Engineering Failure Analysis 46 (2014) 92103

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Fatigue life prediction of parabolic leaf spring under various

road conditions
Y.S. Kong a,b, M.Z. Omar a,c,, L.B. Chua b, S. Abdullah a,c
a
Department of Mechanical and Materials Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, 43600 UKM Bangi,
Selangor, Malaysia
b
APM Engineering & Research Sdn Bhd, Level 4, Bangunan B, Peremba Square, Saujana Resort, Seksyen U2, 40150 Shah Alam, Selangor, Malaysia
c
Centre for Automotive Research, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, 43600 UKM Bangi, Selangor, Malaysia

a r t i c l e

i n f o

Article history:
Received 19 May 2014
Accepted 29 July 2014
Available online 14 August 2014
Keywords:
Parabolic leaf spring
Variable amplitude
Finite element
Fatigue life

a b s t r a c t
Parabolic leaf spring experiences repeated cyclic loading during operating condition. Fatigue life assessment of the parabolic leaf spring is a signicant aspect during the component
design stage. This paper serves to simulate the fatigue life of a parabolic leaf spring design
under variable amplitude loading (VAL). VALs carry the road signal that provokes fatigue
failure on leaf spring. In order to seek for comprehensive leaf spring fatigue assessment,
VALs signal were gathered through measurements from various road conditions such as
highway, curve mountain road and rough rural area road. Subsequently, fatigue life of particular leaf spring design was predicted using nite element (FE) stressstrain model
together with VALs signal as load input. For more conservative way, Morrow and Smith
Watson Topper (SWT) mean stress correction methods were also applied. The results indicate that fatigue life of leaf spring is lowest during rough road mission, followed by curve
mountain road and smooth highway road respectively. Additional design modication to
prolong the fatigue life of the parabolic leaf spring is compulsory. The road VALs has provided even more realistic fatigue life estimation of parabolic leaf spring design when compared to traditional controlled laboratory method.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
In the fast pace automotive technology era, automotive engineers facing challenge in rapid development of complex automotive which can be considered as a compound structure made with many mechanical components subjected to complex
cyclic loading. In the design stage, durability evaluation of the vehicle components is one of the major concerns to be considered, however the experimental assessment is time consuming and expensive. Hence, nite element (FE) method which is
known as state of art technology in design is widely applied in automotive industry to assist in assessing the stress level of
automotive components, especially the suspension system, such as leaf springs [1]. With the enhanced computing
power and advance nite element software nowadays, accessibility of automotive components durability has become more
convenient and possible.

Corresponding author at: Department of Mechanical and Materials Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan
Malaysia, 43600 UKM Bangi, Selangor, Malaysia. Tel.: +60 3 8911 8010; fax: +60 3 8925 2546.
E-mail address: zaidi@eng.ukm.my (M.Z. Omar).
http://dx.doi.org/10.1016/j.engfailanal.2014.07.020
1350-6307/ 2014 Elsevier Ltd. All rights reserved.

