Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/293488935
CITATION READS
1 561
4 authors, including:
Christian Gaier
Engineering Center Steyr GmbH & Co KG, Magna Powertrain
51 PUBLICATIONS 134 CITATIONS
SEE PROFILE
All content following this page was uploaded by Christian Gaier on 08 February 2016.
For lightweight automotive structures, the stiffness and the fatigue behaviour is greatly
influenced by the properties of its joints. The used joining technology, the number and
locations of the joints are of high importance for both engineers and cost accountants.
An overview of common computational procedures including European and national
standards is given for the assessment of the fatigue behaviour of thin sheet structures
with arc welds. The influence of the quality and the size of finite shell elements on the
fatigue result is investigated and it is shown, how this influence can be minimized.
INTRODUCTION
The reduction of the vehicle weight is at least since the drastic raise of fuel prices in the
seventies a basic requirement for the development of new cars, trucks and busses.
Nevertheless there is the general tendency in the opposite direction. The reasons for this
contradiction are on the one hand additional features for more comfort and safety as
well as increased motor torque. On the other hand the potential for mass savings in the
basic structure is not enormous.
Vehicles have to withstand dynamical loads and their components have lots of arc
welds, spot joints and notches. Several technological treatments with positive or
negative influence to the lifetime usually are applied. The stiffness and the operating
strength of sheet metal parts of bodies is strongly influenced by the properties of their
joints. The used joining technology, the design and the position of the welding seams, as
well as the number and the location of the spot joints have a significant technical and
economical impact. This becomes clear from the fact that in fatigue testing of car bodies
more than 90% of the cracks originate from joints [1,2], which is mainly due to the
considerably lower dynamic loading capability of welding seams and spot joints
compared to the base material. Therefore, especially for lightweight structures, the
joining technology plays a key role. An optimum use of the material capabilities is only
possible with optimized weld or spot joint design of a structure. This only becomes
practicable by assessing the joints and welding seams in an early phase of the
development process.
How can engineers handle all important fatigue influences during the development
process? On the one hand, a very good method is to perform fatigue tests with
components or total vehicles. Many influences can be considered without any
mathematical model, nevertheless the results have absolute accuracy. Unfortunately,
experimental tests give only poor information about low stressed areas, so it is very
expensive and time consuming to minimize a component’s mass based on experiments.
Therefore, computational lifetime simulations, mainly based on the finite element
method FEM, are state of the art to reduce or even avoid testing loops, prototyping and
therefore development time and costs (“virtual prototyping”). Nevertheless, release tests
are still unavoidable before serial production starts.
• There is no usable standard for the fatigue assessment of welding seams for a
general FE-model.
• Different models for each welding seam lead to different stress results.
• Adjoined shell elements in two planes normal to each other are often used to
simulate welding seams. They can have completely different stress states. When
post-processing e.g. a T-joint modeled with shell elements, the stresses of all
planes are averaged at the nodes leading to averaged stresses which may be as
low as half of the stress in one plane. Therefore, stresses at the welding seam
nodes may not be appropriate for fatigue evaluation.
• Usually global coordinates are used to evaluate stresses. For welding seams,
however, the stress components crosswise and along the welding seam are of
importance.
• A high-quality assessment of the FE-results for the whole structure with respect
to fatigue life requires an extensive amount of time when using the classical,
manual methods.
At the Engineering Center Steyr, concepts for fatigue life analysis have been developed
during the last two decades, which give a solution to the problems described above. The
concepts have been coded in the user-friendly software FEMFAT [3], allowing a fast,
efficient and high-quality assessment of the fatigue life of FE structures with or without
welding seams and spot joints.
In this paragraph some commonly used methods for fatigue assessment of welding
seams will be presented, which were implemented in FEMFAT.
EUROCODE 3
Eurocode 3 [4] presents a general method for the fatigue assessment of structures and
structural elements made of steel which are subjected to repeated fluctuations of
stresses. Assessment of parent material and welds was treated. The main provisions of
Eurocode 3 rely upon a set of fatigue resistance curves (S/N curves), equally spaced,
upon which a set of constructional details are classified. The S/N curves have a constant
slope of m=3 (for normal stress), or m=5 (for shear stress, hollow section joints and
some particular details). The slope of m=3 changes at 5 106 load cycles to m=5 and
stress amplitudes below a cut off limit at 108 load cycles will be neglected.
The assessment could be done using either nominal stresses or according to the
geometric stress concentration concept (also called the “hot spot stress” method). For
weld seams the stress components related to the seam direction will be used to
constitute an equivalent stress σw and τw respectively.
