You are on page 1of 16

International Journal of Pressure Vessels and Piping 78 (2001) 523±538

www.elsevier.com/locate/ijpvp

Analysis of residual stresses and distortions in


T-joint ®llet welds
Tso-Liang Teng a,*, Chin-Ping Fung b, Peng-Hsiang Chang b, Wei-Chun Yang c
a
Department of Mechanical Engineering, Da-Yeh University, 112, Shan-Jiau Rd., Da-Tsuen, Changhua 515, Taiwan, ROC
b
Institute of System Engineering, Chung Cheng Institute of Technology, Ta-Shi, Tao-Yuan 335, Taiwan, ROC
c
Ordnance Readiness Development Center, Nantou, Taiwan, ROC
Received 5 December 2000; revised 1 August 2001; accepted 7 August 2001

Abstract
T-joint ®llet welds are extensively used in ship engineering and bridge structures. Localized heating from the welding process
and subsequent rapid cooling induce tensile residual stress near the toe of the T-joint in ®llet welds. Welding produces thermal
stresses that cause structural distortions, which in¯uence the buckling strength of the welded structures. This study describes the
thermal elasto-plastic analysis using ®nite element techniques to analyse the thermomechanical behaviour and evaluate the residual
stresses and angular distortions of the T-joint in ®llet welds. Furthermore, this work employs the technique of element birth and
death to simulate the weld ®ller variation with time in T-joint ®llet welds. Also discussed are the effects of ¯ange thickness,
welding penetration depth, and restraint condition of welding on the residual stresses and distortions. q 2001 Elsevier Science Ltd.
All rights reserved.
Keywords: T-joint ®llet weld; Residual stresses; Angular distortions

1. Introduction distortion in relation to considerations of design and safety


is a relevant task.
Metallurgical joints made by welding are extensively For the prediction of the residual stresses and distor-
used in the fabrication industry, including ships, off- tions attributed to welding, previous investigations have
shore structures, steel bridges and pressure vessels. developed several experimental methods, including
Among the merits of such welded structures are high stress±relaxation [1], X-ray diffraction [2,3], ultrasonic
joint ef®ciency, water and air tightness, and low fabri- [4] and cracking [5]. In these methods the stresses are
cation cost. The types of welded joint can be classi®ed determined by experimental methods. With the develop-
into ®ve basic categories: butt, ®llet, corner, lap and ment of computer techniques, the ®nite element method
edge. T-joint ®llet welds are widely employed in for analysing thermomechanical behaviour in welded
ships, bridge structures and supporting frames for pres- structures has been further enhanced. For an analysis
sure vessels and piping. Due to localized heating by the of T-joint ®llet welds, Sasayama et al. [6] used the
welding process and subsequent rapid cooling, residual experimental method to determine the relation between
stresses and distortions can occur near the T-joint. High longitudinal shrinkage deformation and welding para-
residual stresses in regions near the weld may promote brit- meters on long T-joint ®llet welds. This work also
tle fracture, fatigue, or stress corrosion cracking. Mean- presented a formula describing the deformation process.
while, distortion in the base plate may reduce the buckling Meanwhile, Guyot [7] discussed the effect of transverse
strength of structural members. To accurately evaluate T- shrinkage on different types of ®llet welds and deduced
joint ®llet welds, predicting welding residual stresses and the shrinkage formula. Furthermore, Kumose et al. [8]
developed an experimental method to measure angular
distortions in single pass T-joint ®llet welds with differ-
ent welding parameters. Their investigation also consid-
* Corresponding author. Tel.: 1886-3-389-2131; fax: 1886-3-389-2131.
ered ways to improve angular distortion. Nagaraja [9]
E-mail address: g910404@ccit.edu.tw (T.-L. Teng). examined how the T-joint ®llet welds can be treated as
0308-0161/01/$ - see front matter q 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0308-016 1(01)00074-6
524 T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538

