Beruflich Dokumente
Kultur Dokumente
, Zhao Rui
School of Materials Science and Engineering, Hebei University of Technology, No. 8, Duangrongdao Road, Hongqiao District, Tianjin 300130, PR China
a r t i c l e i n f o
Article history:
Received 25 January 2013
Received in revised form9 July 2013
Accepted 11 July 2013
Available online 23 July 2013
Keywords:
Numerical simulation
Welding temperature eld
Heat source model
Pulsed current gas tungsten arc welding
a b s t r a c t
A time-dependent welding heat source model, which is dened as the dynamic model, was established
according to the characteristic of PCGTAW. The parabolic model was proposed to describe the heat ux
distribution at the background times. The recommended Gaussian model was used at the peak times due
to the bell-shaped temperature contour. The dynamic welding heat source was composed of these two
models with a function of time.
To assess the validity of the dynamic model, an experiment was conducted in which the pulsed current
gas tungsten arc deposits on the plate. From the comparison of the experimental and the simulated
values, it can be concluded that the dynamic heat source model, which uses the parabolic model at the
background time, is more realistic and accurate under the same welding conditions.
2013 Elsevier B.V. All rights reserved.
1. Introduction
With the development of the computer and numerical anal-
ysis technologies, the FEM has become a powerful and reliable
technique for prediction in the welding processing industry. The
temperature eld contains sufcient information about the quality
and properties of the welded joint, and determines the distortion,
residual stresses, and reduced strength of a structure in and near
the welded joint. The temperature eld is also the foundation of
the metallurgical analysis and phase change analysis. To obtain an
accurate welding temperature eld, Goldak et al. (1984) reported
that the importance of a good welding heat source model has been
emphasized by many investigators.
Many welding heat source models have been developed up to
now, and the Gaussian model and the double ellipsoidal model are
the most popular models among them. Some good welding heat
source models can accurately predict the temperature eld. How-
ever, most of these models were developed on the assumption that
the heat sources are static and not varied with time in the welding
processes. These models are no longer realistic for some dynamic
welding processes, such as the pulsed current gas tungsten arc
welding (PCGTAW). The objective of this paper is to develop a more
realistic and accurate welding heat source model for PCGTAW.
PCGTAW was developed in 1950s and is widely used in the
manufacturing industry today. In PCGTAW, the welding current is
1
x
2
R
2
b
, R
b
x R
b
(1)
where q(0) is the maximumvalue of heat ux and R
b
is the radius
of the power density.
Z. Tong et al. / Journal of Materials Processing Technology 213 (2013) 23292338 2331
Fig. 2. Infra-red camera images at the background and peak times for both rst and last periods.
Substituting q(0) =43W/mm
2
and R2.8mm which is corre-
sponded with the third curve in Fig. 3a into Eq. (1):
q(x, ) = 43
1
x
2
2.8
2
, R
b
x R
b
(2)
The function image of Eq. (2) is shown in Fig. 3b, which approxi-
mates tothethirdcurveinFig. 3athat represents theradial heat ux
distribution at the background time, which can be clearly observed
in Fig. 3c which combined Fig. 3a with Fig. 3b in the same scale. So
it can be considered that the radial heat ux distribution at back-
ground time is approximate to parabolic shape, and the welding
heat source is a spinning parabolic shape distribution as shown in
Fig. 4.
The spinning parabolic shape model of welding heat source with
the center at (0, 0, 0) to coordinate axes x, y, can be written as:
q(x, y, ) = q(0)
1
x
2
+y
2
R
2
b
(3)
where q(x, y, ) is the power density (W/m
2
).
For r =
x
2
+y
2
which is the radial distance fromthe center of
the heat source, then Eq. (3) can be written as:
q(r) = q(0)
1
r
2
R
2
b
, r R
b
(4)
Conservation of energy requires that:
Q = qUl =
!
q(r)r dr d0 =
R
b
0
q(0)
1
r
2
R
2
b
r dr
2
0
d0 (5)
and produces the following:
Q = qUl = q(0)
R
2
b
2
(6)
q(0) =
2qUl
R
2
b
(7)
Substituting q(0) fromEq. (7) into Eq. (4) gives:
q(r) =
2qUl
R
2
b
1
r
2
R
2
b
, r R
b
(8)
So the dynamic welding heat source model of PCGTAWin one
pulse cycle can be written as:
q(r) =
3qpUlp
R
2
p
exp
3
r
2
R
2
p
1
r
2
R
2
b
, t (tp, t1 ] and r R
b
(at background times)
(9)
where q(r) is the power density (W/m
2
), q
p
the heat source ef-
ciency at the peak time, q
b
the heat source efciency at the
background time, U the arc voltage (V), I
p
the peak current (A), I
b
the background current (A), r =(x
2
+y
2
)
1/2
which is the radial dis-
tance from the center of the heat source (m), R
b
the radius of the
heat source at the background time (m), R
p
the radius of the heat
source at the peak time (m), t
T
=1 pulse cycle time =1/f (s), f the
pulse frequency, t
p
the peak time (s), t
b
the background time (s)
and t
p
+t
b
=t
T
.