σ w = σ ⊥2 + τ ⊥2 Equation 1
τw=τ|| Equation 2
For multi-axial loading normal stress and shear stress should be assessed separately
using linear damage accumulation according to Palmgren/Miner in combination with
one or two slope S/N curves.
The classification of load histories should be done using rain flow or reservoir counting
method.
The fatigue strength could be modified due to the influence of sheet thickness and/or by
a 40% reduction of the compressive part of the stress amplitude at parent material or
annealed welding seams. No further mean stress influence will be considered in
Eurocode 3.
The British Standard 7608 [5] is very similar to Eurocode 3. Again constructional
details were divided into several classes with corresponding S/N curves. In difference to
Eurocode 3 the S/N curves are not equally spaced and the slopes varies from m=3 over
3.5 and 4 to 8. At 107 load cycles the slope changes from m to m+2. No cut off at low
stress amplitudes is allowed if only one cycle is above the cut off limit. There are no
shear stress S/N-curves defined in BS 7608.
The classification of load histories should be done using reservoir counting method.
As well as at Eurocode 3 the fatigue strength could be modified too due to the sheet
thickness influence and by the reduction of compressive stress amplitudes of 40%. No
further mean stress influence will be considered in BS 7608.
The German crane manufacturing standard DIN 15018 [6] defines permissible fatigue
stresses σD,perm(R=-1) for St 37 and St 53-3 in dependency from notch categories and
loading classes. Mean stress influence is considered using the stress ratio R=σl/σu of
lower stress σl to upper stress σu. Thus one can obtain the permissible fatigue stress
σD,perm(R) for a certain notch category, loading class and mean stress (R value
respectively).
The verification of the fatigue strength will be done proofing the following relation (4)
which is similar to v.Mises stress:
2 2 2 2
⎛ σx ⎞ ⎛ σy ⎞ ⎛ σ x ⋅σ y ⎞ ⎛ τ ⎞
⎜ ⎟ +⎜ ⎟ −⎜ ⎟ +⎜ ⎟ ≤ 1,1 Equation 4
⎜σ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ xD, perm ⎠ ⎝ σ yD, perm ⎠ ⎝ σ xD, perm ⋅ σ yD, perm ⎠ ⎝ τ D, perm ⎠
DS 952
In the German rail vehicle standard DS 952 [7] diagrams are given, where permissible
upper normal or shear stress is plotted versus stress ratio R. In each diagram different
curves are given for several notch categories. Thus again one can obtain the fatigue
strength in dependency from a notch category and the mean stress (R value
respectively).
These diagrams are given for the welds of the following materials: St 37, St 52-3,
AlMg3, AlMgMn, AlMg4.5Mn, AlMgSi1, AlZnMg1.
The highest stress in the welding seam σmax or τmax should be determined and this
maximum stress must be smaller than the permissible values coming from the diagrams.
DS 952 and DIN 15018 demonstrate similar mean stress sensitivity at low mean
stresses, but highly variable assessments at high mean stresses. The DS 952 rail vehicle
standard evaluates high mean stresses very conservatively, in contrast to the DIN 15018
crane manufacturing standard.
Radaj, Koettgen, Olivier and Seeger developed an approach which allows the prediction
of mean value and spread width of the dynamic strength limits of welding seams (see
[8]). During 41 test series (about 400 single tests) for different joint geometries, load
configurations and stress relations component S/N curves were determined. All
specimens were annealed.
Figure 2 shows the notch stress fatigue limits calculated from the experimentally
determined component strength using the above described models with r=1mm. In spite
of the different joint geometries a uniform value of the notch stress fatigue limit with
little scatter results. This conclusion is valid both for a constant stress ratio R=-1 and
R=0. The values given in Figure 2 were verified and confirmed for other joint types
using the S/N curve catalog of the DVS (see [9]). Thus it can be stated, that for any
weld joint geometry loaded with normal stress the fatigue strength of root and toe can
be predicted using the described model.
Figure 2: Survival probability [%] against fatigue strength [N/mm2]
PROBLEM DESCRIPTION
The following questions and difficulties arise in connection with fatigue assessment
based on Finite Elements:
• How can essential parameters of a welding seam like the type of the seam, the
orientation of the seam, the seam ends etc. be recognized in the FE structure?
• How can singularities be handled? They appear for example at the ends of
welding seams and are strongly influenced by the local mesh discretization.
• Fatigue assessment of the local welding seam region including relevant welding
seam notch factors at the root, the seam ends or at seam crossings
• Openness for any extension with respect to joint shape and welding seam
configuration.
To make it easily applicable, the welding seam assessment concept has been integrated
in a fatigue life software package [3]. It is based on four key issues:
3. Recognition and classification of all local welding seam areas in the whole
structure.
