An attempt to enhance NUMERICAL MODELS OF ANGULAR DISTORTION by considering the physics of the welding arc

An attempt to enhance NUMERICAL MODELS

OF ANGULAR DISTORTION by considering
the physics of the welding arc
S. Okano, M. Mochizuki, K. Yamamoto and M. Tanaka
ABSTRACT

Recently, it has become necessary to control or reduce weld distortion, which has a negative influence on
structural integrity, without loss of manufacturing efficiency. Some studies on the fundamental characteristics
of weld distortion and its control or reduction have been conducted. However, the results cannot be applied
to all welding processes because such processes are becoming more diversified. For this reason, to under-
stand the fundamental characteristics of weld distortion based on the welding conditions, the heat transport
phenomenon in arc physics must be investigated. In this study, an experiment and numerical simulations are
conducted to investigate the link between arc physics and weld distortion. As a result, the relation between
weld distortion and the heat transport phenomenon is further clarified.

IIW-Thesaurus keywords: Arc physics; Distortion; Energy input; Simulating.

1 Introduction temperature distribution during welding. To clarify the

It is well-known that weld distortion, which has a negative
influence on material properties, on the ability to fabricate
structures, and structural integrity, should be controlled.
There are many methods for controlling or reducing weld
distortion, but most of them involve costly processing in
addition to the welding process. In-process control of
weld distortion becomes preferable to post-weld control
when manufacturing efficiency is considered.
Traditionally, weld distortion is controlled by weld heat
input. Weld heat input is controlled by the following: weld-
ing current and arc voltage from the welding power source,
welding speed from a welding robot; and arc efficiency of
the welding method. However, it is also known that weld
distortion is not totally controlled by weld heat input only.
Therefore, it is important to clarify the factors affecting weld
distortion.
In this study, two types of investigations are performed
to consider the relationship between weld heat input and
weld distortion. The first one is an experiment and the
second one is numerical analysis. In the numerical analy-
sis, an approach to linking arc physics and weld distortion
is presented.
dominant factor affecting weld distortion, it is important
to understand the actual generation behaviour of plastic
strain in relation to the inhomogeneous temperature dis-
tribution. An example is the model of a bar clamped at
both ends. In this case, the generation behaviour of plastic 93
strain in the welding direction can be explained almost
completely. However, the generation behaviour of plastic
strain in the transverse direction is more complex, and it
is difficult to ascertain the dominant factor affecting weld
distortion, such as transverse shrinkage and angular dis-
tortion in the transverse direction. For this reason, one
approach has been to clarify the dominant factor affect-
ing weld distortion in the transverse direction in relation to
the similarity rule of temperature distribution in the plate
thickness section [1].
By considering a non-moving (instantaneous) line heat
source, thermal conduction theory can be used to express
the temperature distribution in a plate thickness section
during welding [2], as stated below:
T = 1 /(2√πkt)2 · Qnet / cρ · exp{−(y2 + z2)/ 4 kt} (1)
that is, k = λ / cρ,
where
ρ is density,

2 Knowledge about dominant c is the specific heat,

λ is the heat conductivity,
factor affecting weld distortion
t is time and
2.1 Similarity rule of temperature
h is the plate thickness,
distribution in a plate thickness section
Qnet is the net heat input,
The cause of weld residual distortion is inherent strain,
which is plastic strain generated due to an inhomogeneous T is temperature.

Doc. IIW-2092, recommended for publication by Study Group SG-212 “The Physics of Welding.”

06 2011 Vol. 55 WELDING IN THE WORLD

N° 05 Peer-reviewed Section
 An attempt to enhance NUMERICAL MODELS OF ANGULAR DISTORTION by considering the physics of the welding arc

In Equation (1), by converting each variable number into

plate thickness h, the temperature distribution is expressed
as follows:

T = (Qnet / h2) · (2 / cρ) · exp{−(X* 2 + Y* 2)/ τ* 2} / πτ* 2 (2)

that is, Qnet = η/V / υ, Y* = y / h, Z* = z / h, τ* = 2√kt / h,

where

η is the arc efficiency.