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

93

In automotive part manufacturing industry such as spring maker, fatigue analysis procedures for the design of products
rely on the techniques that follow relatively simple procedures, apply single phase repeated loading on products until the
components fail. Lab experimental fatigue life prediction has poor description on the real life road condition whereby the
actual condition could be much more severe. Currently, strainlife fatigue analysis calculations could be performed base
on time series data, rainow matrix, or multi-column list of cycles. A complete schedule of multiple time series events
can also be summed to give overall damage. Currently, strain life fatigue analysis is also available with nite element (FE)
method. Instead of strain life analysis, others fatigue analysis approach such as stresslife, crack propagation, creep analysis,
vibration fatigue and multi-axial fatigue also widely adopted in FE method fatigue analysis. For the strainlife method, three
of the most popular strainlife models are CofnManson, Morrow and SmithWatsonTopper (SWT). CofnManson is
performed without mean stress effects while Morrow and SWT consider the mean stress effects [2].
Recently, another strainlife approach known as effective strain damage method (ESD) which is based on crack growth
and crack closure mechanisms was developed. The algorithm is used to account for the cycles sequence effect [3]. In fatigue
analysis, strainlife model combined together with VAL are extensively used. Characterisation of load sequence effect on fatigue crack propagation under VAL has been conducted and proved [4]. Loading sequence is affecting the fatigue level of the
component. Fatigue crack growth test under constant amplitude loading (CAL) and VAL in different environment to nd the
combined effect of load history and environment is emphasised [5]. VAL with imposing multiple overload-underload cycles
is developed to predict fatigue crack growth through probabilistic approach [6]. Multi-axial fatigue with VAL correlation has
been developed to predict fatigue life of rubbers [7]. Probability distribution of fatigue life of steel notched elements under
VAL directly from tensile properties is predicted by using probabilistic stress life (PSN) curves [8]. All the research in VAL
has proven its signicance towards fatigue research where the traditional laboratory experiment unable to understand.
Numerous researchers have proposed methods for fatigue life prediction for automotive components subjected to variable amplitude loadings (VAL). Automotive components require high cycle fatigue life especially the suspension system. Fatigue life of vehicle suspension components such as parabolic springs [9], lower suspension arm [1012] has been analysed
under VALs. Modied stresslife and strainlife method in predicting automotive components that are exposed to VAL have
been suggested [13]. In their research, CAL with random overloads was converted into VAL to predict fatigue crack propagation of steel. Damage tolerance reliability analysis methodology for spot-welded joints on automotive under VAL history
has also been performed [14]. Research on fatigue assessment presents that wavelet bump extraction to summarise load history and obtain fatigue lives under VAL has shown a high correlation to the experiment data [15,16]. Based on these publications, it is obvious that researches are intended to create VAL according to CAL. The reason for this is due to VAL is always
provide the more accurate results.
In this analysis, VAL data of a parabolic leaf spring was collected from three different road conditions which are highway,
curvy and rough road. VALs were then used as the load input for FE fatigue analysis. Previous CAL fatigue test is widely used
to determine the strainlife or stresslife curve for metal products. However, most automotive components especially suspension components are subjected to random loading conditions in which stressstrain cycles uctuate with time. Life estimation of components with CAL is not sufcient to describe the actual loading conditions where the actual conditions are
more damaging. VAL includes occasional severe event which may happen during driving condition. Severe conditions were
not considered during CAL fatigue prediction. Therefore, strain life approaches with road VAL were applied to evaluate the
fatigue life of the parabolic leaf spring for realistic environment fatigue estimation. Other than different road conditions,
effects of Morrow and SWT strain life model were also performed to seek for the difference between fatigue results.
2. Fatigue strainlife approach
Fatigue life of automotive components under service loadings is commonly evaluated by strainlife damage approach
[911,16]. The early stage of fatigue formula was explored about a century ago where Basquin observed the linear relationship between stress and fatigue life in log scale when the stress is limited [2]. The preliminary fatigue formula controlled by
stress is listed in Eq. (1):

ra r0f 2Nf b

where ra is the stress amplitude, r0f is the fatigue strength coefcient, Nf or 2Nf is the number of cycles or reversals to failure,
b is fatigue strength exponent. Later, Cofn and Manson proposed that plastic strain may also be related with fatigue life by
simple power law. Subsequently, Morrow combined the work of Basquin, Cofn and Manson whereby the elastic and plastic
strain devotes to fatigue life as well [17]. Nowadays, among the strain life method, CofnManson, Morrow, and SWT are the
most well-known approaches in nite element model for life assessments. The CofnManson strainlife is mathematically
dened as in Eq. (2):

ea

De r0f
2Nf b e0f 2Nf c
2
E

where ea and D2e is the total strain amplitude, e0f is the fatigue ductility coefcient, c is the fatigue ductility exponent, E is the
modulus of elasticity.

94

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

Another method often referred to as the Morrow meanstress correction relations, taking the mean stress, rm into consideration, form the following equation:

ea

De r0f  rm

2Nf b e0f 2Nf c

2
E

In Eq. (3), rm is taken to be positive for tensile values and negative for compressive values. Morrows equation is consistent with the observation that mean stress effects are signicant at low value of plastic strain while less effect on high value
of plastic strain. Lastly, the strainlife mean stress correction method referred to the SWT method proposed a different
method to consider the mean stress effect by counting maximum stress during one cycle [18]. This relationship was based
on strainlife test data which was obtained at various mean stresses. Thus, the SWT expression is mathematically dened as:

rmax ea E r0f 2 2Nf b r0f e0f E2Nf bc

where rmax = rm + ra and ea is the alternating strain. This is equation is based on assumption that, for different combinations
of strain amplitude ea and mean stress rm, the product rmaxea remains constant for a given life. SWT method indicates that no
damage will occur when the maximum stress is zero or negative, which is not tolerated in the reality. Comparing the Morrow
and SWT method, SWT method predicts conservative life for loads predominantly tensile while Morrow provides more realistic result when the predominant load is in compressive.
Linear damage rule was rst proposed by Palmgren and Miner as mentioned by Verboom and Koten [19]. The linear damage summation rule proposed by PalmgrenMiner and damage of one cycle Di can be expressed as:

