removed. The size of the specimen
is shown in Fig. 1.
Before performing the microscopic observation, the cross-section of
the specimen was polished, and corrosion was carried out using hy-
R. Bai et al.
Fig. 2. Microstructure of weld fusion line obtained using metallographic micro-
scope (200×).
Fig. 3. (a) Experimental setup and (b) load–displacement curves.
drofluoric acid. The corroded specimen was then observed under a met-
allographic microscope (Leica MEF4A) as shown in Fig. 2. It can be ob-
served from the microstructure diagram that the welding zone is divided
into three parts: the fusion zone (FZ), heat-affected zone (HAZ), and base
zone (BZ). The majority of the fusion area comprises the foundry struc-
ture. Chu et al. analyzed the microstructure of this laser-welded joint by
using electron backscattered diffraction (EBSD) and inverse pole figures
(IPFs), which revealed that the HAZ has coarse columnar crystals and
the grain structure of the FZ is much rougher than that of the BZ, see
more details in [31]. In order to more accurately reflect the material
properties, a transition base zone (TBZ) is established between the HAZ
and the BZ, where the crystals gradually shift from fine grains to coarse
columnar ones. So, the four zones are divided in the study according to
their crystals’ structure.
Before the experiment, the aluminum alloy specimen was first spot-
ted using spray speckle, and after drying it, the specimen was fixed on
a stretching machine (Instron 3345) as shown in Fig. 3(a). A cold light
source is used for lighting and the tensile speed is set at 0.5 mm/min.
Fig. 3(b) shows the uniaxial tensile load and displacement curves of the
aluminum alloy specimens with and without the laser welding, both of
which show significant elastic-plastic mechanical behaviors.
2.2. Uneven strain field
While the uniaxial tensile test was performed, as shown in Fig. 3(a),
the surface speckle images of the specimen were collected using a 3D-
DIC system with a frame rate of 2 fps and an image size of 2048 × 1536
pixels. After the collection, the system is calibrated to determine the
internal and external parameters of the 3D-DIC system [32]. A commer-
cial software of DIC-3D (PMLAB, Nanjing Zhongxunwei Co.) [33], was
used to reduce the effect of out-of-plane displacement on the obtained
measurements.
124
Optics and Lasers in Engineering 110 (2018) 122–131
As shown in Fig. 4(a), the 3D-DIC calculation was performed in the
rectangular frame of the aluminum welded specimen to obtain the full-
field strain distribution at different moments. Moreover, Figs. 4(b–d)
show the longitudinal, transverse, and tangential strain of the central
axis at a yield time, whose deformation movies can be downloaded (see
𝜀x .avi,𝜀y .avi and 𝜀xy .avi). It can be observed that the strains in the three
directions of the welded specimen are not uniform, the strain values
around the welding joint are obviously higher than those of the base
material, and there is a gradient change in the strain. These indicate
that the laser welding affects the local mechanical properties of the alu-
minum alloy plate.
The longitudinal strain data of the center axis (the dotted line in
Fig. 4(a)) were extracted as shown in Fig. 5(b), the HAZ at the junction of
the welding joint and the base material exhibit the greatest strain, which
means that the HAZ may first enter the yield stage followed by the FZ,
and BZ has the lowest strain value. Furthermore, the horizontal strain
is distributed evenly around the welding joint and the tangential strain
in the weld has an “X-type” distribution (Fig. 4(d)). As the displacement
load increases, the final failure of the specimen occurs in the HAZ on
one side of the weld as shown in Fig. 6.
2.3. Constitutive models
2.3.1. Elastic constitutive model
Assuming that the elastic behavior of the material is isotropic and
subject to the 2D Hooker’s law, at any position of the welded specimen,
the linear relationship between the stress tensor 𝝈 and the strain tensor
𝜺 is given as follows
⎛𝜎1 ⎞ ⎡𝑄1 𝑄2 0 ⎤⎛𝜀1 ⎞
𝝈 = 𝑸𝜺 𝑖.𝑒. ⎜𝜎 ⎟ = ⎢𝑄 𝑄1 0 ⎥⎜𝜀 ⎟, (1)
⎜ 2⎟ ⎢ 2 ⎥⎜ 2 ⎟
⎝𝜎6 ⎠ ⎣ 0 0 𝑄1 − 𝑄2 ⎦⎝𝜀6 ⎠
𝜕𝑢
where 𝜀1 = 𝜕𝑥 ,𝜀2 = 𝜕𝑣
𝜕𝑦
𝜕𝑢
and 𝜀6 = 𝜕𝑦 𝜕𝑣
+ 𝜕𝑥 . In order to facilitate the use
of VFM to obtain the local elastic parameters of the material, such as
Young’s modulus E and Poisson’s ratio v, two stiffness coefficients are
defined as
𝐸 𝑣𝐸
𝑄1 = and 𝑄2 = = 𝑣𝑄1 . (2)
1 − 𝑣2 1 − 𝑣2
Then the following relationship is obtained as
𝑄21 − 𝑄22 𝑄2
𝐸= and 𝑣= . (3)
𝑄1 𝑄1
2.3.2. Plastic constitutive model
In the mechanical analysis and calculation of the welded structure,
the isotropic hardening plastic constitutive and the kinematic harden-
ing plasticity constitutive are two commonly used constitutive model
[34]. These theories not only can accurately describe the yield and hard-
ening of the material, but also are able to predict the mechanical be-
havior of the material under cyclic loading conditions. Among them
the kenimatic hardening model can accurately characterize the yield
strength and hardening performance of the material under the reverse
load, which can reflect the phenomenon that once the general metal is
hardening along one direction, it will be softening along the opposite
direction. Therefore, the kinematic hardening model has a significant
effect on the prediction of residual stresses and that of plastic strain.
In the theory of stress increment, the stress is gradually updated as
the load step increases, and this renewal process is usually done in the
stress space. In the 2D or 3D stress space, once the stress reaches the
yield state, a yield surface is established in the stress space to describe
the yield criterion. The common yield surface equation was set up using
the Tresca yield criterion and the von Mises yield criterion [34]. If the
strengthening of the material continues in the plastic state, then a yield
function considering the hardening effect is required to be derived. The
kinematic hardening yield function can be written as
𝑓 = (𝜎 − 𝑋)𝑒𝑞 − 𝜎0 ≤ 0, (4)
R. Bai et al. Optics and Lasers in Engineering 110 (2018) 122–131