According to Equation (2), the dominant factor affecting

temperature distribution in the plate thickness section is
Qnet / h2, which is called the heat input parameter. At this
time, the welding heat input Qnet means the heat inten-
sity per welding length [J/mm], not per time [J/s]. The Figure 2 – Evaluation of angular distortion [1]
heat input parameter represents the temperature rise at
one location at one time, and the generation behaviour of
plastic strain at one location at one time. The deformed
state in welded joints is shown in Figure 1. In this fig-
ure, transverse distortion conforms to the similarity rule of
3 Experimental examination
temperature distribution and distortion, which means that for relation between weld heat
the transverse shrinkage per plate thickness is depen-
dent on the heat input parameter, and angular distortion is
input and weld distortion
dependent on the heat input parameter [1].
3.1 Experimental procedure
2.2 Characteristics of angular distortion The steel used in the experiment is SM490. The chemi-
94 cal composition of this steel is shown in Table 1, and the
According to the similarity rule of temperature distribu- configuration of the welded joint used in the experiment
tion in the plate thickness section, angular distortion is shown in Figure 3. The dimensions of the plate are the
is dependent on the heat input parameter. As shown in following: length 200 mm, width 500 mm, and thickness
Figure 2, when the welding method is the same, angu- 6 or 12 mm. The weld length is 150 mm, which leaves
lar distortion produced in any welding conditions can be unwelded sections of 25 mm at both ends of the plate.
evaluated by one curve which depends on the heat input The plate is placed on a sill plate due to the heat transfer
parameter. When the heat input parameter is small, rigid- from the bottom of the welded plate to the air.
ity due to plate thickness prevents the welded plate from
being distorted. In contrast, when the heat input param- The locations of the thermocouple and displacement
eter is large, the temperature gradient through the thick- gauge are shown in Figure 4. The temperature histories
ness direction becomes uniform. Then, angular distortion are measured by five thermocouples, and the deflection is
becomes smaller. As a result, the maximum angular dis-
tortion occurs in the intermediate condition of the heat Table 1 – Chemical compositions [mass %]
input parameter.
Plate thickness
C Si Mn P S
[mm]
6 0.15 0.25 1.42 0.020 0.004
12 0.16 0.28 1.45 0.015 0.003

Figure 1 – Schematic illustration of similarity rule

of temperature distribution and distortion Figure 3 – Geometry of bead-on-plate welded joint
 An attempt to enhance NUMERICAL MODELS OF ANGULAR DISTORTION by considering the physics of the welding arc

Figure 4 – Measurement location of temperature history and vertical displacement

measured by six displacement gauges. The thermocou-
ples are set in holes of 3 mm in diameter, which were
already drilled in the steel. Deflection is calculated as the
sum of displacements on both sides, for example, [(D1)
+ (D2)].
To clarify the different effects of each factor, such as that
between welding current and welding speed, it is impor-
tant to consider the heat transport phenomenon during
welding by examining the link between arc physics and
weld distortion.

4 Investigation of heat input

3.2 Welding conditions
The experimental welding conditions are shown in Table 2.
The welding method is tungsten inert gas (TIG) welding.
The shielding gas is 100 % Ar, the flow rate is 10 l/min. To distribution by numerical analysis
investigate the effect of welding current, welding speed,
arc length and plate thickness individually, 11 welding of TIG arc plasma 95
conditions are examined. The base condition is a welding
current of 200 A, a welding speed of 10 cm/min and an 4.1 Simulation model
arc length of 3 mm. Two approaches for reduction of weld The process using this model is stationary TIG welding
heat input Qnet (= IV/v: J/mm) are implemented. One is using a tungsten cathode of 3.2 mm in diameter with a
reducing the welding current (I-changed), and the other 60° conical tip. Figure 6 shows the calculation domain,
is increasing the welding speed (v-changed). These con-
ditions are applied in both the 6 mm and 12 mm plate
thicknesses. In the case of a plate thickness of 6 mm, the
effect of arc length (L-changed) is investigated, as shown
in Table 2.