predicting thermomechanical behaviour. This investiga-


Nomenclature tion performs thermal elasto-plastic analysis using ®nite
r density element techniques to analyse the thermomechanical
C speci®c heat behaviour and evaluate the residual stresses and angular
T temperature distortions of the T-joint in ®llet welds. Additionally, it
t time also considers the effects of ¯ange thickness, welding
{q} heat ¯ux penetration depth, and restraint condition on residual
Q the rate of internal heat generation stresses and distortions. Information on how to improve
h unit outward normal vector the fabrication process of welded structures is also
hf ®lm coef®cient presented.
TB bulk temperature of the adjacent ¯uid
TA temperature at the surface of the model 2. Analysis model
{Te } nodal temperature vector
{Ds e } nodal stress increment matrix 2.1. Thermo-mechanical model
{Dep } {De } 1 {Dp }
{De } elastic stiffness matrix Welding residual stress distributions are calculated by
{Dp } plastic stiffness matrix a ®nite element method. Fig. 1 presents the analysis
{Ue } nodal displacement vector procedures.
‰BŠ strain-displacement matrix
{DT} temperature increment matrix 2.1.1. Thermal model
{Cth} thermal stiffness matrix In the thermal analysis, a total of 160 load steps
{DTe } nodal temperature increment matrix increasing from 0.001 to 10 s were required to complete
‰MŠ temperature shape function the heating cycle. Only 30 load increments were typi-
sz longitudinal residual stress cally required for the weldment to return to its initial
sX transverse residual stress (room) temperature. The time increments were auto-
matically optimised for each time step by the computer
program. The modi®ed Newton±Raphson method was
used in each time step for the heat balance iteration.
a pattern that was combined by a welding pass at the
This study simulates the weld thermal cycles for SAE
middle and edge of a butt-welded plate. Michaleris and
1020 steel shown in Fig. 2. The convective heat transfer
DeBiccari [10] designed a computational model to esti-
coef®cients on the surfaces were estimated (using engi-
mate buckling and deformation on large and complex T-
neering formulae for natural convection) to be
joint ®llet welds. Meanwhile, Arnold [11] estimated
15 W m 2K 21.
residual stresses in multipass ®llet welds using the ®nite
element code PAFEC. Furthermore, Finch and Burdekin
2.1.2. Mechanical model
[12] discussed the effects of residual stresses on differ-
In the mechanical analysis, the temperature history
ent kinds of T-joint ®llet weld defects using the ®nite
obtained from the thermal analysis was input as a ther-
element code ABAQUS. Finally, Ueda and Ma et al.
mal loading into the structural model. The thermal
[13] developed elastoplastic ®nite-element computer
strains and stresses can be calculated at each time
programs to improve the accuracy of two-dimensional
increment. Also, the ®nal state of residual stresses
symmetric ®nite-element models and help them
will be accumulated by the thermal strains and stresses.
approach three-dimensional models on T-joint ®llet
During each weld pass, thermal stresses are calculated
welds. Their investigation also discussed the effect of
from the temperature distributions determined by the
T-joint weld size, and welding parameters on the weld
thermal model. The residual stresses from each tempera-
residual stresses.
ture increment are added to the nodal point location to
Residual stresses and distortions are unavoidable in
determine the updated behaviour of the model before
welding, and the effects of these stresses and distortions
the next temperature increment. The material was
on welded structures cannot be disregarded. Deter-
assumed to follow the von Mises yield criterion and
mining residual stresses and distortions is thus an
the associated ¯ow rules. Phase transformation effects
important problem. However, accurate prediction of
were not considered in the current analysis due to
residual stresses and distortions induced by the welding
lack of material information, especially at high tempera-
process is extremely dif®cult because the thermal and
tures, such as the near-melting state.
mechanical behaviour in welding include local high
temperature, temperature dependence of material proper- 2.2. Element birth and death
ties, and a moving heat source. Finite element simula-
tion of the welding process is highly effective in The model in this study adopts the technique of
T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538 525

Fig. 1. Flow diagram of the analysis procedure.

