Proceedings of ESDA2006

8th Biennial ASME Conference on Engineering Systems Design and Analysis

July 4-7, 2006, Torino, Italy

ESDA2006-95125

FINITE ELEMENT ANALYSIS OF RESIDUAL STRESSES ON BUTT WELDED JOINTS

Enrico Armentani, Renato Esposito, Raffaele Sepe

Department of Industrial Design and Management, University of Naples Federico II

Piazzale V. Tecchio, 80 – 80125 Naples, Italy
Tel. +390817682450, Fax +390817682466
E-mail: [email protected]

ABSTRACT tensile residual stresses can be a major source of cracking and

Localized heating during welding, followed by rapid fracture in welding whereas compressive residual stresses in
cooling, usually generates residual stresses in the weld and in the the base plate may reduce the buckling strength of structure
base metal. Residual stresses in welding processes give members. However evaluating residual stresses associated to a
significant problems in the accurate manufacture of structures welded joint is extremely complicated. Difficulty in
because those stresses heavily induce the formation of cracks in determining these stresses is emphasized by the thermal
the fusion zone in high strength steels. Therefore, estimating the transient, by the variation of the thermal and mechanical
magnitude and distribution of welding residual stresses and properties of the material with the temperature and by the non
characterizing the effects of certain welding conditions on the linear heat losses. Therefore, it is necessary to know the
residual stresses are deemed necessary. In this work, residual magnitude and distribution of welding residual stresses in order
stresses and distortions on butt welded joints are numerically to facilitate the design and fabrication of welded structures.
evaluated by means of finite element method. The FE analysis Several experimental destructive and non destructive
allows to highlight and evaluate the stress field and his gradient techniques for directly measuring residual stress have been
around the fusion zone of welded joints, higher than any other developed. These techniques include X-ray diffraction method
located in the surrounding area. Temperature-dependent material [1], neutron diffraction method [2,3], layer-removal method,
properties, welding velocity, external mechanism constraints, sectioning method, ultrasonic and magnetic methods, and hole
technique of ‘element birth and death’ and latent heat of fusion drilling method [4,5]. However, it is impossible for any
are also taken into account. Some numerical results are compared experimental technique to obtain a complete distribution of
with experimental data showing a very good correlation. residual stress and distortion in a general welded structure. So
computational simulation plays an indispensable role to resolve
INTRODUCTION these complex problems.
Fusion welding processes are widely employed in several In the last years, several simulation tools based on finite
industrial fields such as aerospace, shipbuilding, automotive, element method have been applied to calculate the residual
and machinery. However when structures are manufactured by stresses. Siva Prasad and Sankaranarayanan [6] used adaptive
welding a non-uniform temperature distribution is produced. grids based on temperature calculation to evaluate residual
This distribution initially causes a rapid thermal expansion stresses. Chen and Kovacevic [7] used a three-dimensional
followed by a thermal contraction in the weld and surrounding model based on a finite element method to simulate a friction
areas, thus generating inhomogeneous plastic deformation and stir welding. Chang and Teng [8] used element birth and death
residual stresses in the weldment when it is cooled. to simulate the weld filler variation with time in multi-pass butt
The presence of residual stresses can reduce the welded joints. Gery et al. [9] used a heat source model based on
performance of welded structures; in fact the residual stress the Goldak’s method and performed a thermal simulation of
distribution is the initial stress of a welded structure and it plate butt joint MIG (Metal Inert Gas) welding processes using
should be combined with the applied service loading. Moreover FE transient heat transfer analysis.

1 Copyright © 2006 by ASME

 ⎛ ∂ 2T ∂ 2T ∂ 2T ⎞ ∂T
k ⋅⎜ 2 + 2 + 2 ⎟ + Q = ρ ⋅C p ⋅
In the present work an elastic-plastic finite element model
⎜ ∂x ∂y ∂z ⎠ ⎟ ∂t
⎝
considering temperature dependent mechanical properties is , (1)
used to evaluate residual stresses. In this study a parametric
model is adopted and the elements birth and death are used in
where ρ, Cp, and k are the density, specific heat capacity and
single-pass butt welded joint to simulate the weld filler
thermal conductivity of the plate material respectively and Q is
variation with time. Then analytical results are compared with
the rate of internal heat generation. Moreover a non-linear
experimental data available in [10].
material is modeled, because the parameters ρ, Cp, k depend on
the temperature. Eq. (1) is a differential equation governing
NOMENCLATURE heat conduction in a solid body. The initial and boundary
conditions of the problem are:
T ( x, y, z ,0) = T0 ( x, y, z )
C specific heat capacity
E Young’s modulus (2)
FZ Fusion Zone
⎛ ∂T ∂T ∂T ⎞
q = − k ⋅ ⎜⎜ + + ⎟⎟ ,
HAZ Heat Affected Zone