3.3 Result of experiment

The results of the experiment on the relationship between
weld heat input and angular distortion are shown in
Figure 5. The measurements of deflection at the centre
[(D3) + (D4)] in all experimental conditions are evaluated
by the heat input parameter Qnet/h2. As shown in this fig-
ure, it is obvious that the angular distortion is not always
represented by the heat input parameter. In particular, an
obvious difference of distortion exists due to the effect of
welding current and welding speed. This tendency is not
Figure 5 – Experimental result
influenced by the plate thickness.
Table 2 – Welding conditions for measurement of residual welding distortion
Base I-changed v-changed L-changed
Welding current, I [A] 200 160 100 200 200 200
Welding speed, v [cm/min] 10 10 10 14 20 10
Arc length, L [mm] 3 3 3 3 3 6
Thickness of base metal,
6, 12 6, 12 6, 12 6, 12 6, 12 6
h [mm]

06 2011 Vol. 55 WELDING IN THE WORLD

N° 05 Peer-reviewed Section
 An attempt to enhance NUMERICAL MODELS OF ANGULAR DISTORTION by considering the physics of the welding arc

jr = −σEr, jz = −σEz (8)

where
t is time,
H is enthalpy,
p is pressure,
vr and vz are the radial and axial velocities,
jr and jz are the radial and axial components of the current
density,
g is acceleration due to gravity,
c is the specific heat (under constant pressure),
λ is the thermal conductivity,
ρ is density,
μ is viscosity,
U is the radiative emission coefficient and
σ is the electrical conductivity.
Figure 6 – Schematic illustration of simulation domain Er and Ez are the radial and axial components of the elec-
tric field defined by
described in two-dimensional cylindrical coordinates with wV wV
rotational symmetry around the arc axis. The anode is Er  , Ez  (9)
wr wz
SM490, the same as that used in the experiment. Argon
shielding gas is supplied from outside the cathode on the where
upper boundary at a flow rate of 1.67 × 105 mm3s-1. Two-
96
dimensional calculations are conducted. V is the electric potential.

The governing equations used in the model are shown as The azimuthal magnetic field Bθ induced by the arc cur-
follows: rent is evaluated by Maxwell’s equation:
1w
The mass continuity equation is  rBT  P0 j z (10)
wU 1 w w r wr
  r Uv r    Uv z  0 (3)
wt r wr wz where
The radial momentum conservation equation is μ0 is the permeability of free space.
wUv r 1 w w wp
wt

r wr
 r Uv r2  
wz
 Uv zv r    j z BT 
wr It is necessary to consider the effects of energy transfer at
(4) the surfaces of the electrodes. The additional energy fluxes
1w § w vr · w § w vr w vz · v
¨ 2r P  P P  2P 2r at the cathode and anode are described as follows:
r wr © w r ¸¹ w z ¨© w z w r ¸¹ r
Cathode: FK = −εαT4 −⏐je⏐φK + ⏐ji⏐Vi (11)
The axial momentum conservation equation is
wUv z 1 w w wp Anode: FA = −εαT4 −⏐j⏐φA
 r Uv r v z    Uv z2    jr BT 
(12)

wt r wr wz wz
(5) where
w § w v z · 1 w § w vr w vz ·
2P  rP  rP  Ug
w z ¨© w r ¸¹ r w r ¨© w z w r ¸¹ ε is the surface emissivity,