element `birth and death' to simulate the weld ®ller model. An element's strain is also set to zero as soon
variation with time in T-joint ®llet welds. All elements as that element is `killed'. Similarly, when elements are
must be created, including those weld ®llers to be born, they are not actually added to the model, but are
`born' in later stages of the analysis. The method simply reactivated. When an element is reactivated, its
proposed does not remove elements to achieve the stiffness, mass, element loads, etc. return to their full
`element death' effect. Instead, the method deactivates original values. Thermal strains are computed for newly
them by multiplying their stiffness by a severe reduction activated elements according to the current load step
factor. Although zeroed out of the load vector, element temperature.
loads associated with deactivated elements still appear
in element-load lists. Similarly, mass, damping, speci®c
heat, and other such effects are set to zero for deacti- 2.3. Veri®cation
vated elements. The mass and energy of deactivated
elements are excluded from the summations of the The proposed method was compared with ®nite
526 T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538

Fig. 2. Simulated weld thermal cycles for SAE 1020 steel.

element and experimental results taken from Ma et al. the width of the ¯ange. The solid lines and broken
[13] and Shim et al. [14] to con®rm its accuracy. Ma et lines in Fig. 3 represent the residual stress computed
al.'s investigation computed the residual stress in T- by Ma et al. and this work, respectively. According to
joint ®llet welds using thermal elastic plastic three- Fig. 3, the residual stress distributions computed by the
dimensional FEM and generalized plane strain FEM. method proposed here show very good agreement with
Fig. 3 portrays the residual stress distributions across those determined by three-dimensional FEM.

Fig. 3. Residual stress computed by Ma et al.'s three-dimensional FEM and present method.
T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538 527

Fig. 4. Geometry of multipass butt weld.

For Shim et al.'s investigation, a specimen was presented experimental results for the problem. Addi-
constructed using multi-pass butt welding, with a length, tionally, the ABAQUS ®nite element package is applied
width and thickness of L ˆ 1000 mm; W ˆ 400 mm; t ˆ as a comparison. As Fig. 5 indicate, the ABAQUS
25:4 mm; respectively, as shown in Fig. 4. The welding package result showed slightly lower tensile transverse
used the submerged arc technique. Pass sequences and stress near the weld centreline. The present method
welding parameters are shown in Table 1. Figs. 5 and 6 tends to the experimental results near the surface. As
portray the distribution of the transverse and longitudi- Fig. 6 indicate, both analysis results show tensile stress
nal >residual stress on the thick plate computed by near the weld centreline.
Shim et al. and the present method. Shim et al. [14] The residual stress calculated using the present method

Table 1
Schematics of pass sequences along with welding parameters for each pass

Pass no. (1±11) Voltage (V) Current (A) Speed (mm sec 21)

1 25 190 3.34
2±5 26 215 4.70
6 25 190 3.37
7±9 26 220 4.70
10±11 27 250 4.70
528 T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538

Fig. 5. Transverse residual stress at the top surface of plate.

Fig. 6. Longitudinal residual stress at the top surface of plate.


T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538 529

Fig. 7. Geometry of T-joint ®llet welds.