Di

1
Nfi

where N is the number of constant amplitude cycles to failure. Further calculation for fatigue damage under VAL, a linear
cumulative damage approach has been dened in the following equation:

Di

X ni
1
nif

where Di is the cumulative fatigue damage, ni is the number of applied cycles, nif is the number of constant amplitude cycles
to failure. The linear damage summation rule does not take into account the effect of the load sequence on the accumulation
of damage due to cycle fatigue loading. The failure will occur when the summation of individual damage values caused by
each cycles reach value of one.
3. Finite element (FE) simulation procedures
For material assignation in fatigue analysis, elastic plastic material model was utilised. In this material model, the material behaves as linear elastic when the equivalent stress is lower than yield stress. The relation between stress and during
plastic deformation was calculated using the RambergOsgood equation:

  10

r
r
e ee ep 0
E

where r is the equivalent stress, E is the material elastic modulus, K0 is the monotonic work hardening and n0 is the monotonic strain hardening exponent. The monotonic work hardening and monotonic strain hardening exponent of steel SAE
5160 are 1940 MPa and 0.05 respectively [20]. Strain life parameters were used to predict fatigue life of the model during
the strainlife approach. Details of strain life parameters are listed in Table 1.
Prior to fatigue simulation, FE analysis was performed on parabolic leaf spring model to obtain the stressstrain result
during vehicle fully loaded load condition. The work ow of fatigue analysis with nite element method is shown in
Fig. 1. First of all, a three-dimensional (3D) model for the parabolic leaf spring was developed through commercial nite
element software Altair Hyperworks. 8-noded hexahedra elements with 8 nodes ordered were used for the solid mesh.

Table 1
Mechanical properties of the SAE 5160 steel [20].
Properties

SAE 5160

Ultimate tensile strength

Yield strength, a (MPa)
Modulus of elasticity, E (GPa)
Fatigue strength coefcient, Sf0
Fatigue strength exponent, b
Fatigue ductility exponent, c
Fatigue ductility coefcient, Ef0

1584
1487
205
2063
0.08
1.05
9.56

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

Boundary conditions

Geometric and FE model

95

Materials & properties

FE stress-strain analysis
(Pre-process, solve, post process)

Leaf spring characteristic

correlation

Variable amplitude loading

from various roads

Strain life based fatigue

assessment

Design validation
Fig. 1. Flow chart of parabolic leaf spring fatigue life analysis.

Front eye with revolute joint

Rear eye with shackle

(Rotation in Y-axis)

(Rotation in Y-axis and

translation in X-axis)

Loading from centre (Z-axis)

Fig. 2. Parabolic leaf spring boundary conditions.

8-node hexahedra elements are recommended for stressstrain analysis in FE method due to higher integration number
which providing greater accuracy compared to tetrahedral elements [21]. Thus, FE model with a total of 92,577 elements
and 115,157 nodes were generated. Minimum length of those elements was controlled within 0.89 mm to avoid extra computational time. A nonlinear implicit quasi static time-integration scheme was applied to perform the analysis. In implicit
scheme, a global stiffness matrix should be assembled and inverted, leading to higher computational duration per loading
step. However, the nonlinear implicit quasi static scheme reduced the inertia generated while remained the accuracy of
the simulation result [21].
Boundary condition of parabolic leaf spring is one of the main concerns during pre-processing in FE analysis. There are
three main mounting regions in the parabolic leaf spring, as shown in Fig. 2. The attachment of the following three parts
was considered in the FE boundary conditions. Front eye region is attaches to a bracket with pin. Therefore, front eye was
allowed to rotate in Y-axis only. Rear eye was attached to shackle and it was allowed to rotate in Y-axis and translate in
X-axis. Centre of the parabolic leaf spring was clamped together with front axle and therefore centre of the parabolic leaf
spring was allowed to translate in X, Z-axis and rotation Y-axis [22]. A concentrated load of was applied from the centre
of the parabolic leaf spring in vertical direction as the result of laden load. The parabolic spring was pushed to deect into
gross vehicle weight condition.
Fig. 3 highlights the location of parabolic leaf spring related to the front axle module of the bus suspension system. Connection of a parabolic leaf spring in suspension front module is presented. Centre of the parabolic leaf springs were attached
to the front axle of the bus through U-bolt clamping. Therefore, the centre of the parabolic leaf spring is assumed to be rigid.
The front axle was then connected to the tyre linkage. At the same time, shock absorbers and anti-roll bars were assembled