Fig. 4. (a) Displacement fields of the area of interest (AOI) region and cloud map of the measured strain fields of (b) 𝜀x (c) 𝜀y (d) 𝜀xy .

where 𝜺p refers to the plastic strain, 𝜶 is the back stress tensor, and 𝜆 is
called the plastic multiplier, which directly drives the plastic strain rate
amplitude. The total stress tensor and its deviator tensor are respectively
represented as 𝜎= ̄ 𝝈 – X and s = s𝜎 – sX , where s𝜎 and sX are the derivative
stress tensors of 𝝈 and X.
Before the damage occurs, the strain is still in the plastic stage when
the total strain is composed of the elastic strain and the plastic strain as
𝜺 = 𝜺e + 𝜺p , and the stress increment is only related to the elastic strain,
which means,
√
3 ̇ − 12
𝜎̇ 𝑖𝑗 = 𝐸𝑖𝑗𝑘𝑙 𝜀̇ 𝑒𝑘𝑙 = 𝐸𝑖𝑗𝑘𝑙 (𝜀̇ 𝑘𝑙 − 𝜀̇ 𝑝𝑘𝑙 ) = 𝐸𝑖𝑗𝑘𝑙 (𝜀̇ 𝑘𝑙 − 𝜆𝐽2 𝑠𝑘𝑙 ). (8)
2
According to the theory of kinematic hardening and with the known
material parameters, there can be a linear relationship between the in-
Fig. 5. Strain distribution at the center axis at a yield moment. cremental kinematic stress tensor and the back stress tensor. In other
word, there is a material constant 𝛾 such as
1
−
𝑿̇ = 𝛾2 𝜶̇ = 𝜆(
̇ 𝐶 𝐽 2 𝒔 − 𝛾𝑿 ),
2
(9)
where 1 𝛾 2 and 𝛾 = 𝛾 1 𝛾 2. Most importantly, when the material
C = 3−1/2 C
has a plastic yield flow, it is necessary to ensure that the following in-
cremental equation holds in order to guarantee that the yield function f
Fig. 6. Position of damage. equals 0 during this period of time.
𝜕𝑓
𝑓̇ = 𝝈̇ + 𝜕𝜕𝑓𝑿 𝑿̇
where 𝝈 is the stress tensor, X is the kinematic stress tensor, 𝜎 0 is the 𝜕√𝝈
1 √ 1 √ 1
3 −2 3 ̇ −2 3 −2 −1
yield stress, ( · )eq denotes the equivalent stress, and C is the hardening = 𝐽 𝑠𝑖𝑗 𝐸𝑖𝑗𝑘𝑙 (𝜀̇ 𝑘𝑙 − 𝜆𝐽 𝑠 𝑘𝑙 ) − 𝐽 𝑠 𝑖𝑗
̇ 𝐶𝐽 2 𝑠𝑖𝑗
𝜆( − 𝛾𝑋𝑖𝑗 )
2 2 2 2 2 2 2
parameter. f < 0 corresponds to the elastic stress state while f = 0 refers =0
to the plastic stress state. This study uses the von Mises equivalent stress (10)
𝜎 eq expressed in the stress deviator sij .
√ and the solution is
√ 3
𝜎𝑒𝑞 = 3𝐽2 = 𝑠 𝑠 , 𝑠𝑖𝑗 𝐸𝑖𝑗𝑘𝑙 𝜀̇ 𝑘𝑙
2 𝑖𝑗 𝑖𝑗
(5) 𝜆̇ = √ 1
3 −2 −1
𝐽 𝑠𝑖𝑗 𝐸𝑖𝑗𝑘𝑙 𝑠𝑘𝑙 +𝐶 𝐽2 2 𝑠𝑖𝑗 𝑠𝑖𝑗 −𝛾𝑠𝑖𝑗 𝑋𝑖𝑗
where sij = 𝜎 ij – 𝛿 ij 𝜎 kk , and 𝛿 ij is the Kronecker symbol. 2 2
𝑠𝑖𝑗 𝐸𝑖𝑗𝑘𝑙 𝜀̇ 𝑘𝑙 (11)
=
Based on the flow rule, 𝝈 and X can be updated by updating the √
3 −2
1 −1
𝐽 𝑠𝑖𝑗 𝐸𝑖𝑗𝑘𝑙 𝑠𝑘𝑙 +2𝐶 𝐽2 2 −𝛾𝑠𝑖𝑗 𝑋𝑖𝑗
following dual variables. 2 2

√ After calculating the plastic multiplier, the stress tensor, kinematic

𝜕𝑓 3 ̇ − 12
𝜺̇ 𝑝 = 𝜆̇ = 𝜆𝐽2 𝒔. (6) stress tensor, and hardening parameters can be renewed. In practice, in
𝜕𝝈 2 order to adapt to the iterative search of the nonlinear VFM, an estima-
√
The equivalent plastic strain can be easily get by 𝑝̇ = 2 𝜀̇ 𝑝 𝜀̇ 𝑝 = 𝜆,
̇
3 𝑖𝑗 𝑖𝑗
tion correction updating algorithm is designed to obtain a more accurate
then according to literature [35]: stress sequence, by introducing a stress supplement factor 𝜂 and a cor-
rection parameter 𝛽 (see Appendix for details). The deformation and
2
𝜶̇ = 𝐶 𝜺̇ 𝑝 − 𝛾1 𝑝̇ 𝜶, (7) movement of the yield surface in the stress space are shown in Fig. 7.
3 1

125
R. Bai et al. Optics and Lasers in Engineering 110 (2018) 122–131

Fig. 7. Change in the yield surface with kinematic hardening.