correlates well with that determined using Ma et al.'s of the webs are assumed to be simultaneously welded
three-dimensional FEM and that found in Shim et al.'s under the same welding conditions. Therefore, the
experiments. Therefore, the procedure proposed here is T-joint ®llet weld can be considered to be symmetrical
considered appropriate for analysing residual stresses and with the Y±Z plane. his work develops a two-
distortions due to welding. dimensional symmetrical generalized plane strain model
to calculate the residual stresses of the T-joint ®llet
weld using the ®nite element method. With the aid
3. Analysis of T-joint ®llet weld of this generalized plane strain condition, the three-
dimensional residual stress components distributed in
3.1. Specimen and material properties the transverse section can be computed by thermal
elasto-plastic analysis using ®nite element tech-
Fig. 7 depicts two plate ®llet weld. The length of the
niques with unit thickness. The model employs two-
®llet weld, the width of the ¯ange and height of the
dimensional four node plane elements, including the
web are assumed to be 500, 200 and 100 mm, respec-
®nite element meshes for the ®llet weld, along with
tively. The plate thickness is 16 mm for the ¯ange and
re®ned meshes used in the weld area. The symmetric
12 mm for the web. The plate material is SAE 1020,
model has 439 elements and 514 nodes as shown in
and the mechanical properties are dependent on the
Fig. 9.
temperature history, as Fig. 8 illustrates.
3.4. Mesh sensitivity study
3.2. Welding conditions
To examine the adequacy of element sizes, the effect
The welding parameters chosen for this analysis of mesh re®nement in the weld area was studied. A new
were as follows: welding method, single pass gas model with re®ned meshes consists of 507 elements
tungsten-arc welding; welding current, I ˆ 260 A; and 585 nodes. Results from two mesh densities with
welding voltage, V ˆ 20 V; and welding speed, the same material model and geometry showed little
v ˆ 5 mm sec21 . For practical welds, the heat sources difference. Therefore, the original FEM model with-
are applied along the weld path. However, this investi- out mesh re®nement in the weld joint is used for this
gation simulates the increment of heat loading on the study.
welding process via the lead temperature curve as
shown in Fig. 2. 3.5. Analysis procedure

3.3. Finite element model for T-joint ®llet welds During each weld pass, thermal stresses are cal-
culated from the temperature distributions deter-
In the T-joint ®llet weld, the welds on both sides mined by the thermal model. The residual stresses
530 T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538

Fig. 8. The mechanical properties of T-joint ®llet weld.

from each temperature increment are added to the plates near the ®llet weld toes. The value of the residual
nodal point location to determine the updated beha- stress near the weld toes is 25 MPa and decreases to
viour of the model before the next temperature zero as the distance from the weld toes increases.
increment. Owing to the locally concentrated heat source, the
temperature near the weld bead and heat-affected zone
rapidly changes with distance from the heat source, i.e.
the highest temperature is limited to the domain of the
heat source, from which lower temperature zones fan
4. Results and discussion out. According to Fig. 10 the temperature non-
uniformity varies the shrinkage through the weldment
4.1. Transverse residual stresses thickness during cool-down and, consequently a high
tensile residual stress occurs on the surface of the
A stress acting normal to the direction of the weld weld toes.
bead is known as a transverse residual stress, denoted
s x . Fig. 10 represents the distributions of the residual 4.2. Longitudinal residual stresses
stress s x along the X direction. A very large tensile
residual stress is produced at the surface of the base A stress acting parallel to the direction of the
T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538 531

weld bead is termed a longitudinal residual stress,


denoted s z . Fig. 11 depicts the distributions of the
residual stress s z along the X-direction. The longi-
tudinal residual stress develops from longitudinal
expansion and contraction during the welding sequence.
Along the weld line, a high tensile residual stress
arises near the weld toes, and then decreases to zero,
®nally becoming compressive as distance from the
weld line on the ¯ange increases. The residual stress
value is 110 MPa, approaching the yield stress of the
material. Due to the self-equilibrium of the weldment,
tensile and compressive residual stress exists at
the weld toes and away from the welding line on the
¯ange.

4.3. Angular distortion

For the angular distortion of a T-joint ®llet weld, the


angular change Du of the ¯ange for T-type joints is
expressed by Du ˆ a=b (for small angular change),
where b is the half length of the ¯ange and a is
the displacement of the Y direction along the ¯ange
edge. This equation describes the angular change of
Fig. 9. Finite element meshes for the T-joint ®llet welds of 439 the ¯ange for the T-joint ®llet weld illustrated in
elements. Fig. 12. In T-type ®llet welding, Fig. 13 represents

Fig. 10. Transverse residual stress distribution along the X direction.