The Finite Element welding seam modeling guidelines were developed to instruct a FE
user how to realize welding seams within a complete structural model so that they can
be assessed in an automated manner. The guidelines mainly give rules for setting
attributes to nodes and elements representing the local welding seam. Different
attributes, like material labels and node colors serve to distinguish the orientation, the
geometry and the type of welding seam. An open scheme permits the extension of the
guidelines to new, not yet included types of seam shapes or welding seam
configurations. The welding seam itself is not modeled. Instead, the seam sheet metals
are represented by shell elements. In the following the shell elements which border on
the welding seam are called “welding seam elements” or simply “seam elements”. The
size of these elements may vary widely, as only the basic stresses (“nominal stresses”)
will be computed in the seam elements. A possibility to automatically correct the
stresses based on the characteristic of the stress curve and the element size is supplied.
The contents of the seam elements’ material tables as well as their thickness are the
same as those of the associated sheet metal.
Figure 3: FEMFAT WELD Fatigue assessment procedure
The ends of the welding seams are especially weak points. The reason for this fact are
the additional notches due to the end of the welding seam. Besides, there is usually an
additional negative effect due to Finite Element stress singularities. To disregard such
unrealistically high stresses, the following measures are taken:
1. The definition of the welding seam ending/starting (e.g. node color 101) is
applied to the end of the welding seam in FE-model.
The additional effort in model preparation is limited to associating certain node and
element attributes to the welding seams
Based on a concept presented by Radaj [11], an approach was developed in [8] which
allows to derive the dynamic loading capabilities of welding seams of arbitrary type and
geometry from the welding seam notch fatigue strength. In a first step, for structures
with almost no residual stresses, the structural S/N curves were determined for different
joint geometries, load configurations and plate thickness (8 to 80mm) by 41 test series.
In the next step, for every test series, a numerical Boundary Element simulation model
was established to compute the linear notch stresses respectively the stress
concentration factors at the welding seam locations endangered to crack initiation. A
radius of 1mm was used to round the seam roots and the seam transitions to the
surrounding basic material. Else, the simulation model was based on the mean values of
the properties of the respective test series. Using the notch factors determined by means
of simulation and the nominal fatigue strength in the crack-sensitive cross-sections, a
fictitious seam notch fatigue strength was calculated. For the different test series, for a
constant stress ratio R a unique value of the seam notch fatigue strength has been found.
Therefore, the fictitious seam notch fatigue strength may be used in combination with
FE submodels together with a constant replacement notch radius for the prediction of
the fatigue life of complex welding seams.
The stress state of the welding seam is determined from FE-results of the stress state in
the seam elements using local seam coordinates. The notch factors for a computation of
the comparable fictitious seam notch fatigue strength are filed in a welding seam
database, covering most widely used seam types. This database has been made for the
sake of efficiency, as the costly process of creating and analyzing sub-models with
respect to S/N curves and local notch factors is done only once. Depending on the type
of welding seam, notch factors for at most 3 load cases at 4 assessment locations per
sheet connected to the welding seam are stored. These are, if existing, the heat affected
zone notches (undercuts) at the transition and the root region for the top or bottom side
of the sheet panel. The notch factors are associated with the stress normal and parallel to
the welding seam in the plane of the sheet panel. The S/N curves used for linear damage
accumulation are adapted to the actual loading case by interpolation [3].
The finite element models of welded components like car and truck frames, body in
white structures, rail vehicle bogies etc. were mostly built up with thin shell elements.
For assessment of welding seams commonly the stresses of the elements adjacent to the
seam will be used. In this case the used stresses are depending strongly on the size of
the elements. Figure 5 shows the distribution of the normal stresses perpendicular to the
seam line of an overlap joint for 4 different element sizes (10x10mm, 5x5mm,
2.5x2.5mm and 1.25x1.25mm). On one hand the peak stress at the joint depends on the
element size, on the other hand in a certain distance from the joint the stress is nearly
the same for all mesh densities.
Figure 5: Stress distribution of overlap joint for different element sizes
The fatigue software FEMFAT provides an automatic stress correction feature for
welding seams. The users have the possibility to define a certain distance from the joint
line. As a default the minimum sheet thickness is used, which is approx. the position of
the weld toe between welding seam and heat affected zone. In reality, fatigue crack
initiates at this position or at the weld root, for which a fatigue analysis is performed
too. FEMFAT interpolates the assessment stresses from the surrounding elements (see
Figure 6).