3. Virtual field methods

3.1. Principle of virtual work

The VFM uses the virtual work equation as the starting point [36].
It uses the measured strain field as the input data, and the integral form
of the virtual work equation is discretized such that it is suitable for the
entire field measurement signal. For different problems, a series of KA
virtual displacement fields are assumed, and the virtual work equation Fig. 8. Geometric description for the tensile and necking deformations.
that contains the Young’s modulus, Poisson’s ratio, and other material
constants can be weakened to solve these parameters. In the quasi-static the virtual field equation can be written as a linear elastic constitutive
state, regardless of the volume force, the classical virtual field equation relation.
has the following analytic form,
∫𝑆 𝝈𝜺∗ 𝑑𝑆 = ∫𝑆 𝜺𝐷𝜺∗ 𝑑𝑆
( ) ( )
𝝈𝜺 𝑑𝑉 =
∗
𝑻̄ 𝒖∗ 𝑑𝑠, (12) = 𝑄1 ∫𝑆 𝜀1 𝜀∗1 + 𝜀2 𝜀∗2 + 2𝜀6 𝜀∗6 𝑑𝑆 + 𝑄2 ∫𝑆 𝜀1 𝜀∗2 + 𝜀2 𝜀∗1 − 2𝜀6 𝜀∗6 𝑑𝑆 (15)
∫𝑉 ∫𝜕𝑉
= 𝐹̄ 𝑢 ∗

where V is the integral region, 𝑻̄ is the external load, 𝜕Vis the load distri-
There are two variables Q1 and Q2 in the above equation. In the VFM
bution area, and 𝜺∗ and u∗ are respectively the virtual strain tensor and
analysis of the elastic material parameter inversion for each sub-region,
virtual displacement vector. This equation describes the equilibrium re-
two types of displacement fields are set as the KA virtual displacement
lationship between the given virtual field and external force, where only
fields. One describes the tensile deformation of the sub-region, and the
the first order continuous differentiable condition is satisfied for the vir-
other describes the sub-region’s necking deformation, both of which
tual displacement field. In the plane stress problem, when the thickness
were used by Sutton et al. [14] and Saranath and Ramji [30] to obtain
of the plate h is known, the corresponding virtual field equation can be
the elastic parameters. Fig. 8 shows these two deformation modes.
written as follows
Virtual field 1 (tensile):
ℎ 𝝈𝜺∗ 𝑑𝑆 = ℎ 𝑻̄ 𝒖∗ 𝑑𝑠. (13) ⎧𝜀∗ = 0
∫𝑆 ∫𝜕𝑆 { ∗
𝑢 =0 ⎪ 1∗
In the uniaxial tensile test of a laser welding specimen, the tensile ⇒ ⎨𝜀2 = 1 (16)
𝑣∗ = 𝑦 ⎪𝜀∗ = 0
force is obtained from the total force 𝐹̄ = ℎ ∫𝜕𝑆 𝑇̄2 𝑑𝑠; thus, in this prob- ⎩ 6
lem, the virtual field equation can be simplified as
Virtual field 2 (necking):
ℎ 𝝈𝜺∗ 𝑑𝑆 = 𝐹̄ 𝑢∗ . (14) { ⎧𝜀 ∗ = 𝑦 2 − 1 𝑏 2
∫𝑆 ⎪ 1
𝑢∗ = 𝑥(𝑦2 − 14 𝑏2 ) 4
In the process of stretching, the strain field distribution (Fig. 4) from ⇒ ⎨𝜀∗2 = 1 (17)
𝑣∗ = 𝑦 ⎪𝜀∗ = 𝑥𝑦
the DIC shows that the strain distribution has a significant relationship ⎩ 6
with the distance from the welding joint, which is symmetrical about After plugging the two virtual displacement fields into the virtual
the welding cross line, and the texture of the aluminum alloy material field equation, a linear system can be established as
changed along the longitude. In contrast, it can be observed that the [ ]
strain in the transverse direction is roughly distributed evenly, showing [ ∫𝑆 𝜀2 𝑑𝑆 ] [ ∫𝑆 𝜀1 𝑑𝑆 ]
1 1
that the horizontal material properties in did not change significantly. ∫𝑆 𝜀1 (𝑦2 − 4 𝑏2 ) + 𝜀2 + 2𝑥𝑦𝜀6 𝑑𝑆 ∫𝑆 𝜀1 + 𝜀2 (𝑦2 − 4 𝑏2 ) − 2𝑥𝑦𝜀6 𝑑𝑆
Based on the above analysis in this section, an assumption can be { } {1 }
made that the entire tensile process is quasi-static and does not take into 𝑄1 ℎ
𝐹𝑙
= . (18)
account the effect of the volume force. Moreover, each of the material 𝑄2 1
𝐹𝑙
ℎ
constants concluding Young’s modulus, Poisson’s ratio, and the plastic
parameters can be expressed as a continuous function of the stretching 3.3. Identification for plastic material parameters
direction coordinate (i.e., the ordinate y).
In the elastic stage, the Young’s modulus E and Poisson’s ratio v in
3.2. Identification for elastic material parameters each of the sub-regions can be obtained after calculating Q by using
Eq. (3). When the laser-welded specimen reached the plastic stage, Avril
As the tensile test specimen can be regarded as a thin plate, it is a et al. [23] and Saranath and Ramji [30] proposed the use of simple vir-
plane stress problem. Within the overall elastic deformation of 0.1%, tual displacement field settings in the plastic parameters identification.

126
R. Bai et al. Optics and Lasers in Engineering 110 (2018) 122–131

Virtual field (for determination of plastic parameters): 4. Results and discussion

