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

net/publication/283462197

Finite Element Calculation and Assessment of Static Stresses in Load-

Carrying Fillet Welds

Conference Paper · January 2003

CITATIONS READS

8 4,884

3 authors, including:

Wolfgang Fricke
Technische Universität Hamburg
347 PUBLICATIONS   4,042 CITATIONS   

SEE PROFILE

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

Fatigue strength improvement by weld toe grinding and weld profiling View project

Fatigue strength of welded structures at sub-zero temperatures View project

All content following this page was uploaded by Wolfgang Fricke on 03 November 2015.

The user has requested enhancement of the downloaded file.

International Institute of Welding Institut International de la Soudure
German Delegation Délégation Allemande

IIW-Doc. XV-1151-03

Finite Element Calculation and Assessment of

Static Stresses in Load-Carrying Fillet Welds
W. Fricke1, M. Mertens2, C. Weissenborn1
1)
Technical University Hamburg-Harburg, Germany
2)
University of Applied Sciences, Bremen, Germany

June 2003

Abstract

Fillet welds have to be designed with a sufficient strength with respect to

static as well as cyclic loads. The present paper deals with the calculation
of the relevant stresses for static strength, which are usually assessed re-
garding the ultimate load-carrying capacity of the weld. After reviewing the
current procedure, where nominal stress components in the weld throat are
determined analytically, the application of the finite element method is
demonstrated by the example of a fillet-welded lifting-eye. The stresses
from coarse as well as fine meshes are compared with the analytical re-
sults. High stress peaks are observed in fine meshes. Some problems oc-
cur when deriving the nominal stress components in the weld from the ele-
ment stresses in the finite element models. A non-linear calculation with the
fine-mesh model shows the re-distribution of stresses after yielding, where
the high stress peaks resulting from the linear stress analysis disappear.
Finally conclusions are drawn with respect to the calculation of nominal
stress components in fillet welds from finite element models, which can be
used in the subsequent strength assessment of the welds.
1 Introduction
Stresses in load-carrying fillet welds of steel structures are usually calculated very simply by con-
sidering the forces and moments which have to be transferred through the weld throat. This leads
to nominal axial and shear forces or, else, to stresses in the weld section which can be checked
against permissible values defined in codes and guidelines. Eurocode, for instance, defines either
a resultant force per unit length or - in the Annex M - an equivalent stress in the throat area, which
is assessed on the basis of the ultimate strength of the weld. DIN 18800 defines a special equiva-
lent stress in the weld throat area corresponding to the vector sum of the relevant stress compo-
nents. The permissible stresses are again based on the ultimate strength of fillet welds.

With increasing utilization of numerical methods such as the finite element method, questions are
raised how to model the weld and evaluate the stresses obtained from different mesh densities. In
very fine meshes, high stress peaks may be computed where it is not clear if these affect the
strength. In this connection it should be kept in mind that under static load normally a ductile failure
occurs, associated with yielding and stress re-distribution. This justifies the relatively simple as-
sessment with the nominal stresses mentioned above.

Many finite element calculations of fillet-welded joints were performed in the past, e. g. Dietrich
(1990), however mostly with the objective to assess the stress concentration at the weld toe or root
with respect to fatigue. The problem, how to handle static stresses obtained from finite element
models, is studied in this paper by considering a real failure case, where the fillet welded connec-
tion of a lifting-eye failed after unintended out-of-plane loading.

The lifting-eye, which is made from a 40 mm thick plate with yield strength 355 MPa and welded to
a 10 mm thick base plate, is sketched in Fig. 1 together with the critical load situation. The force F
is acting at the shackle in a way, that a large bending and torsional moment is created at the foot of
the lifting-eye. After the force reached about 216.5 kN, the weld suddenly broke apart. The fracture
plane was almost vertical in all parts of the weld. Before the fracture, extensive yielding took place,
which deformed particularly the lifting-eye.

F Fz Fy z

-Fx Applied load at

x = -197.0 mm;
y = 60.1 mm;
z = 120.5 mm

y
Inner contour of
the Shackle

Weld throat
thickness
a = 6 mm

x
t (40 mm)
A B
c (165 mm)

Fig. 1: Lifting-Eye investigated

2/13
In this paper, the stresses obtained from a simple calculation of the nominal stresses in the weld
are compared with those computed with coarse and fine-mesh finite element models. The fine-
mesh model is also used for a calculation with non-linear material law, offering the possibility to
consider the stress re-distribution which might have occurred just before the failure.