96

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

Shackle
mounting

Wheel

Parabolic
leaf spring

Shock absorber
mounting

Critical
region
Bracket mounting
Front axle

Anti-roll
mounting

Fig. 3. Front module of parabolic leaf spring suspension.

Translation X &
rotation Y

Translation Z

Translation X &
rotation Y

Fig. 4. Illustration of leaf spring experimental vertical stiffness test setup.

Fig. 5. Comparison of simulation versus experimental spring stiffness.

together with the parabolic leaf springs through shock eyes of the parabolic leaf spring. Another end of shock absorbers and
anti-roll bars were attached to the chassis of the bus.
During nite element geometric nonlinear analysis, parabolic leaf spring undergoes large deformations. The nonlinearities either come from contacts or materials. A general-purpose contact introduced in Radioss solver was applied. The interface stiffness Is was computed from both the master segment Km and the slave segment Ks. The interface stiffness relationship
between the master and the slave is dened in the following equation:

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

Is



Km  Ks
Km Ks

97

Introduced friction formulation was implemented in this contact interface. The most well-known friction law is the Coulombs law of friction. This formulation provides accurate results with only one input parameter, the Coulomb friction coefcient l [23].
4. Experimental works and variable amplitude loading (VAL)
Correlation between the numerical method and experimental method is extremely signicant for fatigue life prediction.
Experimental setup of vertical stiffness measurement was performed in Fig. 4. The parabolic leaf spring was placed at the
spring load tester and a load was applied from the centre of the parabolic leaf spring. Deection of the parabolic leaf spring
with respect to the loading was recorded. The experimental and simulation analysis on vertical stiffness of parabolic leaf
spring is shown in Fig. 5. From Fig. 5, the vertical stiffness of simulation leaf spring model is 276 N/mm while the experimental leaf spring is 289 N/mm. The variation between simulation model and leaf spring sample in lab is about 4.7% and hence
the simulation was expected to have good accuracy in representing the actual spring behavior.
Strain signal from various road conditions are vital for fatigue life prediction of any automobile especially for springs. The
major reason for conducting the VAL tests were because of determining fatigue life under this complex loading is not

Fig. 6. Variable amplitude loading data collection setup on bus.

Spring centre

Maximum principal
strain: 2.66 10-3

Fig. 7. Equivalent maximum principal strain of parabolic leaf spring.

98

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

Spring centre

Maximum principal
stress: 468 MPa

Fig. 8. Equivalent maximum principal stress of parabolic leaf spring.

Fig. 9. Strain time history of the strain gauge for different road conditions (a) smooth highway, (b) curve mountain road and (c) rough rural area road.

possible by any cumulative damage hypothesis [12]. Therefore, for the fatigue life assessment of three different road conditions where the bus could exhibit, life of the parabolic leaf spring must be predicted by conducting actual on road test. The
VAL testing procedure has been summarised in Fig. 6. From Fig. 6, critical region of the component during service loading
should be evaluated through applying strain gauges on particular area. Sampling rate of 200 Hz was used due to suspension
resonant frequency is about only 1215 Hz [24]. For vehicle suspension parts, frequency range of interest for fatigue analyses
is between 40 and 60 Hz [25]. According to sampling theory, the sampling frequency is at least 2.5 times the target

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

99

maximum frequency range. Therefore, sampling rate of 200 Hz was enough to collect data with enough information for fatigue assessment.
Initial FEA was performed to identify the component critical areas to select the suitable location for strain gauge attachment [26]. The location was selected based on the highest strain output from FEA result. After that, rosette strain gauge was
applied on the surface of the parabolic leaf spring. In order to ensure the strain signal is accurate, strain gauge attachment
point of the parabolic leaf spring was polished until mirror surface. The strain gauge was then connected to a data logger and
data logger was connected to a laptop for strain signal monitoring. The bus was loaded with battery to simulate the fullyloaded condition. A total of 80 batteries were located either in compartment or seat of the bus to represent passengers and
luggage. Each battery was weighted about 45 kg and hence a total of 3600 kg load was added. This load exerts has pushed the
parabolic leaf spring into maximum deected conditions. Strain values of the initial readings were reset before testing to
obtain the strain readings due to road loads.