{ ⎧𝜀 ∗ = 0 4.1. Verification using FEM

𝑢∗ = 0 ⎪ 1∗
𝑣∗ = 𝑦
⇒ ⎨𝜀 2 = 1 . (19)
⎪𝜀 ∗ = 0 The important function of the VFM is to extract the material attribute
⎩ 6 parameters that are suitable for the finite element calculation. In this
part, a finite element simulation is used to verify that the incremen-
By collecting the strain field and load data of several time nodes
tal process of the kinematic hardening stress in Appendix can be used
in the nonlinear deformation stage, an objective function of the least
and performs well in the nonlinear VFM in order to identify the plastic
squares problem can be constructed as follows.
material parameters (𝜎 0 , C, and 𝛾).
2 The numerical simulation process is realized by finite element mod-
⎧ ∑𝐾 ⎫ eling in the ANSYS 17.0 platform and a rectangular thin plate with a
⎪ℎ ∫𝑆 𝜎2 (𝜎0𝑘 , 𝑋𝑘 , 𝐶𝑘 , 𝛾𝑘 , 𝑡𝑖 )𝑑𝑆 − 𝐹 (𝑡𝑖 )𝐿 ⎪
∑ ⎪ 𝑘=1 𝑘 ⎪ radius of 50 mm is chosen for the numerical simulation with a 1/4
min 𝐺(𝜎0 , 𝐶, 𝛾) = ⎨ ⎬ , (20)
𝐹 ( 𝑡 ) 𝐿 model. The reason why this model was chosen is that under the uni-
𝑖 ⎪ 𝑖 ⎪
⎪ ⎪ axial tensile load, an uneven strain field is generated near the hole to
⎩ ⎭ verify the effectiveness of this virtual field method. a displacement load
of 0.04 mm is applied at one end of the plate in the y direction, then a
where ti is the acquisition time of the strain field in the stretching
reverse displacement load of −0.08 mm is applied at this edge to reach
test, 𝜎 0 = (𝜎 01 , 𝜎 02 , …,𝜎 0 K ), C = (C1 , C2 , …,CK ), 𝛾 = (𝛾 1 , 𝛾 2 , …, 𝛾 K ) and
the compressed state. The whole process is divided equally into 180 load
X = (X1 , X2 , …, XK ) are the parameter vectors.
steps, while the other end sets the symmetry constraint. Meanwhile, the
Calculated using the incremental iterative scheme of stress (see Ap-
Young’s modulus E = 7e10, Poisson’s ratio v = 0.33, and the yield stress
pendix for details), the vertical direction stress 𝜎 2 (𝜎 0 k , Xk , Ck , 𝛾 k , ti ) has
is set as 𝜎 0 = 134 MPa. The model of shell 181 is taken as the element
strong non-linear characteristics. Thus, the analytical formula of the op-
attribute with the actual parameter of 0.001. After performing the cal-
timization problem (Eq. (20)) is very complex and can be viewed as
culation using FEM, the results of displacement, strain, and load at each
a black box optimization problem with the input variables 𝜎 0 , C, and
step are stored. The triangular mesh node distribution and the strain and
𝛾 and with the output variable G. This problem can be solved by us-
stress distribution of a load step are shown in Fig. 9.
ing the Nelder–Mead simplex method [37], which is a local optimiza-
In this numerical simulation, in order to speed up the computation,
tion method and requires a sufficiently good initial condition to obtain
the meshing should be completed before the VFM identification. The
the optimal solution. In this laser welding problem, each sub-region
triangular mesh grids used here are automatically generated by ANSYS
has independent elastic-plastic parameters, thus causing the Eq. (20) to
as shown in Fig. 9(a). In the plastic VFM calculation, the following least
have too many variables to solve. In order solve this problem, a sub-
squares problem is solved using the Nelder–Mead simplex method.
problem for each region is required to be set up according to the Cauchy– { }2
Schwartz inequality, ∑ ℎ ∫𝑆 𝜎𝑦 (𝜎0 , 𝐶, 𝛾, 𝑡𝑖 )𝑑𝑆 − 𝐹 (𝑡𝑖 )𝐿
min 𝐺(𝜎0 , 𝐶, 𝛾) = . (23)
𝑖
𝐹 (𝑡𝑖 )𝐿
2
⎧ ∑ 𝐾 ∑𝐾 ⎫
⎪ℎ ∫ 𝜎̄ 2 (𝜎0𝑘 , 𝑋𝑘 , 𝐶𝑘 , 𝛾𝑘 , 𝑡𝑖 )𝑑𝑆 − 𝐹 (𝑡𝑖 )𝐿𝑘 ⎪ With the increase in the load step, the updating process was fin-
∑ ⎪ 𝑘=1 𝑆𝑘 𝑘=1 ⎪ ished using the kinematic hardening stress increment in Appendix.
min ⎨ ⎬
𝑖 ⎪ 𝐹 ( 𝑡 𝑖 ) 𝐿 ⎪ The y-directional stress–strain curves of five elements (namely Elm341,
⎪ ⎪ Elm368, Elm745, Elm888 and Elm1017 as shown in Fig. 9(a)) are ob-
⎩ ⎭
{ }2 tained by using the nonlinear VFM, as shown in Fig. 10(a), which are
∑∑ 𝐾 ℎ ∫ 𝑆𝑘 𝜎
̄ 2 ( 𝜎 0 𝑘 , 𝑋 𝑘 , 𝐶 𝑘 , 𝛾 𝑘 , 𝑡 𝑖 )𝑑𝑆 − 𝐹 ( 𝑡 𝑖 ) 𝐿 𝑘
≤ min (21) very close to the FEM calculated ones. Fig. 10(b) gives the convergence
𝑖 𝑘=1 𝐹 (𝑡𝑖 )𝐿 curve of the iterative process. Therefore, the method performs well in
{ }2
𝐾 ∑
∑ ℎ ∫𝑆 𝜎̄ 2 (𝜎0𝑘 , 𝑋𝑘 , 𝐶𝑘 , 𝛾𝑘 , 𝑡𝑖 )𝑑𝑆 − 𝐹 (𝑡𝑖 )𝐿𝑘 the case of this problem.
𝑘
≤ min Because of the influence of illumination, vibration and other factors
𝑘=1 𝑖 𝐹 (𝑡𝑖 )𝐿𝑘
𝐾 in experiment, as well as the accuracy limit of DIC as a numerical algo-
∑
= min 𝑔(𝜎0𝑘 , 𝐶𝑘 , 𝛾𝑘 ), rithm, the deformation field data provided by DIC usually have different
𝑘=1 degrees of noise. In this part, the noise test of the algorithm under this
model is given below, by adding different degrees of Gaussian noise,
𝐾
∑
where Lk is the length of the kth subregion, and 𝐿 = 𝐿𝑘 is the total with signal-noise ratios from 0% to 30%, to the original stain fields. An
𝑘=1 energy loss rate is defined as the following equation to determine the
length of the AOI region. Thus, for the kth sub-region, the sub-problem anti-noise ability of the algorithm.
can be written as | 𝒓𝒆 𝒇 | | 𝒓𝒆 𝒇 |
∫𝑆 ∫Γ |(𝝈 𝒊𝒋 − 𝝈 𝒊𝒊𝒅𝒋 𝒆 )𝑑 𝜀𝑖𝑗 |𝑑 𝑠 ∫𝑆 ∫𝑇 |(𝝈 𝒊𝒋 − 𝝈 𝒊𝒊𝒅𝒋 𝒆 )𝜀̇ 𝑖𝑗 |𝑑 𝑡𝑑 𝑠
𝜀 | | | |
{ }2 𝐸 𝐿𝑅 = = , (24)
∑ ℎ ∫𝑆 𝜎̄ 2 (𝜎0𝑘 , 𝑋𝑘 , 𝐶𝑘 , 𝛾𝑘 , 𝑡𝑖 )𝑑𝑆 − 𝐹 (𝑡𝑖 )𝐿𝑘 | 𝒓𝒆 𝒇 | | 𝒓𝒆 𝒇 |
∫𝑆 ∫Γ |𝝈 𝒊𝒋 𝑑 𝜀𝑖𝑗 |𝑑 𝑠 ∫𝑆 ∫𝑇 |𝝈 𝒊𝒋 𝜀̇ 𝑖𝑗 |𝑑 𝑡𝑑 𝑠
𝑘 𝜀 | | | |
min 𝑔(𝜎0𝑘 , 𝐶𝑘 , 𝛾𝑘 ) = .
𝐹 (𝑡𝑖 )𝐿𝑘
𝑖 where Γ𝜀 is the history of the strain, 𝝈 ref and 𝝈 ide are respectively the
(22) reference stress tensor and the identified stress tensor. The energy loss
rate seems to have a linear relationship with the signal-noise ratio, as
When using VFM to identify the plastic constitutive parameters, the shown in Fig. 11, and moreover, when the signal-noise ratio is below
Nelder–Mead simplex method is usually applied. However, when solv- 1%, the energy loss rate is less than 2%. Therefore, the algorithm is
ing the nonlinear optimization problem by using the simplex method, recommended when the signal-noise ratio is less than 5%.
it is not easy to converge the solution when the number of variables is
greater than 15. Therefore this paper first uses the Nelder–Mead simplex 4.2. Identification of elasto-plastic parameters
method to obtain the solution of the sub-problems (Eq. (22)), and these
solutions are then plugged into the original problem as the initial value In this section, the welding cross line and its aforementioned AOI
for generating the initial population, and finally, the global solution is region, as shown with dot lines in Fig. 4(c), are set as the new area re-
obtained using the particle swarm algorithm [38]. quired to be identified, which is equidistantly divided into 14 sub-zones