The calculation concentrates on the weld denoted with A-B in Fig. 1, where tensile stresses are
expected. The stresses in the opposite weld, which is under compression, are smaller as the un-
welded root gap closes and carries part of the load. This was taken into account in the fine-mesh
model by introducing contact elements.

The base plate below the lifting-eye was simply supported at its lower surface in all calculations.
This simplification might cause differences from the real stress distribution, however, it does not
affect the present study as only calculated stresses are compared with each other which are all
based on the same assumption.

2 Definition of Stresses in Fillet Welds

According to Eurocode and DIN 18800, three nominal stress components are defined in a longitu-
dinal section through the fillet weld, i. e. σ⊥, τ⊥ and τII, which are considered to be relevant for
failure. Normally, a diagonal section through the weld throat is considered, however, also the verti-
cal or horizontal plane along the weld legs can be considered acc. to DIN 18800, see Fig. 2. To
distinguish between the stress components, an additional index ‘v’ for the vertical plane, ‘d’ for the
diagonal plane and ‘h’ for the horizontal plane has been introduced in Fig. 2. It should be noted that
in all cases the reference plane is restricted to the throat thickness ‘a’ so that in a 90°-fillet weld
the stress components correspond to each other in the following way, if the weld is idealized as a
line-connection at its root:

σ⊥ v + τ⊥v σ⊥h + τ⊥h

σ ⊥ v = τ ⊥h and σ⊥d = = (1)
2 2

τ⊥v − σ⊥ v τ ⊥h − σ ⊥h
τ ⊥ v = σ ⊥h and τ⊥d = = (2)
2 2

τII v = τII d = τII h (3)

τ⊥v
σ⊥v σ⊥d τII d τII h
τII v τ⊥d

a τ⊥h

σ⊥h
a a

Fig 2: Nominal Stresses in Fillet Welds

The possibility to consider different planes can simplify the stress analysis by choosing the plane
where the relevant stresses can easily be defined.

According to this definition, the stress analysis can be performed very simply when the forces and
moments in the fillet-welded plate are known. As illustrated in Fig. 3 for a T-joint, the stresses in

3/13
the vertical section of a double-sided fillet weld are calculated from the distributed forces f and
moments m (per unit length) acting in the plate at the weld root position as follows:

fy fz m fx
σ⊥v = τ⊥v = ± x τII v = (4)
2⋅a 2⋅a t⋅a 2⋅a

When using eq. (1) – (3), the stresses in the diagonal section become

fy fz mx
σ⊥d = + ± (5)
2a ⋅ 2 2a ⋅ 2 t ⋅a⋅ 2

− fy fz mx
τ⊥d = + ± (6)
2a ⋅ 2 2a ⋅ 2 t⋅a⋅ 2

fx
τII d = (7)
2⋅a

It is assumed here that moments about the y- and z-axis result in changing forces fy and fz along
the weld.

dx 40

fz 6
(t
fy
fx
mx 165

Fig. 3: Distributed forces and moments Fig. 4: Idealization of the Weld

transferred by fillet welds for Warping Stress Analysis

3 Analytical Calculation of the Stresses in the Weld

Referring to the footpoint of the lifting-eye (i. e. origin of the coordinate system, see Fig. 1), the
following resultant forces and moments are acting:

Fx = -149,38 kN
Fy = 155,13 kN
Fz = 22,22 kN

Mx = -Fy⋅z + Fz⋅y = -17,36 kNm

My = Fx⋅z - Fz⋅x = -13,62 kNm
Mz = -Fx⋅y + Fy⋅x = -21,58 kNm

The resulting stresses in the fillet weld A-B are calculated in the vertical plane, assuming constant
or linear distribution of the nominal stress components in the weld:

Fy Mz x
σ⊥ v = + ⋅ (8)
2 ⋅ a (c + t ) a ⋅ c ( t + c / 3) c / 2

4/13
 Fz Mx My x Mz
τ⊥ v = − − ⋅ + ω (9)
2 ⋅ a (c + t ) t ⋅ a ⋅ c a ⋅ c (t + c / 3 ) c / 2 λ ⋅ Iω

