PROCEEDINGS OF THE 3” INTERNATIONAL GONFERENCE
POWER TRANSMISSIONS ‘99
1-2 GergueR 2909, CHALKIOIKI, GREECE
Bot A. MINAILIDIB
FATIGUE BEHAVIOUR OF WELDED AXLE CASINGS PROVIDING
WELD END FAILURE,
Michail MALIKOUTSAKIS, Alexandros SAVAIDIS, Christos SALPISTIS, Georgios SAVAIDIS
Abstract: Structures such as axles of motor trucks are sensitive to fatigue loading due fo their construction and loading
conditions. The welded components that form the mideseries rear axle under investigation are categorized as thin.
‘walled, with structure thicknesses 1.5<1<25mm. Ths fact makes thelr influence dominant for the durability of the struc
ture in addition tothe operational loading conditions, Thus, a complete and reliable experimental investigation should
include the assessment of the basic fatigue life curves (Wahler curves) as well as data which correspond to sequences
simulating the varlable loading conditions (Gafner curves). This second step is essential in order to verify and enhance
the fatigue life prediction accuracy of conclusions reached by the use of the frst step. This paper focuses to the deter~
rmination of the curves mentioned for characterste load cases that occur to such structures. Wohler curves for three
ferent types of loading (vertical, ongtudinal and torsional loading) are presented. Firs fatigue life data at variable
amplitude loading are also obtained and assessed. n addition, the stress Behaviour in the weld regions under the above
mentioned loading condition is determined based on strain measurements by means of strain gages «and allowable cy-
lie stress values are derived forthe design of such components. The experimental data base achieved can be used for
the verification of calculation models fr the fatigue assessment of welded structures.
Keywords: Fatigue, Welds, Axles, Variable amplitude loading
1, INTRODUCTION
In general, an experimental investigation of fatigue life of
‘components should take into consideration numerous fac-
tots which influence the service life of structures. Mate-
rial properties, geometry and service conditions are usu
ally explicitly determined when dealing with industrial
products,
Fig. 1. Course of experimental fatigue assessment
‘The major issue is to determine the service loads applied
to the component within a realistic framework. In order to
achieve a reliable and yet practical investigation with
‘manageable results the multiaxial service loads should be
broken-down to their nominal components (tension, bend:
ing and torsion}.
The first stop is to determine the basic fatigue life curves
(Wahler curves) of each load component. ‘This is obvi-
ously not adequate to specify a realistic fatigue life for
structures that their service loads are a mixture of all
‘nominal load components and amplioudes. Thus, as a sec-
fond step, a complete and reliable investigation should
include eurves which correspond to sequences simulating
the operational loading conditions for each load compo-
nent (Gafiner curves),
‘The two steps mentioned above enable the determination
of an appropriate damage accumulation law which can be
used to calculate the fatigue life for any given load se-
«quence (measured on test tracks or given in design codes).
2, WELDED COMPONENTS
Special care is essential when dealing with welded com-
ponents of thicknesses t<2Smm, where the notch effect
caused by the weldment geometry is dominant for the
durability of the structure (a weldment with nominal di-
mensions of 4xdmm or greater, which is common for such
structures, represents a 16% thickness enlargement, form=
ing also a stiffer area of significant size). Even though
design codes and international recommendations [1] give
instructions and calculation methods for simple geome-
323tries, the complexity of many components leads to a man-
datory experimental investigation of durability
‘The majority Of welded components fail at areas where
‘welds start or end [1]. This fact is caused by the apparent
complicated 3D geometry, which brings out a high stess
concentration in all directions. The welding. parameters
also have a significant role as this area is difficult to
‘maintain the uniformity of the weldment in terms of
thermal processing and hence residual stresses,
Simple versus complex geometries
Common welds have been categorized by type and ge-
cmety, facilitating the durability assessment of simple
components under uniaxial loading. Nevertheless, when
more complex geometries occur, the international guide.
Lines give only limited theoretical assistance, resorting to
experimental investigations.