⎝ ∂x ∂y ∂z ⎠
I welding current (3)
Q heat generation rate
MIG Metal Inert Gas
S surface where q is the heat flux at the boundary. In welding problems,
T temperature the heat flux q may consist of one or more of the following
T0 room temperature modes: convective heat loss, radiative heat loss and boundary
V welding voltage heat. The heat flux losses on the surfaces of the welded plate
h convection coefficient due to the convection and radiation are given by:
k thermal conductivity q c = h ⋅ (T − T0 ) (4)
q heat flux
qc = ε ′ ⋅ ξ ⋅ (T 4 − T04 )
s plate thickness
t time (5)
v welding velocity where T0 is the room temperature, ξ is the Stefan Boltzmann
α coefficient of thermal expansion constant and (h, έ) are convection and emissivity coefficients,
ε strain respectively for all surfaces of the plate. In this study the flux
έ emissivity coefficient losses due to the radiation are not considered. In addition, a
υ Poisson’s ratio latent heat of fusion is taken into account to consider the phase
ξ Stefan Boltzmann constant transformation.
ρ density
σ stress
ys yield stress 1.2 Mechanical model
σx transversal residual stress The temperature history during the thermal analysis at the
σy longitudinal residual stress nodes of the FE mesh is calculated and saved, for the
η efficiency of heat-input mechanical analysis. During the structural analysis, these
temperature histories are used as thermal loading to produce the
residual stress field. In this way thermal strains and stresses can
1. ANALYSIS MODEL be calculated at each time increment. In thermo-mechanical
A theoretical evaluation of welded induced residual analysis, the plastic deformation of materials is assumed to
stresses and deformation can be achieved in two steps: (a) heat consider the Mises yield criterion and isotropic strain hardening
flow analysis; (b) residual stress and deformation analysis. rule. The rate relationship between thermal stress and strain is

[dσ ] = [D ep ] ⋅ [dε ] − [C th ]
Although the welding process can be simulated in one step by a described by the following equation:
coupled thermo-mechanical analysis, the uncoupled technique
allows to perform all computations more quickly and to utilize ⋅ dT (6)
fewer computer resources.

[D ] = [D ] + [D ] ,
being
1.1 Thermal model ep e p
(7)
In the thermal analysis, the transient temperature
distribution T of the welded plate is a function of time t and where [De] is the elastic stiffness matrix, [Dp] is the plastic
coordinates x, y, z, and the balance relation of heat flow of a stiffness matrix, [Dep] is total stiffness matrix, [Cth] is the
volume bounded by an arbitrary surface S is given by the thermal stiffness matrix, dσ, dε, and dT are the stress, strain and
following equation: temperature increment, respectively.

2 Copyright © 2006 by ASME

Figure 1. Geometry of the butt welded joint.

Figure 2. Finite element mesh for one plate.

3 Copyright © 2006 by ASME

2. FINITE ELEMENT MODEL FOR BUTT JOINT As Figs. 5-8 indicate, when temperature increases, the
Young’s modulus and the yield stress decrease while the
2.1. Geometry, materials properties and welding thermal expansion increases, whereas Poisson’s ratio decreases
parameters slightly until 1100 K and then increases up to 1800 K.
Fig. 1 shows the geometry of the welded joint considered
in this study. The plates of the joint are 8 mm thick and the
joint is 500 mm long in the welding direction; they are joined
by a single-pass butt weld. The material of the plate and 2

E x10 5[N/mm 2]
welding are made of mild steel, and the material thermal
properties are taken from Prasad and Sankaranarayanan [6] and
these vary according to the temperature history, as illustrated in
Figs. 3 and 4. The convective heat coefficient on the surfaces 1
was estimated to be 15 W/mm2°K according to [10].
The total effective heat generation rate input Q = ηIV is
assumed to be 1650 W; it corresponds to the following
welding parameters: welding current I = 110 A, welding 0
500 1000 1500 2000
voltage V = 20 V, welding velocity v = 3,33 mm s-1 and the
T [K]
efficiency of heat-input η = 0,75.
Moreover material mechanical properties (E, ys, α, υ) are Figure 5. Variation of Young’s modulus with
dependent on the temperature as showed in Figs. 5-8. temperature.