127
R. Bai et al. Optics and Lasers in Engineering 110 (2018) 122–131

Fig. 9. (a) 1/4 FEM model mesh with 𝜀y strain field and (b) the corresponding 𝜎 y stress field under the tensional displacement load of 0.04 mm.

Fig. 10. (a) The identified 𝜀y –𝜎 y curves and (b) the changing of square error in iteration.

Fig. 11. The relationship between the signal-noise ratio and energy loss rate.
of a length of 0.4 mm. After identifying each sub-zone using VFM, the
distribution of the elasto-plastic parameters along the y direction is ob-
tained. Firstly, the elastic parameters of the 14 sub-zones were identified
using the elastic VFM, and the distributions of the elastic modulus E and
the Poisson’s ratio v with the distance from the welding cross line are
shown in Fig. 12. From the inversion results, it can be seen that the elas-
tic modulus of the BZ is the highest, with an average of 72 MPa, followed
by that of the TBZ being between 35–44 MPa, and then by that of the
FZ with approximately 40 MPa. The elastic modulus of HAZ is between
25–65 MPa, which has the lowest one of all the regions. The Poisson’s
ratio of the FZ is approximately 0.4, followed by those of the HAZ and
TBZ values between 0.22 and 0.35, and the BZ has the lowest Poisson’s
ratio at approximately 0.2. The result of the elastic modulus is close to
that of Bai’s research [39], and the result of the Poisson’s ratio is very
similar to that of Song’s work [6].
Using the VFM of the nonlinear constitutive parameter identification
for isotropic materials mentioned in Section 3, the AOI area that is com-
posed of 14 sub-zones is identified in order to obtain the kinematic plas-
tic constitutive parameters. The stress increment process in Appendix