The energy conservation equation is α is the Stefan-Boltzmann constant,

wU H 1 w w φK is the work function of the tungsten cathode,
  r Uv r H    Uv z H 
wt r wr wz
(6) Vi is the ionization potential of the arc plasma gas,
1 w § rO w H · w § O w H ·
¨ ¸ ¨ ¸  jr Er  j z E z  U
r wr © c wr ¹ wz ©c wz ¹ je is the electron current density,

The current continuity equation is ji is the ion current density,

1w w φA is the work function of the anode, and

 rjr    j z  0 (7)
r wr wz T is temperature.
 An attempt to enhance NUMERICAL MODELS OF ANGULAR DISTORTION by considering the physics of the welding arc

For the cathode surface, FK needs to be included in

Equation (6) to account for thermionic cooling by the
emission of electrons, radiative cooling and ion heat-
ing. Similarly, for the anode surface, FA is required in
Equation (6) to account for radiative cooling and thermi-
onic heating. Furthermore, for the cathode surface, the
electron current and the ion current are considered sepa-
rately and defined based on the Richardson-Dushman
equation of thermionic emission, as follows:
§ eI ·
jR AT 2 exp ¨  K ¸ (13)
© kBT ¹
a) Effect of welding current
where

kB is Boltzmann’s constant and

A is Richardson’s constant, which is dependent on the

cathode material.

The ion current density ji is then assumed to be

ji = j − jR (14)

where the total current density j is

⏐j⏐ = ⏐je⏐ + ⏐ji⏐ (15)

b) Effect of welding speed
The convection in the weld pool is influenced by the shear
stress caused by the convective flow of the cathode jet,
the Marangoni force induced by the gradient in the surface 97
tension of the weld pool, the buoyancy due to gravity, and
the electromagnetic pinch force due to the arc current. If
only the driving forces of the weld pool convection at the
boundary between the weld pool and the arc plasma are
considered, the shear stress caused by the plasma flow is
already included in μ in Equation (4). Therefore, only the
Marangoni force is added, which is given by [3]:
w § wJ ·§ w T ·
MA ¨ ¸¨ ¸ (16)
w z © w T ¹© w r ¹ c) Effect of arc length

where Figure 7 – The result of numerical simulations

of TIG arc plasma
γ is the surface tension of the weld pool.

It is assumed in this study that the variation of surface
tension at the weld pool surface decreases linearly with
increasing temperature (= –0.46 mN/mK) [3].
The governing and auxiliary equations were solved itera-
tively by the SIMPLEC numerical procedure [4, 5]. The
other approximation and boundary conditions are given in
previous papers of the present co-authors [3]. In addition,
in the simulations, the effect of welding speed is consid-
ered by quasi-three-dimensional thermal conduction in
the anode material.
4.2 Heat input distribution from TIG arc
plasma to base metal
Heat input distribution from TIG arc plasma to base metal
is included in the numerical simulation. Figure 7 shows
the simulation results. The effects of welding current,
welding speed and arc length are shown in Figures 7 a),
7 b) and 7 c), respectively. As shown, the total heat input
per unit time (J/s) and its distribution, which means the
intensity of heat flux, is greatly influenced by welding
current, but not so influenced by welding speed. The arc
length greatly affects the distribution of heat input, but
not the total heat input. Total heat input increases slightly
as arc length increases. It is important to distinguish the
different characteristics of heat input from the TIG arc
plasma to the base metal in order to understand the dif-
ferent effects of welding current, welding speed and arc
length.
The next investigation is a weld distortion analysis using
the heat input distribution obtained from the above-men-
tioned numerical simulation of TIG welding arc plasma.