Fig. 2. Examples of simple welds
Specimens under investigation
The specimens used are axle casings of mid-series motor
fucks with peripheral components welded to the main
bart. These casings are categorized as thinwalled with
Structure thicknesses t<25mm, The material used is a
‘common low-strength steel welded via fillet welds
Fig. 3. Overview of the axle casing under investigation,
‘assembled on the chassis of the truck
‘The welded components are the hats which form the area
ofthe exle-leaf spring connection and the brake flange. A
closer look at these areas reveals the complexity of the
{eometty as well es the variations of the weld-end topol.
ony.
(Leaf spring hats (0) Brake ange
Fig. 4. Weld geometries and typical variations
324
3. EXPERIMENTAL PROCEDURE,
Test rig
‘The specifications of the manufacturer dictate the simul
tion ofthe whoie structure, a requirement which is met k
the test rig construction shovn in Fig. 5. This enables ty
study of the axle easing's behaviour a5 close to servin
conditions as possible. The experimental procedure 1
conducted by assembling the whole axle tothe test i i
contradiction to just extracting specimens from the’ as
sumed critical areas. This is of decisive importance be
cause the stiffness of the structure is kept intact and all
assumptions concerning the translation of operational
loads to their experimental equivatent loads do not affect
the resultant fatigue life,
Fig. 5. Overview of the test rig
The test rig shown above provides the same support types
as the actual assembly on the chassis ofthe truck (Fig 3).
Furthermore the londing is introduced via an actuator ot
the same point where the whecl ofthe trick is fastened, in
cases of simulating the vertical and horizontal compo.
nents of losding, and at the point where the braking sys-
tom is assembled, in case of braking moment,
Load, displacement and stress monitoring
The load measured on the load cell and the displacement
of the actuator are always monitored, Strain gages applied
{0 42 points monitor the stress behaviour of the axle case
ing. The critical ones have been identified through a series
of preliminary measurements for all lod components,
‘These critical points where then monitored throughout the
whole procedure. The stress and load histories gathered,
allow the fatigue assessment in terms of both, nominal
loud andi local stress values,
All measurements showed a quasilinear correlation be-
tween force, displacement and stresses. In addition, a
clear image of the critical areas and the most appropriate
Positions for creating reference points that would be es.
sential for further analysis has been obtained,
In onder to verify the repeatability of the experiments,
‘measurement points far away fiom stress concentration
areas where also monitored. This action was essential due
to the light differences in stress concentration factors ap-
pearing for the geometrical variations of the weld ends —
failure points
Failure points ~ Failure Criteria
In case of load type I and 3 (vertical load and braking
‘moment, respectively) failure occurs on the inside rearcend of the brake flange. Even though the faiure area
2 same, load ease 1 develops a weld toc failure
ne, in contrast to load case 3, where the crack initi-
rom the weld root
(load ype -weldioe——(P) Load ppe 3 weld root
Fig. 6. Failure points ~ Load types 1 andl 3
+ fact commands also the use of different failure erite-
For load type 1, the crack length of 2mm, which is
ly detectable via visual inspection, was chosen as a
istic value for failure criteria. A very rapid propaga-
‘of the erack afler this length has been observed, lead-
to total fracture.
the other hand, load type 3 failure initiates at the very
‘ eycle area forming an initial crack of 2mm. This
‘tops having no influence on the stiffness of the
tcture, Later on a secondary erack path is formed di-
lly on the weldment, which propagates slowly. A loss
siiffness able to be measured is observed only after
ck lengths of approximately 72 to 75 mm, a value de:
ed to be the failure criteria
ad type 2 (longitudinal.oading) leeds to feilure at the
tside weld end area of the spring hats, an area where
0 different weldment meet.
Fig. 7. Failure point — Load type 2
he behaviour of the crack initiation and propagation is
miler to the one observed at load type 1. Therefore, a
etectable crack length of 2mm is also decided as failure
teria while the transgression of this length also leads to
otal separation, The evitical weld toe, ie. the one where
se crack initiates, is that of the upper spring hat, an area
vhich is closer to the actuator than the lower spring hat,
5 itcan clearly be seen,
|. TEST RESULTS
The fatigue life results are plotted against both nominal
oad and critical point stresses for each load type. This
Dourse of action outlines potential differences in slopes
snd highlights the spectrum of stress concentration factors
achieved because of geometrical variation,
In addition to the fatigue life investigation, a pretiminary
frequency response analysis was conducted inside the
spectrum of 0.1 and 4.5 Hz because of the big mass of the
rig which oscillate during the fatigue experiments. How=
ever, no worth mentioning influence of any kind to the
resulting stresses was revealed.
