See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269100823

Finite Element Modelling of Filled Weld Connections

Conference Paper · December 2008

DOI: 10.1533/9781782420484.9.427

CITATIONS READS

0 602

2 authors:

Ślęczka Lucjan Aleksander Kozlowski

Rzeszów University of Technology Rzeszów University of Technology
29 PUBLICATIONS   62 CITATIONS    84 PUBLICATIONS   140 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

stability View project

PhD - Rotation capacity of stiffened bolted end-plate beam-to-column connections. View project

All content following this page was uploaded by Ślęczka Lucjan on 04 February 2019.

The user has requested enhancement of the downloaded file.

DFE2008 Design, Fabrication and Economy

Finite element modelling of fillet weld connections

Lucjan Ślęczka, Aleksander Kozłowski
Rzeszów University of Technology, ul. Poznańska 2, 35-082 Rzeszów, Poland,
[email protected]

Abstract

The paper describes the finite element modelling method employed to analyse welded
connections. The strength phenomena of fillet welds is based on ductile crack initiation and
propagation, so finite element model should include proper modelling of stress-strain
relationship, including hardening and softening effects, incorporation of geometrical
nonlinearity and different material models for base and weld metals and also for heat affected
zone. The occurrence of large plastic strains requires the use of the true stress-strain
relationship. Nonlinear stress-strain curve gives possibility to model local yielding and
redistribution of stresses. “Element dead” option available in the FE programs trace initiation
and propagation of ductile crack in this places, where ductility of welds is exhausted.
Keywords: Steel structures, connections, fillet welds, FE-modelling.
1. Introduction
Fillet welds are one of the most common components of steel connections.
In structural application, they are usually designed with some level of overstrength,
due to their small ductility, compared with other joint parts. The strength criteria for
statically loaded connections used in various codes were developed in last century,
based on results of extensive tests, (Design rules 1976). Such design rules in many
cases are likely conservative because of the large scatter of the test results and
limited number of tests. Now, finite element simulations give possibility to model
accurate behaviour of fillet weld connection up to failure. In many cases such failure
analysis estimating the real strength is necessary. The emphasis in such analysis is
usually placed on the assessment of overall behaviour, Jamshidi & Birkemoe (2005),
Rodriguez et al. (2004), Mellor et al. (1999). Only several researches have focused
on ductile failure of welded connection. Such studies, e.g. Wang and al. (2006),
have been performed mainly for aluminium joints, where fracture occurs indeed in
heat affected zone (HAZ). Failure analyses of welded connections are usually
performed with special user defined constitutive material model, which makes them
difficult to perform for non-highly advanced users. “Element dead” option in the
commercially available FEM programs give possibility to model initiation and
propagation of ductile crack of welds in easy way, provided that some basic
phenomena are included in numerical model.
2. Basic phenomena in FE modelling of welded joints
Finite element modelling of weld connections has to take into account some basic
phenomena to achieve accurate results.
2.1 Material heterogeneity
The welded joint consists of three different zones: weld metal, base metal and heat
affected zone. The weld and base metal are usually chemically and structurally
compatible, but in typical steel building structures the nominal material properties
 DFE2008 Design, Fabrication and Economy

of weld metal are usually higher then those of base metal. The weld metal strength
level and their ductility is also influenced by welding conditions. The heat affected
zone (HAZ) is area adjacent to the weld, where properties of base metal are changed
by heat input. For medium strength steel grade the yield strength and ultimate
strength of HAZ are usually increased. Moreover, the specific regions exist in HAZ
zone with evidently different microstructure and mechanical properties. The range
of HAZ depends strongly on welding thermal cycle. Normally, the width of HAZ
zone is narrow, usually has less than 2÷3 mm, after application of modern welding
technologies. Figure 1 shows zones of weld metal, base metal and HAZ in fillet
weld connection. Such simplified division can be used in FEM modelling.

Figure 1. Zones and contact surfaces in fillet weld connection: a) weld, b) base metal, c) heat
affected zone (HAZ).
2.2 Stress – strain relationship
The behaviour of fillet welds is affected by yield strength fy, but in particular the
strength is based on ultimate strength fu and ultimate strain εu, so stress-strain
relationship should be modelled in true (not engineering) values. Figure 2 shows
simplified multilinear true stress – true strain σ-ε curve, which can be implemented
in finite element models, where large plastic deformation occurs. Last branch of this
line in σ-ε relationship, where values of stress exceed fu, simulates the behaviour of
material after the necking. According to Faella et al. (2000) Table 1 gives suggested
values of σ-ε curve parameters for constructional steels.

fu Eu
1
Eh
fy 1
True stress σ

E
1
εy εh εm εu
True strain ε

Figure 2. σ-ε multilinear constitutive law.