06 2011 Vol. 55 WELDING IN THE WORLD

N° 05 Peer-reviewed Section
 An attempt to enhance NUMERICAL MODELS OF ANGULAR DISTORTION by considering the physics of the welding arc

5 Weld distortion analysis based The material properties used in the analysis are the same
as those of commonly used steel [6]. Also, T is the tem-
perature of the base metal and w is the internal heat gen-
on arc plasma simulation eration. Traditionally, the distribution of the internal heat
generation in the weld distortion analysis is set as one
5.1 Simulation model (finite element chooses by each performer. In this analysis, internal heat
model) generation was already calculated by numerical simulation
of TIG arc plasma. Internal heat generation is expressed
Thermal elastic-plastic analysis of weld distortion is per-
by the Gaussian distribution, as shown by Equation (18):
formed using the finite element method (FEM). Figure 8
shows the three-dimensional finite element model and Q ­°  x  vt 2 ½° ­ y2 ½
w  x, y,t  exp ®  ¾ exp ®  2¾ (18)
the mesh divisions used in the analysis. In the analysis, a S R2 °¯ R 2 ¿° ¯ R ¿
half model is used because of symmetry of the welding
phenomenon. A moving heat source is considered in the where
analysis.
Q is the total heat input per time,
5.2 Thermal conduction analysis R is the radius of the Gaussian distribution,
The physical properties of the steel used for the thermal
conduction analysis are determined with consideration of v is the welding speed and
the temperature dependency. That is, the heat flux due to t is time.
the heat transfer from the plate surface is considered. In
the thermal conduction analysis, the governing equation is The total heat input Q and the radius of Gaussian distribu-
tion R can be determined by numerical simulation of TIG
wT w § wT · w § wT · w § wT ·
Uc ¨ Ox ¸ ¨ Oy ¸ ¨ Oz ¸ w arc plasma. The heat input distribution used in the weld
wt wx © wx ¹ wy© wy ¹ wz© wz ¹ distortion analysis is shown in Figure 9. The Gaussian
(17)
distribution is denoted by a full line and the result of the
where arc plasma simulation is plotted. Both distributions dem-
onstrate good agreement. Therefore, these distributions
98
ρ is density,
are used in the distortion analysis as the heat flux input
c is specific heat, on the surface.
λi is the i-component of thermal conductivity.
5.3 Thermal stress analysis
In the elastic-plastic analysis, the governing equations are
as follows. The equation of the strain-displacement rela-
tion is
{ε} = [A]{U} (19)
where

{ε} and {U} are the tensor of strain and the displacement,
and

[A] is the matrix that includes the differential operator.

a) Finite element model of a bead-on-plate
The constitutive equation is
{σ} = [De]{εe} = [De]({ε} − {εt}) (20)
where

{σ} is the tensor of stress, and

{εe} and {εt} are the tensors of elastic strain and thermal
strain, respectively.

[De] is the elastic stress–strain matrix.

The virtual work principle, shown in the following, can sub-

stitute for the equation of equilibrium:
b) Mesh division in weld
³ G ^H ` ^V ` dV  ³ G ^U` ^F ` dV  ³ G ^U` ^T ` dS
T T T
Figure 8 – Finite element model of a bead-on-plate 0
VV VV SV
and cross-section of mesh division in weld (21)
 An attempt to enhance NUMERICAL MODELS OF ANGULAR DISTORTION by considering the physics of the welding arc

a) I = 100 A, v = 10 cm/min and L = 3 mm d) I = 200 A, v = 14 cm/min and L = 3 mm

b) I = 160 A, v = 10 cm/min and L = 3 mm e) I = 200 A, v = 20 cm/min and L = 3 mm

99

c) I = 200 A, v = 10 cm/min and L = 3 mm f) I = 200 A, v = 10 cm/min and L = 6 mm