5. Results and discussion

FE nonlinear implicit scheme was applied to obtain the stress strain results of the parabolic leaf spring. The principal
strain contour plot of the parabolic leaf spring is shown in Fig. 7. From Fig. 7, the maximum principal strain of
2.66  103 of the parabolic leaf spring under centre vertical loading from centre was observed. Furthermore, maximum
principal stress of 468 MPa was observed in the stress contour plot shown in Fig. 8. The experimental strain data collection
was determined based on the maximum stress and strain region where it was located at 150 mm from the centre of the parabolic leaf spring.
Experimental collected VAL data from various road conditions are shown in Fig. 9. Fig. 9 depicts the strain versus time
plot of the parabolic leaf spring under various road conditions. As shown in Fig. 9, the road conditions were classied into
smooth highway road, rough rural area road and curve mountain road. For smooth highway road, the value of strain generated was much smaller than rough and curve mountain road due to the highway road was mostly straight and with good
surface roughness. Rough rural area road created highest strain value among the selected roads. This may due to numerous
potholes on the road and rough road surface. Third strain signal was collected during the curve mountain road. This condition was important for parabolic leaf spring fatigue prediction due to the shifting of centre of gravity of the bus. The loads
were concentrated at the front suspension when the vehicle was going downhill and more loads could be created at the front
parabolic leaf spring. Besides that, the left and right shifting of the vehicle mass also generated additional loadings.
Subsequently, VAL was used to conduct FE based fatigue simulation through nCode DesignLife software as shown in
Fig. 10. In this analysis, VALs were used as time series input to the parabolic leaf spring together with parabolic leaf spring

Fig. 10. View of fatigue life simulation using nCode DesignLife software.

100

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

under vehicle fully loaded condition. Stress and strain data from FE nonlinear implicit analysis was adopted as the input to
the DesignLife strain life solver. For uniaxial analysis, principal stress or strain of FE elements were combined to form absolute maximum principal value.
Fatigue damage contour of simulated parabolic leaf spring model is shown in Fig. 11. The fatigue damage was concentrated at the middle cantilever region of the parabolic leaf spring. As seen from Fig. 11, a rough road condition generates
highest fatigue damage of parabolic leaf spring (5.853  104) compared to smooth highway road (1.550  107) and curve
mountain road (9.793  106). Therefore, parabolic leaf spring shall fail due to fatigue in shorter period than smooth highway and curve mountain roads. In conjunction to that, fatigue life of the parabolic leaf spring was calculated and fatigue life
contour is shown in Fig. 12. The minimum fatigue life of 1.708  103 was obtained under rough road conditions. For curve
road and highway road, 1.021  105 and 4.349  107 cycle life were obtained respectively. During the parabolic leaf spring
design stage, minimum of 2.00  105 of cycle life is required [9]. When the bus with this parabolic leaf spring design is used
for highway cruising, the lifetime of the parabolic leaf spring is long and sufcient enough. However, additional design modications would be required when the bus is used for mountain and rural area. Parabolic leaf spring design with higher fatigue life is required since the life cycle of parabolic leaf spring cannot surpass the minimum life requirement. Fatigue life of
leaf spring was determined based on the stress level and therefore, this parabolic leaf spring shall be designed to possess
lower stress level to extend its fatigue life on rural and rough road.
Fatigue life and damage of FE base fatigue simulation were summarised in Tables 2 and 3. In real environment, prediction
of fatigue life is always much deviated from actual. Many algorithms have been developed to correlate the fatigue life prediction method to experiment to have higher accuracy fatigue results. Two different mean stress correction factors, Morrow

Fig. 11. FE analysis base fatigue damage, smooth highway: (a) Morrow and (b) SWT, curve mountain road: (c) Morrow and (d) SWT, rough rural area road:
(e) Morrow and (f) SWT.

101

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

Fig. 12. FE analysis base fatigue life, smooth highway: (a) Morrow and (b) SWT, curve mountain road: (c) Morrow and (d) SWT, rough rural area road: (e)
Morrow and (f) SWT.