128
R. Bai et al.
Fig. 12. Elastic modulus and Poisson’s ratio distributions from the FZ to the BZ.
is used in the nonlinear VFM in order to calculate the distribution of the
yield strength (Fig. 13(a)) and those of the kinematic hardening param-
eters C and 𝛾 (Fig. 13(b)). The calculation results are very interesting,
although the HAZ is the most vulnerable from the perspective of the fi-
nal break position; the FZ and TBZ have yield strengths higher than that
of the HAZ, which indicates that the HAZ firstly enter its material yield
state under the tensile load. However, with the increase in the load, the
different elastic moduli and the differences between the parameters C
and 𝛾 cause the plastic strain rate of the HAZ to be much higher than
those of the FZ and TBZ, which results in the final destruction in the
HAZ.
In order to more intuitively describe the nonlinear relationship of
the individual positions that are different from the welding joint, the
stress–strain curves of the seven sub-zones closest to the welding cross
line are smoothed and their results are shown in Fig. 14. From these
curves, it can be observed that all these sub-zones seem to enter the
nonlinear material state simultaneously. Furthermore, the plastic strain
rate of the location with a distance of 0.4 mm from the weld line is
the highest of all the sub-zones, but another sub-zone (1.2 mm from the
weld) of the HAZ has an evidently different hardening rate from the one
of the sub-zone (0.8 mm), which confirms that even in the same region,
the material properties are very distinct. In addition, the area near the
BZ in the TBZ has the highest yield strength, followed by the welding
joint. This identification results also show that the yield strengths of the
HAZ is obviously lower than the neighbor FZ and TBZ. Meanwhile, the
HAZ has the highest plastic strain rate, which may be one of the reasons
why the final failure occurs in the HAZ.
Fig. 13. Distributions of (a) the yield strength and (b) the hardening parameters C and 𝛾.
129
Optics and Lasers in Engineering 110 (2018) 122–131
Fig. 14. Stress–strain curves of various zones of the weld in the tensile direction.
Bai et al. [39] obtained the stress–strain curve corresponding to some
points from the longitudinal strain field at different moments. Based on
the experimental results, the stress–strain curves of 12 points on the ver-
tical line of the welding cross line of the welded specimens show that
the welding joint and HAZ exhibit typical elasto-plastic mechanical be-
havior. The stress state begins entering into the plastic stage from the
elastic stage when the strain is approximately 0.1%, which is nearly half
of the no-welding aluminum alloy’s 0.2%. On using a Vickers microme-
ter hardness tester (MVC-1000B) to measure the cross-section hardness
of the microscopic observation sample, it was found that the hardness
distribution on the vertical line has an approximately "w" shape. Obvi-
ously, the hardness of the BZ is the highest followed by that of the FZ,
and the HAZ’s hardness is the lowest.
In this paper, the division of the AOI region is obtained from the
strain field cloud map. It can be observed from Fig. 5 that there is a
significant over-platform (called the TBZ) between the HAZ and BZ,
and thus, this division is different from Bai’s. From the inversion re-
sults in Fig. 13(a) it can be observed that the TBZ also has a lower yield
strength, such that there is a necessity for further research on it. How-
ever, different division strategies do not affect the inversion results of
the distribution of material properties. Their work [39] on the hardness
distribution is similar to the variation trend of the elastic modulus in
this paper and is a little different to that of the yield strength of this
study. Actually, the hardness is a comprehensive indicator for reflecting
stiffness, strength, and other mechanical properties of a material. The
elastic modulus and yield strength also respectively reflect the stiffness
and strength of the material, such that there is a certain link between
them. It is believed that the lower yield strength in the FZ and HAZ de-
R. Bai et al. Optics and Lasers in Engineering 110 (2018) 122–131