0.06 800

0.05 600
k [W/mm K]

ys [N/mm 2]

0.04 400

0.03 200

0.02 0
500 1000 1500 2000 500 1000 1500 2000
T [K] T [K]

Figure 3. Thermal conductivity k. Figure 6. Variation of yield stress with temperature.

800
2.5

700 2
α x10 [mm/K]
C [J/kg K]

1.5
600
-5

1
500
0.5

400 0
500 1000 1500 2000 500 1000 1500 2000
T [K] T [K]

Figure 4. Specific heat capacity C. Figure 7. Variation of thermal expansion coefficient

with temperature.

4 Copyright © 2006 by ASME

 vector is zero, element loads associated with deactivated
elements still appear in element-load lists. In the same way,
0.5 mass, damping, specific heat, and other such effects are set to
zero for deactivated elements. An element’s strain is also set to
zero as soon as that element is killed. Likewise, when elements
are ‘‘born’’, they are not actually added to the model, but are
ν

0.4 simply reactivated. When an element is reactivated, its

stiffness, mass, element loads, etc. return to their full original
values.
The weld deposit is modelled by activation of elements
0.3 immediately behind the centre of the heat source. When
500 1000 1500 2000 activated, an element adopts the nodal temperatures from the
T [K] surrounding elements already active. The front nodes do not
share the nodal temperatures with elements and are therefore
Figure 8. Variation of Poisson’s ratio with set to ambient temperature. Elements activated at the start of a
temperature. time increment consequently acquire an amount of energy
before the volumetric heat flux becomes effective within the
2.2 Computational simulation element. Such an error, as that caused by the inactive fillet
To solve thermo-mechanical problems several finite elements in the front ellipsoid are accounted for by adjusting
element models can be applied. The appropriate model should the net heat input.
be chosen considering the type of results expected from the The dynamic activation of fillet elements in a mechanical
model. The 2D models have important advantages in model is based on average element temperature. Fillet elements
computing time and necessary computer resources. At the same containing the maximum nodal temperature are activated at
time a 2D analysis neglects the heat propagation in the solidification temperature. The temperature dependent element
direction perpendicular to the considered plane. In this way the activation enhances numerical stability. The filler will
calculated temperatures turn out to be over-predicted compared dynamically couple the plate in distorted geometrical
to the analogous 3D scheme. This work develops a two- configuration. The coupling is obtained through constraint
dimensional symmetrical plane stress model (assuming zero equations, applied and updated incrementally at the end of each
out-of-plane stresses) to estimate the residual stresses. The time step.
model employs two dimensional four-node plane elements
including the finite element meshes for the butt welded joint. 3. RESULTS AND DISCUSSION
The 2D finite element mesh used in our computation is shown Two-dimensional finite element analyses have been carried
in Fig. 2. The symmetric model has 1614 elements and 1685 out, based on the models and physical properties and welding
nodes and the minimum element size in the mesh is 1 mm. process parameters introduced above. The temperature field
In the thermal analysis, a total of 480 load steps, and residual stress distribution are computed and analysed;
corresponding to t = 150 s, were required to complete the details of the characteristics of temperature field and residual
heating cycle, while another 88 load steps (up to t = 2096 s) stress distribution are discussed in the following paragraphs.
were required for the weldment to return into its initial room
temperature. The incremental time in the analysis is checked 3.1 Temperature distribution
automatically by the computer program and the full Newton- Fig. 9 shows the temperature distribution during welding
Raphson method was used to obtain the incremental processes at welding time of 117 s. The centre of the welding arc
calculation. at this time is at the position of x = 0 mm and y = 390 mm. The
temperature field has a normal distribution with the highest
temperature at the centre of the welding arc; as the distance
2.3 Element birth and death increases from the welding arc centre, the temperature decreases
The model in this study adopts the technique of element monotonously and continuously at different rates in all directions,
birth and death; this procedure can be used for modelling the either in longitudinal direction (parallel to the weld bead) or in
effects of structural changes. In finite element models created transverse direction (normal to the weld bead).
for welding applications, the process of filler metal addition For a more detailed and quantitative analysis of the
can be simulated. In the case of metal cutting, the removed temperature field, isotherms are given in Fig. 10. The welding
parts an be also modelled. All elements must be created, time for this figure is also 117 s. Owing to the geometrical
including those weld fillers to be born in later stages of the symmetry with respect to the weld centreline, only one half of
analysis. To achieve the “element death” effect the elements are the plate is provided. The peak temperature around the torch is
not actually removed. These are deactivated by multiplying calculated to be just above 2036 K suggesting melted material
their stiffness by a severe reduction factor. Although the load in the fusion zone. High temperatures are present near the