Fx
τII v = (10)
2 ⋅ a (c + t )

where ω = sectorial coordinate of the welded connection

(ωmax = 1006 mm2 at the corners)
λ = (G ⋅ IT ) /(E ⋅ Iω )
E,G = Young’s modulus and shear modulus
IT = St. Venant torsional rigidity (2.55⋅106 mm4)
Iω = sectorial moment of inertia (830⋅106 mm6)

The last term in eq. (9) considers the warping restraint effect due to the supported plate. The sec-
torial coordinate ω as well as IT and Iω were determined by assuming the welded connection as
a closed, thin-walled section along the weld root line with the thickness a, see Fig. 4. The resulting
warping stresses correspond to shear stresses τ⊥,v in the weld.

In the diagonal section through the weld, the stress components become, using eq. (5) – (7):

Fy + Fz Mx My 2 x Mz  ω 1 2x 
σ⊥d = − − ⋅ +  + ⋅  (11)
2 a (c + t ) 2 t ⋅ c ⋅ a ⋅ 2 a ⋅ c (t + c / 3 ) 2 c 2  λ ⋅ Iω a ⋅ c ( t + c / 3 ) c 

− Fy + Fz Mx My 2 x Mz  ω 1 2x 
τ⊥d = − − ⋅ +  − ⋅  (12)
2 a (c + t ) 2 t⋅c ⋅a⋅ 2 a ⋅ c (t + c / 3 ) 2 c 2  λ ⋅ Iω a ⋅ c ( t + c / 3) c 

Fx
τII d = (13)
2 ⋅ a (c + t )

4 Stress Calculation with a Coarse Finite Element Model

In a first step, the stress calculation is studied at a 2D-model of the welded connection, see Fig. 5.
The fillet weld is modelled by two triangular elements in order to obtain stresses in the diagonal
section through the weld in addition to those in a vertical and horizontal section. The unwelded root
gap has also been modelled, which is part of the section through the model. A vertical force is ap-
plied at the top, which creates a unit stress resultant in the weld throat area.

'1'⋅t
F=
2a

t = 40 mm

a = 6 mm v
d
v,d, h
10 mm h

Fig. 5: Coarse-2D-Mesh of the Fillet Welded Connection

5/13
Isoparametric elements with and alternatively without mid-side nodes have been chosen to model
the plates and welds. As the element stresses along the sides are linearly distributed, if reduced
integration is chosen in case of mid-side nodes, it is sufficient for the force equilibrium to consider
only the average stresses, i. e. the values at the mid-side point.

When looking at the element stresses along the sections shown in Fig. 5, it should be noted that
some stresses are acting also at the faces of the root gap, although the gap is correctly modelled.
This is due to the fundamental principle of the finite element method, that the elements are only
connected at the nodes and that nodal forces are always associated with certain stress distribu-
tions in the adjacent elements. It will be shown later that the stresses at the face of the root gap
have to be included when deriving the nominal stresses in the weld.

The relationship between element stresses and nodal forces is illustrated in Fig. 6 by a distributed
load on the side of a 2D element having the thickness t. The pressure distribution shown in the left
part of Fig. 6 leads to an equivalent nodal force F at only one of the two nodes, if a linear dis-
placement function is assumed. Vice versa, a force at one node leads to a corresponding stress
distribution along the element side. The resulting effective breadth of the element is l / 4 , which
may be utilized when estimating the force at the gap face simply from the stress at the node i.

p0 ⋅ l ⋅ t
p0 F=
-p0/2 4
i i
(t) (t)

Fig. 6: Distributed Load at the Side of a Finite Element and Equivalent Nodal Force

Taking instead the average stresses at the element sides above the sections shown in Figs. 5 and
7, the nominal stresses in the weld can be derived from equilibrium conditions as follows:

Vertical section (v-v):

σ ⊥ v = σ1 ⋅ 2 − τ 2 ⋅ l / a (14)

τ ⊥ v = τ1 ⋅ 2 + σ 2 ⋅ l / a (15)

Diagonal section (d-d):

σ ⊥ d = σ1 − τ 2 ⋅ l (a 2 ) + σ 2 ⋅ l (a 2 ) (16)

τ ⊥ d = τ1 + τ 2 ⋅ l (a 2 ) + σ 2 ⋅ l (a 2 ) (17)

Horizontal section (h-h):

σ ⊥ h = σ1 ⋅ 2 + σ 2 ⋅ l / a (18)

τ ⊥ h = − τ1 ⋅ 2 − τ 2 ⋅ l / a (19)

l = width of the adjacent element at root gap. In the case that only one element is arranged
over the thickness t, l has to be set to (t/2) because the element stress is distributed to
both fillet welds.