termines the lower hardness in the zones while the variation trend of the surface, it is necessary to determine the boundary of the elastic incre-
elastic modulus may affect the trend of the hardness distribution. Mean- ment and the plastic increment. So a stress supplement factor 𝜂 is defined
while, it can be considered that the uneven strain field in the vertical to help solving this problem. Meanwhile, due to the errors in the plastic
direction is also probably determined by the uneven hardness distribu- incremental process and the effect of rounding errors, the updated stress
tion and causes gradient changes across the vertical line. Therefore, the often deviates from the yield surface and corrects by a factor 𝛽.
stiffness, strength, and hardness can also be used to confirm each other Step 0: Stress initialization, t = 0, 𝜎𝑖𝑗𝑡 = 0;
using their distributions. Step 1: Estimation of the stress at step t + 1:
All the results in this section indicate that the thermal effect of laser
𝑑 𝜎̂ 𝑖𝑗𝑡 = 𝐸𝑖𝑗𝑘𝑙 𝑑𝜀𝑡𝑘𝑙
welding will reduce the performance of local materials. The internal
𝜎̂ 𝑖𝑗𝑡+1 = 𝜎𝑖𝑗𝑡 + 𝑑 𝜎̂ 𝑖𝑗𝑡
and external factors affecting the fracture toughness of high strength
aluminum alloys were reviewed by Pan and Li [40] from the perspective Step 2:
of composition and grain structure. The grain size greatly affects the Case 2.1 When𝑓 (𝜎𝑖𝑗𝑡 , 𝑋𝑖𝑗𝑡 ) ≤ 0 and 𝑓 (𝜎̂ 𝑖𝑗𝑡+1 , 𝑋𝑖𝑗𝑡 ) ≤ 0, the stress is in the
fracture toughness of the aluminum, and the coarser grain size usually elastic loading or unloading phase, and thus,
results in lower fracture toughness. In addition, the size of the grain
𝜎𝑖𝑗𝑡+1 = 𝜎̂ 𝑖𝑗𝑡+1
can be adjusted by the power, speed, and time of the laser welding,
and thus, larger recrystallized grains can be avoided by some optimal Turn to Step 1;
welding parameters. Case 2.2 When𝑓 (𝜎𝑖𝑗𝑡 , 𝑋𝑖𝑗𝑡 ) = 0 and 𝑓 (𝜎̂ 𝑖𝑗𝑡+1 , 𝑋𝑖𝑗𝑡 ) > 0, after the plastic
multiplier is calculated according to Eq. (11), the stress tensor and kine-
5. Conclusions matic stress tensor are updated using
𝑑 𝜎𝑖𝑗𝑡 = 𝐸𝑖𝑗𝑘𝑙 (𝑑 𝜀𝑡𝑘𝑙 − 𝜆𝑠̇ 𝑡 ), 𝜎̃ 𝑡+1 = 𝜎 𝑡 + 𝑑 𝜎 𝑡
𝑖𝑗 𝑖𝑗 𝑖𝑗 𝑖𝑗
In this paper, the displacement and strain fields near the welded 𝑡 ̇ − 𝜆𝐶
𝑡 ̇ 𝑋 𝑡 ), 𝑋̃ 𝑡+1 = 𝑋 𝑡 + 𝑑 𝑋 𝑡
𝑑 𝑋𝑖𝑗 = 𝛾(𝜆𝑠 𝑖𝑗 𝑖𝑗 𝑖𝑗 𝑖𝑗 𝑖𝑗
joints of an aluminum-alloy laser-welding tensile specimen are mea- 𝜎0
𝛽= 𝑡+1 ̃ 𝑡+1
sured suing the 3D-DIC system. Four AOI regions near the welded joint 𝑓 (𝜎̃ 𝑖𝑗 ,𝑋𝑖𝑗 )
are identified by microscopic observation and the gradient changes in 𝜎𝑖𝑗𝑡+1 = 𝛽 𝜎̃ 𝑖𝑗𝑡+1 , 𝑋𝑖𝑗𝑡+1 = 𝛽 𝑋̃ 𝑖𝑗𝑡+1
the strain (fusion zone, heat-affected zone, transition base zone, and
where 𝛽 is a correction factor used to ensure that the yield function
base zone). Considering the metal properties near the welded joint, a
holds.
nonlinear constitutive stress incremental process based on the kinematic
Turn to Step 1;
hardening is proposed and applied to the nonlinear VFM. In order to
Case 2.3 When𝑓 (𝜎𝑖𝑗𝑡 , 𝑋𝑖𝑗𝑡 ) < 0and𝑓 (𝜎̂ 𝑖𝑗𝑡+1 , 𝑋𝑖𝑗𝑡 ) > 0, there are an elastic
obtain an accurate distribution of the material properties, the various
loading or unloading state and a plastic strengthening state in the inter-
zones were subdivided. Under the proposed nonlinear VFM, the elasto-
val [t, t + 1], thus it is necessary to distinguish them. A stress supplement
plastic constitutive parameters of each zone were obtained by using the
factor 𝜂 is introduced here to estimate the stress boundary between elas-
measured field data and the stress–strain relationship of each zone was
tic state and plastic state.
established. The inversion results show that the heat-affected zone has
𝜎0
the lowest elastic modulus, which is consistent with the hardness test 𝜂= 𝑡+1 ,𝑋 𝑡 )
𝑓 (𝜎̂ 𝑖𝑗 𝑖𝑗
results. The localized Poisson’s ratio of the laser-welded part decreases ⌢𝑡
with the increase in the weld distance. The fusion zone, heat-affected 𝜎 𝑖𝑗 = 𝜂 𝜎̂ 𝑖𝑗𝑡+1 + (1 − 𝜂)𝑋𝑖𝑗𝑡
⌢𝑒 ⌢𝑡
zone, and a part of the transition zone have a lower yield strength than 𝑡 )
𝑑 𝜀 𝑖𝑗 = 𝑆𝑖𝑗𝑘𝑙 (𝜎 𝑘𝑙 − 𝜎𝑘𝑙
the base zone, and the heat-affected zone is the most vulnerable as it
After the plastic multiplier is calculated according to Eq. (11), the
has the lowest yield strength and the highest plastic strengthening rate.
stress tensor and kinematic stress tensor are updated using
⌢𝑡 ⌢𝑒 ⌢𝑡 ⌢𝑡 ⌢𝑡
Acknowledgments 𝑑 𝜎 𝑖𝑗 = 𝐸𝑖𝑗𝑘𝑙 (𝑑 𝜀𝑡𝑘𝑙 − 𝑑 𝜀 𝑘𝑙 − 𝜆̇ 𝑠 𝑘𝑙 ), 𝜎̃ 𝑖𝑗𝑡+1 = 𝜎 𝑖𝑗 + 𝑑 𝜎 𝑖𝑗
⌢𝑡
𝑑 𝑋̃ 𝑖𝑗𝑡 = 𝛾(𝜆̇ 𝑠 𝑘𝑙 − 𝜆𝐶
̇ 𝑋 𝑡 ),
𝑖𝑗 𝑋̃ 𝑖𝑗𝑡+1 = 𝑋𝑖𝑗𝑡 + 𝑑 𝑋𝑖𝑗𝑡
The authors thank the National Natural Science Foundation of China 𝜎0
𝛽= 𝑡+1 ̃ 𝑡+1
(Nos. 11772081, 11572070, 11472070), and the Fundamental Research 𝑓 (𝜎̃ 𝑖𝑗 ,𝑋𝑖𝑗 )
Funds for the Central Universities of China(DUT18ZD209). 𝜎𝑖𝑗𝑡+1 = 𝛽 𝜎̃ 𝑖𝑗𝑡+1 , 𝑋𝑖𝑗𝑡+1 = 𝛽 𝑋̃ 𝑖𝑗𝑡+1
Turn to Step 1.
Supplementary materials
References
Supplementary material associated with this article can be found, in
the online version, at doi:10.1016/j.optlaseng.2018.05.017. [1] Wang Z, Hu FY, Cui AY, Liu HD. Laser welding technology research and development
trend (in Chinese). New Technol New Process 2016;3:42–4.
[2] Moraitis GA, Labeas GN. Residual stress and distortion calculation of laser beam
Appendix. The estimation correction scheme of stress welding for aluminum lap joints. J Mater Process Technol 2008;198(1):260–9.
incremental updating for kinematic hardening [3] Anawa EM, Olabi AG. Control of welding residual stress for dissimilar laser welded
materials. J Mater Process Technol 2008;204(1–3):22–33.
[4] Deng D, Murakawa H. Numerical simulation of temperature field and residual stress
In this section, the design of the nonlinear stress increment process in multi-pass welds in stainless steel pipe and comparison with experimental mea-
of the material is presented in order to extract the plastic parameters surements. Comput Mater Sci 2006;37(3):269–77.
of the kinematic hardening, which is applied in nonlinear VFM [18] to [5] Lockwood WD, Reynolds AP. Simulation of the global response of a friction stir weld
using local constitutive behavior. Mater Sci Eng A 2003;339(1):35–42.
invert the nonlinear constitutive of each zone of the welding part. Based
[6] Song GR, He CF, Li XY, Liu ZH, Wu B. Determining the change of the HAZ mechanical
on the kinematic hardening theory, the stress increment update proce- properties of TA15 titanium alloy by electron beam welding with ultrasonic method.
dure is described below. Here, a pre-estimated stress tensor is given by In: Proceedings of the second international conference on heterogeneous materials
𝜎̂ 𝑖𝑗𝑡+1 . If the estimated stress state stays inside the yield surface in the mechanics; 2008. p. 1362–5.
[7] Ambriz RR, Chicot D, Benseddiq N, Mesmacque G, Torre SDDL. Local mechanical
next step, it is believed that the estimated one is the real one. Else if properties of the 6061-T6 aluminium weld using micro-traction and instrumented
the estimated stress state goes beyond the yield surface in the next step, indentation. Eur J Mech – A/Solids 2011;30(3):307–15.
there will be two situations corresponding to the original situation. If [8] Yoneyama S, Arikawa S, Kusayanagi S, Hazumi K. Evaluating J-integral from dis-
placement fields measured by digital image correlation. Strain 2014;50:147–60.
the original stress state was on the yield surface, only the plastic stress [9] Reynolds AP, Duvall F. Digital image correlation for determination of weld and base
updating is concerned. If the original stress state located inside the yield metal constitutive behavior. Weld Res 1999;60:476–81.