5 Copyright © 2006 by ASME

fusion zone (FZ), which defines the heat affected zone (HAZ). itself where the heat source is going to pass. The heat inputs
The thermal gradients are seen to be very steep in front of generated by moving its source along the welding line are
the heat source; furthermore near the weld line, the distance gradually transferred in all directions of the plate by conduction
between the isotherms is increased as the cooling rate and convection.
decreases. The torch also presents a very small area in front of

2400

2200

2000

1800

1600
Temperature [K]

1400

1200

1000

800

600

400
500
150 400
100
50 300
m]
Tran 0 200 y [m
sv er io n
se po -50 o sit
sitio 100 al p
n x -100 din
[mm gitu
] -150
0 Lon

Figure 9. Temperature distribution during welding process (t = 117 s).

150
Transverse position x [mm]

100

380
50
38
470 0
570 1400
47
0 670 770 1000 1950
0
0 50 100 150 200 250 300 350 400 450 500
Longitudinal position y [mm]

Figure 10. Isotherms [K] during welding process (t = 117 s).

6 Copyright © 2006 by ASME

 Fig. 11 and 12 show the temperature distribution of the 1400
welded plate along the transverse direction (x) calculated Present FEM
using the technique of element birth and death when the Experimental (Ref.10)
welding torch passes the midsection (y = 250 mm) and 7,5 s 1200
after the passing of the arc. They show that the temperature in
FZ (x < 3,2 mm) is substantially higher when the welding
torch just passes the plane but decreases rapidly with time.
Fig. 13 illustrates the temperature distribution along 1000
longitudinal direction when the torch is positioned at y = 250 mm.

Temperature [K]
It shows a quick temperature drop close to FZ (x = 3.2 mm) but
temperatures further away from FZ increase at different rates which 800
are proportional to their transverse distance (x) from the weldment.
From Fig. 14 it can be noted that the variations of
temperature versus time have the same tendency for all four
600
selected locations. Before arrival of the welding arc (t < 9 s),
the temperature remains close to the room one. When the
welding torch approaches the cross section at y = 30 mm (t = 9
s), the temperatures of these points increase to their respective 400
maximum at different rates. As the welding torch moves away
from the cross section (t > 9 s), the temperatures of these
points decrease gradually at different rates with the proceeding
200
of the welding process. Although the overall varying tendency 0 10 20 30 40 50 60 70 80
of these temperatures versus time curves is essentially the T rasverse distance [mm]
same, there are great differences in the maximum temperatures
that can be reached. Figure 12. Transverse temperature distribution at
midsection (y = 250 mm) 7.5 s after passing arc.

2200 2200
Present FEM x=0 mm
2000 2000 x=6 mm
Experimental (Ref.10)
x=12 mm
1800 1800
x=18 mm

1600 1600
Temperature [K]
Temperature [K]

1400 1400

1200 1200

1000 1000

800 800

600 600

400 400

200 200
0 10 20 30 40 50 60 70 80 0 50 100 150 200 250 300
T rasverse distance [mm] Longiyudinal distance [mm]

Figure 11. Transverse temperature distribution at Figure 13. Longitudinal temperature distribution at
midsection (y = 250 mm) during passing arc (t =75 s). time of 75 s.

7 Copyright © 2006 by ASME

 2200 300
x=0 mm Present FEM
2000 x=6 mm Experimental (Ref.10)
250
x=8 mm
1800
x=12 mm

1600 200

Residual stress σy [MPa]

Temperature [K]

1400
150
1200

100
1000

800 50

600
0
400

200 -50
0 10 20 30 40 50 60 70 80 90 100 0 25 50 75 100 125 150 175
T ime [s] T ransverse distance [mm]

Figure 14. Computed temperature histories of points Figure 15. Longitudinal stress distribution at
(y = 30 mm) with different distances to central line. midsection (y = 250 mm).
3.2 Residual stresses distribution 16
Residual stresses are evaluated using the same mesh that Present FEM
was used for calculating the temperatures.
14 Experimental (Ref.10)
Normal stresses parallel to the direction of the weld are
called longitudinal residual stresses and denoted by σy. The
longitudinal residual stresses develop from longitudinal 12
expansion and contraction during the welding sequence. Fig.
15 shows the distribution of the residual stresses σy along 10
Residual stress σx [MPa]