DFE2008 Design, Fabrication and Economy

Table 1. σ-ε curve parameters for constructional steels (Faella et al. 2003)
Steel fy fu εh/εy E/Eh E/Eu εu/εy εu
grade MPa MPa [-] [-] [-] [-] [%]
S235 235 360 12.3 37.5 523.2 ≥ 357 ≥ 40
S275 275 430 11.0 42.8 447.6 ≥ 425 ≥ 56
S355 355 510 9.8 48.2 381.7 ≥ 479 ≥ 81
2.3 Geometrical nonlinearities and process of fracture
FE analysis of ductile fracture requires consideration of the material properties at the
necking phenomenon, which can be reached by stress – strain curve presented in
Figure 2, and also by application of the true ultimate strain εu. The ultimate strain εu
corresponds to the last point input for the strain-stress curve. When such strain is
reached in given element integration point, the corresponding finite element is
removed from FE model. This “element dead” option based on ultimate strain, give
possibility to model initiation and propagation of ductile crack in easy way, in the
commercially available FEM programs. The kinematic description in such nonlinear
analysis should take into account large displacement/large strain formulations.
2.4 Contact surfaces
The friction forces between plates in lap, welded connections are small enough to
neglect them. But modelling of frictionless contact conditions between plates
(Figure 1) can be beneficial. It will prevent the adjoining plates from penetration,
when their deformations occur, during load increase.
2.5 Residual stresses
It is well known that after the welding residual stresses occur in the connection.
They usually produce tension in the weld metal and compression in the surrounding
base metal, in case of one-pass welding. Because the resultant of residual tensile
stresses is compensated by resultant of residual compressive stresses, overall
capacity of welded joint is the same as for joint without residual stresses, so there is
tendency to omitting them in numerical simulations. But in case of large three
dimensional connections, residual stresses cause biaxial or triaxial stress state, and
ductility of the joint can be decreased.
3. Validation of finite element modelling of ductile fracture
All numerical calculations were performed using the FE package ADINA 8.3,
(ADINA R&D, Inc. (2005)). Figure 3 shows comparison of FE modelling results of
rounded specimen with 4 mm diameter with test results. The main aim of this
analysis was comparing numerical results of ductile fracture with data from tests.
The specimen was cut out from weld metal, done using E42 4 B42 electrodes. Its
mechanical characteristic were as follow: yield strength fy=425 MPa, ultimate
strength fu=550 MPa and ultimate strain εu=ln(A0/Af)=1,38, where A0, Af are initial
section area of the specimen and final section area (after the fracture) respectively.
Shape of stress-strain curve was modelled according to Table 1. Material model is
based on the Huber-von Mises yield condition, an associated flow rule using the
Huber-von Mises yield function and on an isotropic hardening rule. One half of the
 DFE2008 Design, Fabrication and Economy

specimen was modelled, using the 2D, 9 nodes axisymmetric elements. The size of
mesh was as 0,2x0,2 mm.
In spite of simplified σ-ε curve modelling as in Figure 1, quite good accuracy is
achieved in post critical behaviour (Figure 3 and 4). The difference in elongation
measured on 60 mm base is 18% between test and FEM results, and the difference
between ultimate force (when total fracture occurs) is only 0,1%.
30000

25000

20000 Test
Force [N]

ADINA
15000

10000

5000

0
0 5 10 15
Elongation [mm]

Figure 3. Modelling of tension test of rounded specimen.

a) b)

Figure 4. Necking of the specimen a), and necking of FE model b) after the fracture.
4. Parametric study of lap connections with transverse fillet welds
4.1 Effect of weld thickness
A view of investigated lap connections with fillet welds is shown in Figure 5. The
joint consist of three plates, which are connected using two fillet welds. There were
analysed connections with varying weld thickness: a=0,2t, a=0,3t, a=0,4t, a=0,5t
and a=0,6t, where thickness of the plate t is equal to 12 mm. Due to symmetry only
half of the joint was modelled, Figure 6. One multilinear σ-ε material model was
used to simulate weld, base and HAZ metal, with parameters described in previous
chapter (in joint with weld thickness a=0,6t central plate was modelled using elastic
material model to ensure weld failure). Average size of FE elements in weld zone
was equal to 0,25x0,25 mm. 2D, plane strain, 9-nodes elements were used with
thickness equal to 1,0 mm. The frictionless contact was assumed between upper and
the central plate. No residual stresses have been modelled.
DFE2008 Design, Fabrication and Economy

Figure 5. General view of analysed fillet weld connection.

Figure 6. FE model of the joint.

a) b) c)

d) e) f)

g) h)

Figure 7. Evolution of yielding and fracture of weld a=0,4t with applied load: a) elastic stage
F=1004 N, b) elastic-plastic stage F=1445 N, c) F=1673 N, d) F=2170 N, e) F=Fmax=2639 N,
f) propagation of fracture F=2292 N, g) F=1268 N, h) total fracture F=0 N.
 DFE2008 Design, Fabrication and Economy

The weld failure mode obtained from FE simulation is shown in Figure 7. It is based
on tearing the parent metal along the weld contour, parallel to direction of loading.
This form of fracture is very similar to that observed experimentally, Grondin at al
(2002), Callele at al. (2005).
Figure 8 shows joint response curves with different weld thickness, where weld
deformation ∆ is assumed as measured difference between displacement of point “1”
and “2” in direction of loading (Figure 5). It can be noticed that analysis show the
same initial stiffness of the welds, approximately equal 50000 N/mm2 in elastic
stage.
4500
Force, F [N]

4000
3500
3000
2500
2000
1500
a0_6
1000
a0_4 a0_5
500
a0_2 a0_3
0
0 1 2 3 4 5 6
Deformation ∆ [mm]

Figure 8. Response of the connections with different weld thickness.