130
R. Bai et al.
[10] Texier D, Zedan Y, Amoros T, Feulvarch E, Stinville JC, Bocher P. Near-surface me-
chanical heterogeneities in a dissimilar aluminum alloys friction stir welded joint.
Mater Des 2016;108:217–29.
[11] Acar M, Gungor S, Ganguly S, Bouched P, Fitzpatrick ME. Variation of mechanical
properties in a multi-pass weld measured using digital image correlation. Soc Exp
Mech Inc 2009.
[12] Yan J, Sutton MA, Reynolds AP, Adeeb S, Horsley D. Characterization of heteroge-
neous response of pipeline steel weld using digital image correlation. 2006.
[13] Dong J, Liu ZW, Gao JX. Multi-parameter inversion and thermo-mechanical defor-
mation decoupling using I-DIC. Exp Mech 2017;57(1):31–9.
[14] Sutton MA, Yan JH, Avril S, Pierron F, Adeeb S. Identification of heterogeneous con-
stitutive parameters in a welded specimen: uniform stress and virtual fields methods
for material property estimation. Exp Mech 2008;48(4):451–64.
[15] Saranath KM, Ramji M. Local zone wise elastic and plastic properties of electron
beam welded Ti-6Al-4V alloy using digital image correlation technique: a com-
parative study between uniform stress and virtual fields method. Opt Lasers Eng
2015;68:222–34.
[16] Banan MR, Hjelmstad KD. Parameter estimation of structures from static response.
II: Numerical Simulation Studies. J Struct Eng 2015;120(11):3259–83.
[17] Perry MJ, Koh CG, Choo YS. Modified genetic algorithm strategy for structural iden-
tification. Comput Struct 2006;84(8):529–40.
[18] Liu ZW, Gao JX. Deformation-pattern-based digital speckle correlation for coefficient
of thermal expansion evaluation of film. Opt Express 2011;19(18):17469–79.
[19] Dong J, Liu ZW, Gao JX. Multi-parameter inversion and thermo-mechanical defor-
mation decoupling using I-DIC. Exp Mech 2017;57(1):31–9.
[20] Pierron F, Grédiac M. The virtual fields method. Extracting constitutive mechanical
parameters from full-field deformation measurements. New York: Springer; 2012.
[21] Avril S, Grédiac M, Pierron F. Sensitivity of the virtual fields method to noisy data.
Comput Mech 2004;34(6):439–52.
[22] Grédiac M, Toussaint E, Pierron F. Special virtual fields for the direct determination
of material parameters with the virtual fields method. 1–Principle and definition.
Int J Solids Struct 2002;39(10):2691–705.
[23] Avril S, Pierron F, Pannier Y, Rotinat R. Stress reconstruction and constitutive pa-
rameter identification in plane-stress elasto-plastic problems using surface measure-
ments of deformation fields. Exp Mech 2008;48(4):403–19.
[24] Notta-Cuvier D, Langrand B, Lauro F, Markiewicz E. An innovative procedure for
characterizing a coupled elastoplastic damage model of behavior using the virtual
fields method. Int J Solids Struct 2015(69-70):415–27.
131
Optics and Lasers in Engineering 110 (2018) 122–131
[25] Valeri G, Koohbor B, Kidane A, Sutton MA. Determining the tensile response of ma-
terials at high temperature using DIC and the virtual fields method. Opt Lasers Eng
2017;91:53–61.
[26] Pierron F, Vert G, Burguete R, Avril S, Rotinat R, Wisnom MR. Identification of the
orthotropic elastic stiffnesses of composites with the virtual fields method: sensitivity
study and experimental validation. Strain 2007;43(3):250–9.
[27] Pierron F. A novel method for measuring the through-thickness shear moduli of
anisotropic plates from surface deformation measurements. Compos Part A Appl Sci
Manuf 2009;40(12):1815–25.
[28] Cao Q, Xie H. Characterization for elastic constants of fused deposition modelling–
fabricated materials based on the virtual fields method and digital image correlation.
Acta Mech Sin 2017;33(6):1–9.
[29] Louëdec GL, Pierron F, Sutton MA, Reynolds AP. Identification of the local elas-
to-plastic behavior of FSW welds using the virtual fields method. Exp Mech
2013;53(5):849–59.
[30] Saranath KM, Ramji M. Local zone wise elastic and plastic properties of electron
beam welded Ti–6Al–4V alloy using digital image correlation technique: a com-
parative study between uniform stress and virtual fields method. Opt Lasers Eng
2015;68:222–34.
[31] Chu Q, Bai R, Jian H, et al. Microstructure, texture and mechanical properties of
6061 aluminum laser beam welded joints. Mater Charact 2018;137:269–76.
[32] Sutton MA, Orteu JJ, Schreier HW. Digital image correlation for shape, motion and
deformation measurements. New York, USA: Springer; 2009.
[33] http://www.pmlab.com.cn/.
[34] Lurie AI, Belyaev A. Theory of elasticity. Berlin Heidelberg,: Springer; 2005.
[35] Chaboche JL. Time-independent constitutive theories for cyclic plasticity. Int J Plast
1986;2(2):149–88.
[36] Grédiac M. Principe des travaux virtuels et identification. 1989, 309-ii(1): 1–5.
[37] Nelder JA. Mead RA simplex method for function minimization. Comput J
1965;7(4):308–13.
[38] Lin SW, Ying KC, Chen SC, Lee ZJ. Particle swarm optimization for parameter de-
termination and feature selection of support vector machines. Expert Syst Appl
2008;35(4):1817–24.
[39] Bai RX, Wei YP, Lei ZK, Jiang H, Tao W, Yan C, Li XL. Local zone-wise elastic-plastic
constitutive parameters of laser-welded aluminium alloy 6061 using digital image
correlation. Opt Lasers Eng 2018;101:28–34.
[40] Pan ZJ, Li WX. Research status and prospect of fracture toughness of high strength
aluminum alloy. Chin J Mater Res 2002;16:14–17.