The stresses of the elements marked with X in Figure 6 are used for evaluation of the
current weld element. The weighted average stress components are calculated using the
following relationship:
n σ i, j
∑i =1 di
σj = n
Equation 5
1
∑
i =1 d i
σj … stress component for the evaluation of the current weld element (transverse,
longitudinal and shear component).
σi,j … j stress components (transverse, longitudinal and shear components with regard to
the local weld coordinate system) of the i elements at the assessment point.
n … number of elements utilized for the evaluation of the current weld element at the
assessment point.
Figure 6: Automatic stress correction of FEMFAT WELD
This procedure works for both linear and quadratic shell elements as well as for weld
seam start/end. Thus FEMFAT provides the possibility to use a constant assessment
distance from the joint independently from the element size.
Another possibility to reduce the element size influence is to use a grid point force
dependent approach as e.g. the equilibrium-equivalent structural stress approach
according to Dong et al. [12].
EXAMPLES
A multi axial fatigue simulation of a closed rear axle subframe was performed using the
MAX module of FEMFAT. Seven load channels with measured load time histories
were used. Nonlinear bearing behaviour was considered by separating the loading in
positive and negative direction. The position of the weldings were defined within the FE
model using node colors or displacement coordinate system labels. The seam type was
determined using particular element material labels according to modelling guidelines.
With this FE model linear stress analyses were performed for each unit load case. In
Figure 8 the maximum Mises stress values of the dominant channel (loading in
longitudinal direction) are shown.
Figure 8: Maximum v.Mises stresses of a rear axle subframe (critical load case)
The stress results of the unit load cases were combined using the measured load time
histories for evaluation of the fatigue life. Parent material and weldings were assessed
simultaneously in one computation run. The structural stresses at the welded joints
coming from the stress analyses were extrapolated to notch stresses by using notch
factors from a database. In this database different notch factors for many joint types
were stored in addition to S/N curve data and Haigh diagrams. The notch factors were
obtained either from submodel analysis or experimental results.
Figure 9 shows the damage values (inverse of life) of one multi axial loading cycle. In
spite of high stresses in larger areas the damage results are focusing on the weld joints.
The benefits of this method are:
• No care must be taken of the detailed geometry of the welding seam in complex
structures.
• Independence from the mesh density using the automatic stress correction.
• Open database (adjustable by the user) for FE weld seam notch factors , S/N
curves and Haigh diagrams.
Figure 9: Damage distribution for multi-axial loading for one load cycle
TRUCK CABIN
A fatigue life analysis of a multiaxially dynamically loaded truck cabin were performed
with FEMFAT MAX. The FE model was generated based on the CAD model of the
cabin including spot and arc weldings (a total of 900.000 nodes). The arc welding seams
were modeled according to the modeling guideline of FEMFAT WELD.
As boundary condition unit loads were applied at both front bearing points in
longitudinal, vertical and transverse and at both rear bearing points in vertical and
transverse direction. For these 10 unit load cases linear inertia relief stress analysis were
performed, where the external forces were set in equilibrium with inertial forces. The
linear stresses provided this way and the gravity loadings were scaled and superposed
by load history data of a rough road test. These superposed stresses were used for a
fatigue analysis of the welding seams. In Figure 10 the damage distribution of a critical
area of the truck cabin can be seen at the lower corner of the front windshield.
Figure 10: Critical area with welding seam at the corner of the windshield
The high damage at the welding seam were confirmed by experiment (see Figure 11).
The crack started at the weld seam and continued in the base material. Based on the
fatigue analysis and on the experimental results the design of the cabin was improved.
CONCLUSION
An overview has been given over some common methods and standards for the
assessment of welding seams. The starting point are FE models which are easily
adapted according to guidelines for welding seams. The fatigue assessment is based on
notch factors and S/N curves stored in databases. Generally, the elements surrounding
the actual joint are used to assess the welding seam. It has been shown, how the
influence of the mesh quality and the element size can be minimized. The realization of
the concepts within the user-friendly, reliable fatigue software FEMFAT allows a fast
and efficient assessment of the operating strength of such structures. It is suitable for
structures which are pre-optimized with respect to fatigue in an early phase of the
design process. They require less tests leading to shorter development phases, higher
quality and therefore less costs.
REFERENCES
3. FEMFAT, Version 4.5 (2004), Users Manual, Engineering Center Steyr, St.
Valentin, Austria.
4. EUROCODE 3 : Design of steel structures . Part 1.1: General rules and rules for
buildings. ENV 1993-1-1.
5. British Standard 7608 (1993), “Fatigue design and assessment of steel structures”.
6. DIN 15018 Part 1 (1984), “Cranes; principles for steel structures, stress analysis”.
7. DS 952, Germany.