600

500
a0_6
σ=F/1*a, [MPa]

400
a0_4 a0_3
300
a0_2
200
a0_5
100

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Normalized deformation, ∆/a [-]

Figure 9. Normalized response curves.

Figure 9 shows the same response curves, but in normalized values. Normalized
deformation is obtained by dividing measured deformation of weld ∆ by their
thickness a. Values at ordinate axis show stress at throat of the weld, σ=F/1.a. As
can be seen, average normalized deformation ∆/a at ultimate load is equal
DFE2008 Design, Fabrication and Economy

approximately to 0,2, which is value quite close to this obtained by Grondin et al.
(2002).
4.2 Effect of HAZ modelling
The joint with weld thickness a=0,4t was selected for further comparison. Figure 10
shows the joint response curves with, and without HAZ modelling. Model “a0_4”
has only one material constitutive model for base, weld and HAZ metal. Model
“a0_4_HAZ_1” has one material model for weld and base metal, with mechanical
properties as described in chapter 3, and the second one with fy,HAZ=625 MPa,
fy,HAZ=750 MPa and εu,HAZ=0,9 for HAZ metal. Also models with decreasing value of
εu,HAZ, equal to εu,HAZ=0,7 and 0,5 respectively were analysed. But in these cases,
there was no difference in total joint response compared to “a0_4_HAZ_1” model.
As can be seen, increase of yield strength of HAZ decreases a little deformation of
weld. The weld failure pattern, is independently of HAZ modelling.
600

500 a0_4
a0_4_HAZ_1
σ=F/1*a, [MPa]

400

300

200

100

0
0 0.2 0.4 0.6 0.8 1
Normalized deformation, ∆ /a [-]

Figure 10. Effect of HAZ modelling.

4.3 Effect of weld metal ductility
The next analysed parameter was influence of weld metal ductility on behaviour of
connection. Also the joint with weld thickness a=0,4t was selected for comparison,
marked as “a0_4” (Figure 11). The next models, “a0_4_weld_1” ÷ “a0_4_weld_3”,
have the same geometry, and also one material model for weld, base and HAZ metal
properties. The shape of σ-ε material constitutive law was constant, but values of
ultimate strain was decreasing from εu=1,0 for “a0_4_weld_1”, εu=0,8 for
“a0_4_weld_2” and εu=0,7 for “a0_4_weld_3”. It can be seen, that along with
decreasing weld metal ductility, also ultimate capacity (maximum stress) is
decreasing.
5. Conclusions
The ultimate capacity of fillet weld connections is governed by the strength of weld
metal fy, fu and its ultimate strain εu. Described above methods of FE modelling give
 DFE2008 Design, Fabrication and Economy

possibility to analyse fillet weld connections in fast and easy way. Average total
calculation time of described above FE models was equal approximately to 400s on
the desktop computer.

600
a0_4
500 a0_4_weld_1
a0_4_weld_2
σ=F/1*a, [MPa]

400 a0_4_weld_3

300

200

100

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized deformation, ∆/a [-]

Figure 11. Effect of weld metal ductility.

References
ADINA R&D, Inc. (2005) ADINA Theory and Modeling Guide. Report ARD 05-7.
Watertown, MA 02472 USA.
Callele, L. J., Grondin, G. Y. & Driver, R.G. (2005) Strength and behaviour of
multi-orientation fillet weld connections. Structural Engineering Report No. 255,
University of Alberta.
Design rules for arc-welded connections in steel submitted to static loads. (1976)
Welding in the World. Vol. 14. No. 5/6 1976.
Faella, C., Piluso, V. & Rizzano. (2000) Structural steel semirigid connections.
Theory, Design and Software. CRC Press.
Grondin, G.Y., Driver R.,G. & Kennedy D. J. L. (2002) Strength of transverse fillet
welds made with filler metals without specified toughness. Research report,
University of Alberta.
Jamshidi, A.K. & Birkemoe P.C. (2005) Strength analysis of fillet welded steel
connections. In: 4th European Conference on Steel and Composite Structures,
Maastricht, June 8-10, Volume B, paper 1.12.
Mellor, B.G., Rainey, R.C.T. & Kirk N.E. (1999) The static strength of end and
T fillet weld connections. Materials and Design 20, No.4, 193-205.
Rodriguez, J. L., Alvarez, E. R. & Moreno, F. Q. (2004) Study of the distribution of
tensions in lap joints welded with lateral beads, employing three dimensional
finite elements. Computers and Structures 82, No.15-16, 1259-1266.
Wang, T., Hopperstad, O. S., Larsen P. K & Lademo O.-G. (2006) Evaluation of a
finite element modelling approach for welded aluminium structures. Computers
and Structures 84, No.29-30, 2016-2032.

View publication stats