Similar equations can be established also for the elements below the sections.

6/13
 Vertical Section (v-v): Diagonal Section (d-d): Horizontal Section (h-h):

l l l

Fillet
weld
a 2 τ1 a 2 τ1 a 2
σ1
τ2 τ⊥v τ2 σ1 τ2 τ1
τ⊥d
σ2 σ2 σ⊥d σ2 σ1
σ⊥v
Fillet weld Fillet weld σ⊥h τ⊥h

Fig. 7: Calculation of Weld Stresses from Stresses in Elements

above Different Sections through the Weld

Table 1 shows the stresses calculated in this way in comparison with the analytical results. Several
of the weld stresses evaluated acc to eq. (14) - (19) are not in good agreement with the stresses
calculated analytically. The additional results in brackets, considering only the stresses σ1 and τ1
directly at the weld element, are even worse. Many values are below 50 % of the analytical ones.
This clearly shows that at least in coarse meshes the stresses of the adjacent elements need to be
included.

Table 1: Nominal stresses in the weld of the 2D-model

Calculation Method Vertical Section Diagonal Section Horiz. Section

v-v d-d h-h
(stress ref. to a) (stress ref. to a)
Analytical results σ⊥ 0 0.707 1
from eq. (14) - (19) τ⊥ 1 0.707 0

from nodal σ⊥ 0.4174 1.0022 1.0000

forces τ⊥ 1.0000 0.4120 0.4174
4-node
elements σ⊥ 0.9976 (0.4342)* 1.5732 (0.7025)* 1.0145 (0.3465)*
from element
stresses τ⊥ 1.0008 (0.3328)* 0.3865 (0.3125)* 0.7683 (0.2049)*

from nodal σ⊥ 0.4723 1.0410 1.0000

8-node
forces τ⊥ 1.0000 0.3732 0.4723
elements
(reduced σ⊥ 0.7816 (0.4115)* 0.8102 (0.2702)* 0.9421 (0.5486)*
from element
integrat.)
stresses τ⊥ 1.1400 (0.7463)* 0.2748 (0.2583)* 0.5573 (0.1872)*
*) if only the stresses σ1 and τ1 directly at the weld element are considered

Alternatively, the stresses can be evaluated from the nodal forces in one of the three sections.
These give the best results because nodal forces satisfy equilibrium conditions in finite element
models. However, a post-processing program might be necessary for their evaluation. From a
comparison of the stresses obtained from the analytical equations with those evaluated from nodal
forces (Table 1), it can be seen that additional stresses act in the weld which are caused by the
restraining of deformations in transverse direction. This means that also the analytical results con-
tain simplifications.

7/13
In the second step, the lifting-eye was modelled with a coarse mesh, see Fig. 8, which is even
somewhat coarser than the 2D-model. The shackle was modelled by truss elements between the
lifting-eye and the load introduction point (Fig. 1). 8-node isoparametric elements were used, cor-
responding to the 4-node elements in the 2D-model. The weld stresses were calculated in the ver-
tical section from the forces at each pair of nodes and afterwards transferred to the diagonal sec-
tion as well as referred to the weld throat thickness a.

The results are shown in Fig. 9 (FE coarse) in comparison with the analytical results and with fine-
mesh results which will be described below. The calculated stresses show a similar course as the
analytical ones, however, the peaks at the ends of the weld calculated analytically are obviously
not really existing.

The differences in τII indicate that the torsional moment Mz creates also this type of shear stress
in the fillet welds in addition to σ⊥ and τ⊥.