532 T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538

Fig. 11. Longitudinal residual stress distribution along the X direction.

the changes of angular distortion (Du ) with cooling in the upper portion exceeds that in the lower one.
time. This ®gure reveals that the angular distortion Fig. 13 reveals that, after 20 s of weldment cooling,
downward is about 0.006 rad when the weldment has the angular distortion upward is approximately
cooled for 2 s. This is because the thermal expansion 0.003 rad, and almost does not change. This is because
the upwards bend of the ¯ange due to plastic deforma-
tion in the upper portion exceeds that in the lower
portion.

4.4. Effect of ¯ange thickness

Fig. 14 presents the transverse residual stress s x


along the X direction for 10, 16 and 22 mm, related to ¯ange
thickness. All of the stress distributions indicate
tensile stresses near the weld toes, which then decrease
to zero as distance from the weld toes increases. Fig. 15
depicts the longitudinal residual stress distributions
along the X direction, for 10, 16 and 22 mm, related
to ¯ange thickness. All of the stress distributions
show tensile stresses near the weld line, that then
decrease to become compressive with increasing
distance from the weld line. Figs. 14 and 15 reveal
that with increasing ¯ange thickness, the residual stress
increases. Thus, ¯ange thickness affects the maximum
residual stress near the weld toe of the ¯ange in the
following two ways: (1) With increasing ¯ange thick-
ness, the temperature nonuniformity varies the thermal
expansion and shrinkage during cool-down, and, conse-
Fig. 12. Angular distortion (Du ) in T-joint ®llet welds. quently, the residual stress increases. (2) A thicker
T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538 533

Fig. 13. Angular distortion of T-joint ®llet weld.

Fig. 14. Transverse residual stress distributions for different ¯ange thickness.
534 T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538

Fig. 15. Longitudinal residual stress distributions for different ¯ange thickness.

¯ange strengthens the internal restraint and increases


residual stress.

4.5. Effect of welding penetration depth

In the welding process, different weldment thick-


nesses require different weld penetration depths to
avoid a non-penetration defect. This research selects
elements and controls the heat input to investigate the
effect of different weld penetration depths on residual
stresses and distortions. All simulation models have the
same dimensions and the same weld toe length. As Fig.
16 shows, the weld penetration depth is assumed to be 0
or 6 mm.
Figs. 17 and 18 present the distributions for the
different welding penetration depth of the residual
stresses s x and s z along the X direction. The resi-
dualstress for a 6 mm penetration depth ®llet weld
is smaller than that in a 0 mm penetration depth weld.
This difference is because the larger penetration
depth corresponds to an increase in heat input or a
reduction of the welding speed, enlarging the heat
affected zone and reducing the temperature variation
of the upper and lower surfaces of the ¯ange.
Furthermore, the distortion and welding residual stress
decrease.

4.6. Effect of restraint conditions

In order to reduce T-joint ®llet weld angular


distortion, an external clamp is frequently applied
Fig. 16. Different types of penetration depth. to the ¯ange, as Fig. 19 illustrates. This research
T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538 535

Fig. 17. Transverse residual stress distribution for different penetration depths.

investigates the effect of restraint conditions and reveals that the angular distortion with restraint is
restraint position on angular distortions and residual smaller than when the ¯ange is unrestrained. When
stresses. the applied restraint position is ®xed at 39.8 mm, this
Fig. 20 presents the angular distortion of the computation provides a minimum angular distortion
¯ange with various restraint positions. The ®gure 0.002 rad.

Fig. 18. Longitudinal residual stress distribution for different penetration depths.
536 T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538

5. Conclusions

This research employs the ®nite element method


to evaluate residual stresses and angular distortions in
T-joint ®llet welds. The technique of element birth and
death is used to simulate the weld ®ller variation with
time in T-joint ®llet welds. Additionally, it discusses the
effects of ¯ange thickness, welding penetration depth
and restraint condition of welding on residual stresses.
Based on the results in this study, we conclude the
following:

1. For transverse residual stresses, a high tensile


stress is produced near the ®llet weld toe. As
distance from the weld toe increases the stress
Fig. 19. Restraint condition in T-joint ®llet welding. approaches zero.
2. For longitudinal residual stresses, a very large
tensile stress occurs near the weld toe, and a
Figs. 21 and 22 show the distribution of the compressive stress appears away from the weld
restrained and unrestrained residual stresses s x , s z bead.
along the X direction. The value of the residual stress 3. The temperature distribution along the ¯ange thick-
in the restrained model is smaller than that in the ness causes ®llet weld angular distortions, which
unrestrained model. This difference occurs because bend the ¯ange up.
when the restraint is removed after welding, the ¯ange 4. With increasing ¯ange thickness, the internal
is slightly bent by the released restraint force, and this restraints are increased and the tensile residual stress
induces compressive stress on the top surface of the ¯ange near the ®llet weld toe increases.
and tensile stress on the bottom surface of the ¯ange. This 5. With increasing penetration depth or heat input
new pattern of stress allows the tensile residual stress near in ®llet welding, the tensile residual stress near the
the toe to be reduced. This phenomenon means that the ®llet weld toe decreases, and can also improve non-
restraint used to prevent angular distortion is also effective penetration defects.
in reducing the tensile residual stress near the weld toe. 6. In a restrained ®llet weld, the tensile residual stress

Fig. 20. The angular distortion of the ¯ange with various restraint positions.
T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538 537

Fig. 21. Transverse residual stress distribution with restraint and in the unrestrained condition.

Fig. 22. Longitudinal residual stress distribution with restraint and in the unrestrained condition.
538 T.-L. Teng et al. / International Journal of Pressure Vessels and Piping 78 (2001) 523±538

and angular distortion near the toe can be reduced [6] Sasayama T, Masubuchi K, Moriguchi S. Longitudinal deformation
after the restraint force is released. When the applied of long beam due to ®llet welding. Welding J 1955:581±2.
[7] Guyot F. A note on the shrinkage and distortion of welded joints.
restraint position is changed at the boundary, a mini- Welding J 1947:519±29.
mum angular distortion can be obtained. [8] Kumose T, Yoshida T, Abbe T, Onoue H. Predicting of angular
distortion caused by one-pass ®llet welding. Welding J 1954:945±56.
[9] Nagaraja NR, Estuar FR, Tall L. Residual stresses in welded shapes.
References Welding J 1964:295±306.
[10] Michaleris P, DeBiccari A. Prediction of welding distortion. Welding
[1] Pang HL, Pukas SR. Residual stress measurements in a Cruci- J 1997:172±81.
form welded joint using hole drilling and strain gauges. Strain [11] Arnold J, Robin FD, Goff P. Predicting residual stresses in multi-pass
1989:7±14. weldments with the ®nite element methods. Comput Struct
[2] Cheng W, Finnie IA. Method for measurement of axisymmetric axial 1989;32(2):365±78.
residual stresses in circumferentially welded thin-walled cylinders. [12] Finch DM, Burdekin FM. Effect of welding residual stresses on
J Engng Mater Technol 1985;107:181±5. signi®cance of defects in various types of welded joint. Engng Fract
[3] Chandra U. Determination of residual stress due to Girth-Butt welds Mech 1992;41(5):721±35.
in pipes. ASME J Pressure Vessel Technol 1985;107:178±84. [13] Ma NX, Ueda Y, Murakawa H, Madea H. FEM analysis of 3D weld-
[4] Chu SL, Peukrt H, Schnider E. Residual stress in a welded steel plate ing residual stresses and angular distortion in T-type ®llet welds.
and their measurements using ultrasonic techniques. MRL Bull Res Transaction of JWRI 1995;24(2):115±22.
Dev 1987;1(2):45±50. [14] Shim Y, Feng Z, Lee S, Kim D, Jaeger J, Papritan JC, Tsai CL.
[5] Masubuchi K, Martin DC. Investigation of residual stresses by use of Determination of residual stresses in thick- section weldments. Weld-
hydrogen cracking. Welding J 1961;40:553s±63s. ing J 1992:305±12.

You might also like