the x direction in the midsection (y = 250 mm). The self

equilibrium of the weldment is such that the tensile and 8
compressive residual stresses are present at the weld bead and
away from the welding line on the plates. Along the weld line,
high tensile residual stresses are present in regions near the 6
welding line, due to a resistance contraction of the material as
cooling begins, and then decrease to zero, finally becoming 4
compressive. The computed stress values are in very good
agreement with the experimental values available in [10]. 2
Normal stresses orthogonal to the direction of the weld
are known as transverse residual stresses, denoted by σx.
0
Tensile residual stresses are present in regions near the
welding line, then decrease close to zero as the distance from
the welding line increases. Fig. 16 shows the distribution of -2
0 25 50 75 100 125 150 175
the residual stress σx along the x direction in the midsection
T ransverse distance [mm]
(y = 250 mm). Once more numerical results show good
compatibility with experimental results near the welding line.
Fig. 17 and 18 show the distribution of the residual stresses Figure 16. Transverse stress distribution at
σy and σx along the y direction. As these figures reveal, the stress midsection (y = 250 mm).
distributions are almost symmetrical at the middle of the plate.

8 Copyright © 2006 by ASME

 4. CONCLUSIONS
250 To evaluate residual stresses in a welding process, 2D
nonlinear thermal and thermo-mechanical analyses are
achieved using FE analyses. Moreover the technique of
element birth and death is used to simulate the welding filler
200 varying with time in weldments. Furthermore numerical
results are compared with experimental data. The main results
and conclusions are summarized as follows:
Residual stress σy [MPa]

1. temperature near the weld bead and the HAZ decrease

150 rapidly with the distance from the center of the heat source;
2. large tensile longitudinal residual stresses are present
near the welding bead, and the magnitude of these stresses
equals the yield stress. Instead compressive stresses appear
100 away from the welding line;
3. high transverse residual stresses are produced near the
weldments, then decreasing close to zero as the distance from
the welding line increases.
50
REFERENCES
[1] Murugan, S., Sanjai, K. Rai, Kumar, P.V., Jayakumar, T.,
Baldev Raj, Bose, M.S.C., 2001, “Temperature distribution
0 and residual stresses due to multipass welding in type 304
0 100 200 300 400 500
stainless steel and low carbon steel weld pads”, International
Longitudinal distance [mm]
Journal of Pressure and Piping, 78, pp. 307-317.
[2] Man Jin Park, Hee Nam Yang, Dong Y. Jang, Jong Sung
Figure 17. Longitudinal stress distribution along the Kim, Tae Eun Jin, 2004, “Residual stress measurement on
y direction (x = 0). welded specimen by neutron diffraction” Journal of Materials
Processing Technology, 155–156, pp. 1171–1177.
25 [3] Owen, R.A., Preston, R.V., Withers, P.J., Shercliff, H.R.,
Webster, P.J., 2003, “Neutron and synchrotron measurements
of residual strain in TIG welded aluminum alloy 2024”
0
Materials Science and Engineering, A346, pp.159-167.
[4] Schajer, G.S., 1988, “Measurement of non-uniform residual
stresses using the hole-drilling method,” Part. I and II, Journal
of Engineering Materials and Technology, 110(4), pp. 338-343.
-50 [5] George Roy, Mac Braid, Guowu Shen, 2003, “Application
Residual stress σx [MPa]

of ADINA and hole drilling method to residual stress determi-

nation in weldments”, Computers and Structures, 81, pp. 929–935.
[6] Siva Prasad, N., Sankaranarayanan, T.K., 1996,
“Estimation of residual stresses in weldments using adaptive
-100 grids”, Computers and Structures, 80(6), pp. 1037-1045.
[7] Chen, C.M., Kovacevic, R., 2003, “Finite element
modeling of friction stir welding – thermal and thermo-
mechanical analysis”, International Journal of Machine Tools
& Manufacture, 43, pp. 1319-1326.
-150
[8] Chang, P.H., Teng, T.L., 2004, “Numerical and
experimental investigations on the residual stresses of the butt
welded joints”, Computational Materials Science, 29, pp. 511-522.
[9] Gery, D., Long,H., Maropoulos, P., 2005, “Effects of
-200 welding speed, energy input and heat source distribution on
0 100 200 300 400 500 temperature variation in butt joint welding”, Journal of
Longitudinal distance [mm] Materials Processing Technology, 167, pp. 393-401.
[10] Argyris, J.H., Szimmat, J., Williams, K.J., 1985, “Finite
Figure 18. Transverse stress distribution along the y element analysis of arc welding process”, Numerical Methods
direction (x = 0). in Heat Transfer, 3, pp. 1-33.

9 Copyright © 2006 by ASME