3. Evaluation of the dynamic model of welding heat source
in PCGTAW
One experiment was conducted in which the pulsed current
gas tungsten arc was deposited on the plate. The thermocouple
was used to measure the temperature eld at the given points,
then the experimental values were compared with the simulated
values to assess the validity of the dynamic welding heat source
model.
Due to the lack of data on material properties, material mod-
eling has always been a critical issue in the welding simulation.
2332 Z. Tong et al. / Journal of Materials Processing Technology 213 (2013) 23292338
Table 2
The chemical composition of AA7075.
Elements Zn Mg Cu Cr Mn Fe Si Ti Al Impurities
wt.% 5.16.1 2.12.9 1.22.0 0.180.28 0.30 0.50 0.40 0.20 Bal. 0.15
Sattari-Far and Javadi (2008) reported that some simplications
and approximations are usually introduced to deal with this prob-
lem, which are necessary because of the scarcity of material
data and numerical problems when trying to model the actual
high-temperature behaviors of the material. Here we select the
Aluminum Alloy 7075 as the base metal; the chemical composi-
tion is shown in Table 2. The thermal properties of AA7075 shown
Fig. 3. The establishment of the parabolic distribution (a is referred to Traidia and
Roger, 2011).
in Fig. 5 were reported by Guo et al. (2006) which are temperature-
dependent, the emissivity is assumed to be 0.6, and the fusion
temperature range is 477638
C.
3.1. Experimental procedure
3.1.1. Experiment preparation
The plate of AluminumAlloy 7075was cut to the requiredsize of
80mm80mm8mm. To measure the temperature in the weld-
ing process, the K type NiCrNiSi thermocouple was used. The
positions of the thermocouples in the plate were shown in Fig. 6.
The thermocouples were glued to a depth of 4mm, through the
blindholes whichwere drilledfromthe bottomof the plate; the hot
end diameter of the thermocouple was 1.5mm, the cold end was
connected to a multichannel temperature measuring instrument
to acquire the thermal cycle, and the same method was introduced
by Karunakaran and Balasubramanian (2011).
3.1.2. Welding
Bead-on-plate welds were made using the PCGTAWon the sur-
face of the plate along with the center line. The welding parameters
are shown in Table 3.
3.2. FEMcalculation
3.2.1. Finite element model
Only half of the plate was selected to analysis for its symmetry.
To reduce the calculation time, the zone near the welding bead has
been modeled with a ner mesh, while the zone further away from
the welding bead has been modeled with a coarser one. Solid70
and Surf152 were used to mesh the model; the surface has been
Fig. 4. Heat source conguration for the spinning parabolic shape.
Z. Tong et al. / Journal of Materials Processing Technology 213 (2013) 23292338 2333
Fig. 5. Thermal physical properties of AA7075: (a) specic heat and density and (b)
conductivity.
Fig. 6. Schematic diagramof welded plate used in the experiment.
Table 3
Welding parameter.
Process parameter Actual Simulated
Welding current
Peak current 180A 180A
Background current 60A 60A
Arc voltage 14V 14V
Welding speed 1.962.03mm/s 2mm/s
Pulse frequency 1Hz 1Hz
% Pulse on time 50% 50%
Electrode W2%Th
Electrode diameter 3.2mm
Arc length 2mm
Torch angle 60
Shielding gas Argon 99.9%
Flowrate 15L/min
coated with Surf152 to represent the convective heat exchange.
The FEMmodel is shown in Fig. 7.
3.2.2. Welding heat source
In this research, the APDL programming languages of ANSYS
were applied to realize the moving load of the heat source. A
local coordinate systemwas established, and the center of the heat
source coincided with the original point of the local coordinate,
then the heat source moved gradually under the control of the loop
command in APDL.
To evaluate the validity of the dynamic heat source model, two
simulation tests were implemented under the same welding con-
ditions, which are described in Table 4. The parameters in the
dynamic welding heat source model are not easy to decide, so a
further study is needed.
3.2.3. Initial condition and boundary conditions
The ambient temperature is 28
C.
Heat loss due to radiation is modeled using StefanBoltzmanns
law:
q
r
= o [(1
s
+273)
4
(1
0
+273)
4
]
where is emissivity which is 0.6 and o =5.6710
8
W/m
2
C
4
is
dened as the StefanBoltzmann constant.