Figure 9 – Heat input modelling for weld distortion analysis

where ficient, are considered in the analysis. No restraint is used

– during the welding process.
{F } is the tensor of body force per unit volume, and
– 5.4 Result of weld distortion analysis
{T } is the tensor of surface force per unit area.
Vσ is the volume of the body and
Sσ is the area for which boundary conditions are given.
δ{ } is the amount of virtual change,
{ }T is the transposed tensor.
Equations (19), (20), and (21) are essential in the elastic-
plastic analysis.
Work hardening of the material and temperature depen-
dency of the mechanical properties, which include yield
stress, Young’s modulus and the thermal expansion coef-
The result of the weld distortion analysis is shown in
Figure 10. Deflection in each welding condition is evalu-
ated by the heat input parameter, as shown in Figure 5.
The tendency of deflection against the welding condition,
such as differences in the welding current and welding
speed, is precisely simulated. In addition, the amount of
deflection is also quasi-quantitatively simulated by the
analysis. Thus, it is expected that the prediction of distor-
tion by the weld distortion analysis can be improved by
heat source modelling based on the phenomenon of arc
physics.
In this study of the link between arc physics and weld
distortion, even though weld pool behaviour, such as

06 2011 Vol. 55 WELDING IN THE WORLD

N° 05 Peer-reviewed Section
 An attempt to enhance NUMERICAL MODELS OF ANGULAR DISTORTION by considering the physics of the welding arc
Figure 10 – Results of the weld distortion analysis
convective heat transport, is not exactly considered, the
analytical result of deflection by weld distortion quasi-
quantitatively corresponds to the experimental result.
6 Summary
In this study, two types of investigations of the relation-
ship between weld heat input and weld distortion are
100
performed. In the experimental investigation, it is clarified
that weld distortion is not always precisely evaluated by
the weld heat input and that each factor affecting the
weld heat input, such as welding current, arc voltage and
welding speed, should be considered individually. Based
on the results, weld distortion analysis is conducted with
consideration of a more detailed heat input distribution
from the arc plasma to the base metal, which is calcu-
lated by numerical simulation of TIG arc plasma. A quasi-
quantitatively precise analysis of weld distortion is per-
formed. For a more precise analysis, more details of the
heat transport phenomenon, such as weld pool behaviour,
and temperature dependency of the material properties
should be employed.
References
[1] Satoh K. and Terasaki T.: Effect of welding conditions
on welding deformations in welded structural materials,
Journal of the Japan Welding Society, 1976, vol. 45, no. 4,
pp. 302-308 (in Japanese).
[2] Rosenthal D.: Mathematical theory of heat distribu-
tion during welding and cutting, Welding Journal, 1941,
vol. 20, no. 5, pp. 220-234.
[3] Tanaka M. and Lowke J.J.: Prediction of weld pool pro-
files using plasma analysis, Journal of Physics D: Applied
Physics, 2007, vol. 40, no. 1, pp. R1-R23.
[4] Tashiro S., Tanaka M., Nakata K., Iwao T., Koshiishi F.,
Suzuki K. and K. Yamazaki: Plasma properties of helium
gas tungsten arc with metal vapor, Science and Technology
of Welding & Joining, 2007, vol. 12, no. 3, pp. 202-207.
[5] Tashiro S., Tanaka M., Nakatani M., Furubayashi M. and
Yamazaki Y.: Properties of mass and heat transfer for tube
cathode arcs, Quarterly Journal of the Japan Welding
Society, 2007, vol. 25, no. 1, pp. 3-9 (in Japanese)
[6] Okano S., Mochizuki M. and Toyoda M.: Angular distor-
tion reduction by in-process control welding using back
heating source, Materials Science Forum, 2007, vols. 580-
582, pp. 577-580.
About the authors
Mr. Shigetaka OKANO (okano@mapse.eng.osaka-u.ac.jp) and Prof. Dr. Masahito
MOCHIZUKI (mmochi@mapse.eng.osaka-u.ac.jp) are both with Department of
Materials and Manufacturing Science, Graduate School of Engineering, Osaka
University, Osaka (Japan). Mr. Kentaro YAMAMOTO (k-yama@jwri.osaka-u.ac.jp)
and Prof. Dr. Manabu TANAKA (tanaka@jwri.osaka-u.ac.jp) are both with Joining
and Welding Research Institute, Osaka University, Osaka (Japan).