Constant amplitude loading
Load type 1 simulates the vertical loads coming from road
‘anomalies while load type 2 the longitudinal loeds acting
at the center of the wheel, Both cases together cover a
wide spectrum of loads occurring during straight-ahead
diving (and partially during braking) and are suspected to
be critical in ftigue failure. In addition, the behavior of
the torsion introduced via the brake flange during braking
is examinod soparately in load type 3.
At least 3 different load lovels were investigated for each
Toad type. The corresponding fatigue life results and the
Wohler curves determined by means of statistical analy
for various values of probability of survival are shown in
Figures 8, 9 and 10,
rote
Fig. 8. Fatigue life curve for load case I
Fig. 9, Fatigue life curve for load ease 2
325Fig. 10, Fatigue life curve for load case 3
‘The slopes of the experimentally determined Wohler
curves agree well with those proposed by intemational
uidelines [1], Additionally, the fatigue life scatter bands
between 10% and 90% probability of survival were found
to be satisfactory natrow. Conceming the scatter band
width, a more detailed analysis of the broken specimens
revealed that there is another geometrical factor that could
contribute to this. When the hats are pressed down on the
axle casing, some inevitable voids are created, slightly
altering the surfaces in contact
Variable amplitude loading
‘The same series of experiments is being carried out using
«4 Gaussian-like sequence of load amplitudes with 201396
reversal points, This sequence is transformed to a sinysot=
dal pulse and applied to the specimens for different values
of maximum load, The ready results up to date are few
because of the time needed to complete each experiment
(10 to 20 times longer than the respective experiments
with constant amplitude). Thus, the available experimen-
{al data for variable ampliuide loading presented here is
limited to load type 1
}
i
i
Fig. 11. Fatigue lives at constant and variable amplitudes
‘for load case I
5. CONCLUSION
Wobler curves and Ga8iner data as well as the scatter of
fatigue life for failure-critical welded joints in bodies of
‘commercial vehicle axles have been experimentally de-
{ermined, In addition, an experimental stress-strain analy-
326
sis of the complete axle structure has besn performed for
three different cyclic loading cases,
The slope of the Wahler curves has been found out to be
in satisfactory agreement with the corresponding recom-
mendation of the IW guideline [1] (Slope k=3). The fa-
‘igue lite scatter expressed as the ratio of fatigue lives for
probabilities of faiture of 90% and 10% is quite narrow.
The experimental data determined here can be used for
the durability assessment of the axle bodies taking meas-
ured operational cyclie loads or standardized load spectra
[2] imto consideration, Additionally, the stress results in
conjunction with the determined W8hler curves can be
used for the validation of modeling methodologies by
imcans of finite elements as well as of fatigue calculation
methods such as the Foot Spot Stress Approach or the Bf
fective Notch Concept. These topics will set the further
research activities on the subject.
REFERENCES
[1] Hobbacher A., Recommendations for fatigue desiga
of welded joints and components, IIW document
XULL-2151-07 | XV-1254-07, 2007.
[2] Gtubisic V., Determination of Load Spectra for De-
sign and Testing, International Journal of Vehicle
Design 15 (1994), 8-26.
CORRESPONDENCE,
| Michail MALIKOUTSAKIS, Dipl
Ing,
Aristotle University of Thessaloniki
Faculty of Engineering
Lab. of Machine Elements
‘& Machine Design
54124 Thessaloniki, Greece
mmalikou@auth gr
‘Alexandros SAVAIDIS, Dr. Ing,
National Technical University
of Athens
Department of Mechanios
Zografou Campus
15773 Athens, Greece
asavaidis@tellas gr _|
Christos SALPISTIS, Assist. Prof,
Aristotle University of Thessaloniki
Faculty of Engineering,
Lab. of Machine Elements
& Machine Design
54124 Thessaloniki, Greece
chrsalp@meng.auth ge
Georgios SAVAIDIS, Assist Prof,
Aristotle University of Thessaloniki
Faculty of Engineering
Lab. of Machine Elements
& Machine Design
54124 Thessaloniki, Greoce
gsavaidis@meng auth.gr _