Table 2
Fatigue damage prediction using the different strain life models.
Road condition

Morrow

SWT

Smooth highway
Curve mountain
Rough rural area

2.299  108
5.236  106
5.853  104

1.550  107
9.793  106
5.305  104

Table 3
Fatigue life prediction using the different strain life models.
Road condition

Morrow

SWT

Smooth highway
Curve mountain
Rough rural area

4.349  107
1.910  105
1.708  103

6.453  106
1.021  105
1.885  103

and SWT were used to evaluate the fatigue life and damage calculation under variable amplitude loading. In smooth highway
and curve mountain road conditions, SWT strain life model has presented the fatigue life and damage in a more conservative
way. However, in rough road condition where the range strain signal was wider, the SWT showed a slightly lower fatigue

102

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

damage and higher life cycles result. As reported by Kadhim et al. [12] in strain life model for fatigue prediction, SWT model
yields smaller deviation than Morrow strain life model when compared to experimental results. This is the reason that why
SWT is preferred in this analysis.
In pseudo static analysis, Morrow and SWT models are not accurate when compared to experimental fatigue lives. VALs
still consists of some setbacks when Palmgren Miner damage rule were applied. The reason is due to the loads and cycle
interaction effects are not taking into consideration when VAL loadings are applied. Validity of PalmgrenMiner rule also
discussed by Johannesson et al. [27] in which the order of cycle application to the material causes residual stress and threshold effects when small cycles interspersed with large cycles. Zakaria et al. [28,29] has published the load sequence effects on
fatigue life which obviously illustrated the signicance of load effects on component fatigue life.
In leaf spring design perspective, there were a few shortfalls which was difcult to resolve. Leaf spring is a suspension
component where shot peen has been applied on the surface to induce compressive residual stresses [30]. Complete shot
peening could improve the fatigue life of component in huge range [31]. In fatigue life simulation, shot peening effect of
the parabolic leaf springs was neglected. Hence, the predicted fatigue life of the parabolic leaf spring is lesser than actual
condition. In the current study, default setting values were used for analysis. Nevertheless, the simulation results provide
a good concept during the fatigue design of parabolic leaf spring. Further experimental work on shot peening could be
embedded in FE method to encounter the shot peening effects in fatigue life prediction.
6. Conclusions
Fatigue life predictions of parabolic leaf spring on smooth highway, curve mountain road and rough rural area road were
presented. FE model together with VAL was used as the input to the fatigue life and damage simulation. FE method has provided the critical region of the parabolic leaf spring where the strain gauge was attached. Strain signal represents the parabolic leaf spring loading history when the bus travels on the particular road. Through this analysis, prediction of actual leaf
spring life based on various road conditions was obtained where it could never be predicted in lab condition. Based on the
simulation result, rural area rough road condition contributed highest damage to the parabolic leaf spring followed by curve
mountain road and smooth highway road. Therefore, life of parabolic spring during rough road operations is lowest, and the
second is mountain road condition. Smooth highway consists of highest life compared to others. Parabolic leaf spring redesign is required to be reconsidered when the bus is used for curve and rough roads. During the fatigue simulation, both Morrow and SWT model was adopted to estimate the fatigue life in more conservative way. With the simulation results,
description of fatigue behavior under various road terrains is possible and when bus usage in particular environment is
known, the duration for parabolic leaf spring to fail in fatigue could be estimated.
References
[1] Kong YS, Omar MZ, Chua LB, Abdullah S. Stress behavior of a novel parabolic spring for light duty vehicle. Int Rev Mech Eng 2012;6(3):61720.
[2] Bishop NWM, Sheratt F. Finite element based fatigue calculations. U.K.: NAFEMS; 2000.