Fig. 8: Coarse Finite Element Model of the Lifting-Eye

5 Stress Calculation with a Fine-Mesh Finite Element Model

In addition, a fine-mesh model of the lifting-eye and the shackle was created, see Fig. 10. The
same element type was used as for the coarse-mesh model. As mentioned before, contact ele-
ments were included to realistically model the contact under compressive stresses in the root gap.
Contact does not occur along the weld A-B in Fig. 1, so that the results are still comparable.

In a first step, linear-elastic material behaviour was assumed. The fine mesh yields relatively high
stress peaks in the welded connection. Fig. 11a shows the distribution of equivalent von Mises
stresses in the most highly stresses region of the weld (at x = -55 mm). The highest equivalent
stress of 1719 MPa in the upper area certainly depends on the mesh density chosen and might
even be higher in finer meshes.

For comparison with the above calculations, an averaged stress has been evaluated in the diago-
nal section, taking again the stresses in the adjacent element at the root gap into account, as
shown in the middle of Fig. 7. The results are included in Fig. 9 (FE fine, linear). The differences to
the results from the coarse mesh are partly relatively large. Reasons for this are seen in a different
load transfer in the finer mesh as well as in effects coming from the contact in the root gap.

8/13
 1000
900 analytical
800 FE (coarse)
FE (fine; linear)
700
FE (fine; nonlinear)
600
500
400
300
200
100
0
-100
-200
-300
-100 -80 -60 -40 -20 0 20 40 60 80 100
x [mm]
600
analytical
500
FE (coarse)
400 FE (fine; linear)
FE (fine; nonlinear)
300

200

100

0
-100 -80 -60 -40 -20 0 20 40 60 80 100
x [mm]
200

100

-100

-200
-100 -80 -60 -40 -20 0 20 40 60 80 100
x [mm]

Fig. 9: Nominal Stress Components in the Diagonal Section through the Weld
(for ‘FE coarse’ derived from nodal forces and for ‘FE fine’ from
element stresses including adjacent element at root gap)

9/13
 Fig. 10: Fine-Mesh Finite Element Model of the Lifting-Eye and Shackle

Fig. 11: Distribution of Equivalent von Mises Stresses in the Most Highly-Stressed Part
of the Weld: a) Linear Analysis and b) Nonlinear Analysis

Neglecting the stresses in the adjacent element at the root gap would lead to a substantial under-
estimation of the weld stresses: the maximum value of σ⊥ would be 540 MPa (instead of
667 MPa) and that of τ⊥ would be 192 MPa (instead of 250 MPa). This underlines the necessity
to include the stresses in the gap element in the calculation of the nominal weld stresses, except
for very fine mesh densities.

In a second step, a nonlinear material law was introduced, because the elastic limit is far exceed-
ed. Fig. 12 shows the material law assumed for both the base and weld material.

10/13
 600

500

Stress σ [MPa]
400

300

200

100

0
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%

Strain ε

Fig. 12: Assumed Material Law for the Nonlinear Analysis

The analysis shows extensive yielding in major parts of the weld as well as in parts of the lifting-
eye. Therefore, the stress distribution changes completely. In the same way as for the linear analy-
sis, average stresses have been evaluated in the diagonal section. The resulting nominal stress
components in the weld are included in Fig. 9 (FE fine, nonlinear), showing the re-distribution of
stresses when reaching the ultimate limit state.

Fig. 11b shows the distribution of nonlinear equivalent stresses in the section at x = -55 mm again.
The comparison with Fig. 11a demonstrates the effects of yielding.

6 Stress Assessment
The stresses can - as mentioned in the Introduction - be assessed on the basis of an equivalent
stress. Table 2 gives the highest values of the equivalent stress computed from the averaged nom-
inal stress components in the weld throat, i. e. the equivalent von Mises stress as prescribed in
Eurocode 3

σeq = σ2⊥ + 3 ⋅ τ2⊥ + 3 ⋅ τII2 (20)

and the vector sum of the three components as defined in DIN 18800

σ v = σ 2⊥ + τ 2⊥ + τII2 (21)

Table 2: Equivalent stresses in the most-highly stresses part of the weld

calculated from averaged nominal stress components in weld throat

Type of analysis Equivalent von Vector sum of

Mises stress stress components
Analytical 1350 MPa 1104 MPa
FE (coarse, linear) 982 MPa 828 MPa
FE (fine, linear) 794 MPa 714 MPa
FE (fine, nonlinear) 505 MPa 480 MPa