3.2.4. Latent heat of phase transition
During the welding process, melting and solidifying will occur
in the welding pool, it will absorb or release latent heat in the
phase transition, which is dened as latent heat of phase tran-
sition. Lei et al. (2006) use the enthalpy method to deal with the
Fig. 7. FEMmodel.
2334 Z. Tong et al. / Journal of Materials Processing Technology 213 (2013) 23292338
Table 4
List of the simulation test.
Simulation test 1 Simulation test 2
Heat source model Dynamic Model 1 Dynamic Model 2
Model description Use Gaussian model at peak times; use parabolic model at
background times
Use Gaussian model both at peak times and background times, but
different values of parameters were used, respectively
Parameters in model
Ip =180A, I
b
=60A, U=14V, f =1Hz Ip =180A, I
b
=60A, U=14V, f =1Hz,
Pulse on time =50%, Rp 5.0mm, R
b
2.8mm, qp 0.68, q
b
0.62. Pulse on time =50%, Rp 5.0mm, R
b
2.8mm, qp 0.68, q
b
0.62.
Notes: The parameters in heat source models are difcult to decide. To simplify the problems, the same parameters in Traidia and Roger (2011) were used for test 1 and test
2 under the same welding condition.
latent heat, and dene the materials enthalpy which varies with
the temperature:
^H(1) =
1
0
(1)c(1) d1
where (T) is the density of the material varying with temperature
(kg/m
3
) and c(T) is the specic heat of the material varying with
temperature (J/(kgK)).
Murugan et al. (2000) reported that the release or absorption of
latent heat can also be considered in the numerical analysis by an
articial increase in the value of the specic heat over the melting
temperature range.
3.2.5. Others
In the meshed nite element model, the number of the Solid70
element is 848,000, the number of the Surf152 element is 46,640,
and the number of nodes is 887,814 in total.
The heat source denedina local coordinate systemmoves with
time, the former loadstepis deletedwhenthe heat source moves to
the next step. Considering both the calculation time and the com-
puters capacity, the minimum size of element is 0.2mm, and the
cooling time is xed to 20s.
Fig. 8. Top viewof temperature distribution: (a) at 20.3s (peak time) and (b) at 20.8s (background time) computed by the Dynamic Model 1.
Z. Tong et al. / Journal of Materials Processing Technology 213 (2013) 23292338 2335
Fig. 9. Top viewof temperature distribution: (a) at 20.3s (peak time) and (b) at 20.8s (background time) computed by the Dynamic Model 2.
Fig. 10. Longitudinal cross-section of temperature distribution: (a) at 20.3s (peak time) and (b) at 20.8s (background time) computed by the Dynamic Model 1.
2336 Z. Tong et al. / Journal of Materials Processing Technology 213 (2013) 23292338
Fig. 11. Longitudinal cross-section of temperature distribution: (a) at 20.3s (peak time) and (b) at 20.8s (background time) computed by the Dynamic Model 2.
4. Results and discussion
4.1. Temperature eld
4.1.1. Top view of temperature distribution
Figs. 8 and 9 which are in the same scale, showthe temperature
eld computed by the Dynamic Model 1 and the Dynamic Model
2, respectively, and including the time 20.3s (peak time) and 20.8s
(background time) for each. To illustrate the difference of the tem-
perature eld between the peak time and the background time in
the welding process, the same area region near the weld pool was
magnied in the same scale.
Comparing the two temperature elds in Fig. 8a and b, it can
be seen that the high temperature region at 20.3s is larger than
that at 20.8s. Due to the cyclic variation of the heat input, there is
a thermal uctuation in the temperature eld, which corresponds
to the real dynamic welding process. FromFig. 9a and b, the same
conclusion above can be obtained.
Table 5
Peak temperature comparison of the experimental and simulated results.
Measuring
point
Methods
a
Peak
temperature (
C)
Difference
b
(%)
Point A
c
Experimental 402.5
FEM(Dynamic Model 1) 397.6 1.2
FEM(Dynamic Model 2) 393.9 2.1
Point B
c
Experimental 285.8
FEM(Dynamic Model 1) 276.7 3.2
FEM(Dynamic Model 2) 272.5 4.7
Point C
c
Experimental 327.2
FEM(Dynamic Model 1) 317.2 3.1
FEM(Dynamic Model 2) 312.3 4.6
a
Experimental: use PCGTAWweldingparameter is showninTable 3; base metal
AA7075, chemical compositionis showninTable 2. The descriptionof the Dynamic
Model 1 and Dynamic Model 2 are listed in Table 4.
b .
Difference (%) =(Calculated value Experimental value)/Experimental value.
c .
The position of the measuring points is depicted in Fig. 6.
In Fig. 8a and b, the maximum temperatures are 892
C and
779
C and
849