[3] DuQuesnay DL. Applications of overload data to fatigue analysis and testing. Application of automation technology in fatigue and fracture testing and
analysis, vol. 4; 2002. p. 16580.
[4] Huang X, Torgeir M, Cui W. An engineering model of fatigue crack growth under variable amplitude loading. Int J Fatigue 2008;30:210.
[5] Lee EU, Glinka G, Vasudevan AK, Iyyer N, Phan ND. Fatigue of 7075-T651 aluminium alloy under constant and variable amplitude loadings. Int J Fatigue
2009;31:185864.
[6] Kocanda D, Jasztal M. Probabilistic predicting the fatigue crack growth under variable amplitude loading. Int J Fatigue 2012;39:6874.
[7] Harbour RJ, Fatemi A, Mars WV. Fatigue life analysis and predictions for NR and SBR under variable amplitude and multiaxial loading conditions. Int J
Fatigue 2008;30:123147.
[8] Zheng Z, Wei J. On the prediction of PSN curves of 45 steel notched elements and probability distribution of fatigue life under variable amplitude
loading from tensile properties. Int J Fatigue 2005;27:6019.
[9] Ahmad Refngah FN, Abdullah S, Jalar A, Chua LB. Life assessment of a parabolic spring under cyclic strain loading. Eur J Sci Res 2009;28(3):35163.
[10] Abdullah S, Al-Asady NA, Arifn AK, Rahman MM, Nopiah ZM. FEA based fatigue life assessment of an automobile lower suspension arm using various
strainlife models. In: Proceedings of the 4th WSEAS international conference on applied and theoretical mechanics; 2008.
[11] Rahman MM, Kadirgama K, Noor MM, Rejab MRM, Kesulai SA. Fatigue life prediction of lower suspension arm using strain life approach. Eur J Sci Res
2009;30(3):43750.
[12] Kadhim NA, Abdullah S, Arifn AK. Effective strain damage model associated with nite element modelling and experimental validation. Int J Fatigue
2012;36:194205.
[13] Taheri F, Trask D, Pegg N. Experimental and analytical investigation of fatigue characteristics of 350WT steel under constant and variable amplitude
loading. Mar Struct 2003;16:6991.
[14] Mahadevan S, Ni K. Damage tolerance reliability analysis of automotive spot-welded joints. Reliab Eng System Safety 2003;81:921.
[15] Abdullah S, Al-Asady, Arifn AK, Rahman MM. A review on nite element analysis approaches in durability assessment of automotive components. J
Appl Sci 2008;8(12):2192201.
[16] Abdullah S, Choi JC, Giacomin JA, Yates JR. Bump extraction algorithm for variable amplitude loading. Int J Fatigue 2006;28:67591.
[17] Morrow J. Fatigue design handbook. Warrandale: Advances in Engineering, Society of Automotive Engineers; 1968.
[18] Smith KN, Watson P, Topper TH. A stressstrain function for the fatigue of metals. J Mater 1970;4:76778.
[19] Verboom G, Koten H. Vortex excitation: three design rules tested on 13 industrial chimneys. J Wind Eng Indust Aerodyn 2010;98(3):14554.
[20] Boardman B. Crack initiation fatiguedata, analysis, trends and estimation. SAE Technical Paper; 1982. 820682.
[21] Altair Engineering, Inc. Altair Users Guide, India: Altair Hyperworks; 2013.
[22] Shimoseki M, Hamano T, Imaizumi T. FEM for springs. New York: Springer; 2003.
[23] Kuttler K, Shillor M. Dynamic bilateral contact with discontinuous friction coefcient. Nonlinear Anal: Theory Meth Appl 2001;45:30927.
[24] Gillespie TD. Fundamentals of vehicle dynamics. U.S.A: Society of Automotive Engineers; 1992.
[25] Sener AS. Determination of vehicle components fatigue life based on FEA method and experimental analysis. Int J Electron Mech Mechatron Eng
2011;2(1):13345.

Y.S. Kong et al. / Engineering Failure Analysis 46 (2014) 92103

103

[26] Aykan M, Celik M. Vibration fatigue analysis and multi-axial effect in testing aerospace structures. Mech Systems Signal Process 2009;23:897907.
[27] Johannesson P, Svensson T, Mare J. Fatigue life prediction based on variable amplitude tests methodology. Int J Fatigue 2005;27:95465.
[28] Zakaria KA, Abdullah S, Ghazali MJ, Azhari CH. Inuence of spectrum loading sequences on fatigue life in high-temperature environment. Eng Failure
Anal 2013;30:11123.
[29] Zakaria KA, Abdullah S, Ghazali MJ. Comparative study on fatigue life behaviour of AA6061 and AA7075 alloys under spectrum loadings. Mater Des
2013;49:4857.
[30] Juvinall RC, Marshek KM. Fundamentals of machine component design. 4th ed. Asia: John Wiley & Sons (Asia); 2006.
[31] Torres MAS, Voorwald HJC. An evaluation of shot peening, residual stress and stress relaxation on the fatigue life of AISI 4340 steel. Int J Fatigue
2002;24:87786.