Considerable differences can be seen between the three linear analyses. The analytical analysis
over-estimates the stresses particularly at the ends of the weld, which would result in a conserva-

11/13
tive weld design in the example. The finite element method gives a more realistic distribution of the
stresses and, hence, a more economic design. On the other hand, the nonlinear analysis shows a
complete re-distribution of stresses. This occurs particularly in welded connection with an uneven
distribution of linear stresses. How far the associated plastic reserve can be utilized can only be
verified by tests. In this connection it should be kept in mind that the stress analysis might still con-
tain uncertainties, e. g. a changed load distribution due to the supporting structure. Furthermore,
weld imperfections may affect the strength.

7 Conclusions
From the stress calculations for the example of a fillet welded lifting-eye, the following conclusions
are drawn:

1. The nominal stress components σ⊥, τ⊥ and τII in the fillet weld, which are relevant for the
static strength acc. to several codes (e. g. Eurocode, DIN 18800), are usually calculated by rel-
atively simple equations (force per sectional area or moment per section modulus). In addition
to the diagonal section through the weld, also the sections parallel to the adjacent plates may
be considered in case of 90°-fillet welds, if the welds are idealized as a line connection and the
stresses are referred to the throat thickness a. The stress components can easily be trans-
ferred into stresses in the diagonal section.

2. Assumptions have to be made particularly for the stress distribution due to a torsional moment,
which can create stresses acting transverse or parallel to the weld. Large additional stresses
may be due to warping restraint, as the example has shown.

3. Simple finite element models with plate or shell elements, which are directly connected at the
welded joint, yield sectional line forces and moments which can be transformed into the nomi-
nal stress components in the weld.

4. Coarse 3D finite element meshes, where the fillet weld is represented by only one or two solid
elements in the cross section, yield results which may under-estimate the stresses in the weld
by more than 50 %, if these are derived from the stresses only in the weld elements or in the
opposite elements in the plates. This is due the fact that the nodal forces between the elements
create also stresses in the adjacent elements at the root gap which have to be considered
when calculating the averaged weld stresses. Even in this case, the differences can still be rel-
atively large, as the example has shown. Only reliable are stresses derived from nodal forces
acting in a section through the weld and referred - as usual - to the throat thickness.

5. Fine-mesh models show high stress peaks mainly close to the weld toe and probably weld root.
For a static strength assessment, the nominal stress components in the weld can again be de-
rived by averaging the element stresses. The diagonal section is recommended to avoid a sec-
tion through high stresses at the weld toe. However, the stresses in the adjacent elements at
the root gap should be included also in fine-mesh models. Their neglect has resulted in the ex-
ample, where three elements were arranged over the throat thickness, in an under-estimation
by more than 20 %.

6. Before reaching the ultimate limit state, yielding of the weld and the adjacent base material re-
distributes the stresses, as shown by the example. The stress peaks found in the linear analy-
sis of a fine 3D finite element mesh are no more existing. This justifies the relatively simple
analysis of averaged stresses in the weld throat.

7. The equivalent stresses calculated in accordance with the codes show large differences be-
tween the analyses. Particularly the analytical formulae yield high stresses at the ends of the
weld under in-plane bending and torsional loads, which would result in a conservative design.
Here, the finite element analysis gives more realistic values, as far as the stresses are aver-
aged through the weld throat. A further stress re-distribution along the welds due to yielding
can increase the load-carrying capacity considerably, however its utilization in design is ques-
12/13
 tionable due to uncertainties related to modelling aspects as well as possible weld imperfec-
tions.

8 References
Eurocode 3 (1994): Design of Steel Structures, Part 1-1 General Rules and Rules for Buildings
(prEN 1993-1-1). European Committee for Standardisation, Brussels.

Dietrich, R.A. (1990): Finite Element Analyses for Stress Assessment of a Welded T-Joint with
Double Fillet Weld and Undercuts under Bending Load (in German). Schweißen und Schneiden
42, pp. 279-284.

DIN 18800 (1990): Steel Structures - Part 1; Design and Construction (in German). German
Standard Organisation.

13/13

View publication stats