Está en la página 1de 8

Engineering Failure Analysis 80 (2017) 464–471

Contents lists available at ScienceDirect

Engineering Failure Analysis


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

Review

The needs of understanding stochastic fatigue failure for the MARK


automobile crankshaft: A review
S.S.K. Singha,b, S. Abdullaha,b,⁎, N. Nikabdullahc
a
Department of Mechanical & Materials Engineering, Faculty of Engineering & Built Environment, Universiti Kebangsaan Malaysia, 43600 UKM
Bangi, Selangor, Malaysia
b
Centre for Automotive Research, Faculty of Engineering & Built Environment, Universiti Kebangsaan Malaysia, 43600 UKM Bangi, Selangor,
Malaysia
c
Department of Mechanical Engineering, Universiti Malaysia Pahang, Lebuhraya Tun Razak, 26300 Gambang, Pahang, Malaysia

AR TI CLE I NF O AB S T R A CT

Keywords: This paper reviews the fatigue failure mechanisms for the automobile crankshaft under service
Failure loading through the stochastic point of view. Fatigue failure of crankshafts are reviewed in
Fatigue general, as it is a major concern due to the uncertainties that arise i.e. randomness in structural
Stochastic materials, the geometric shape of the component and randomness of service loads. There has
Crankshaft
been very little research carried out in assessing the fatigue failure using the stochastic process in
Durability
predicting the fatigue life of crankshafts. This review paper discusses the durability aspects of the
component and is followed by a review of the characteristics of loading and the stochastic fatigue
failure effect on the components. In addition, the stochastic approach from empirical model
aspect using a safe-life approach from the more recent advances in computational methods to
assess stochastic fatigue failure was discussed and reviewed in the context of this paper. The
integration between the empirical and probabilistic methods can be quantified using statistical
models, which evaluate the damage that leads to fatigue and eventually fatigue failure. Hence,
this review provides a platform for understanding the stochastic fatigue failure for an accurate
predictive prediction on the structural integrity of components, especially in the automobile
industry.

1. Introduction

The ability to monitor the resilience to wear and the potential for fatigue is a critical safety factor in automotive components. The
crankshaft is a crucial element located at the heart of an engine, and any fatigue failure will be catastrophic giving the impact on the
other connecting subcomponents. The fatigue fractures of rotating shafts is generally due high stress concentration from shaft dia-
meter and sharp corner at the intersection between two different diameters of the shafts [1]. Besides the geometric design, torsional
vibration of the shaft that are related to the vibration modes at the natural frequencies of the shaft. The actual torsional vibration is
due factors such as mass of inertia, damping and stiffness based on the design of the shaft. Hence, making the shafts and crankshaft
not exceptional from this occurrences because these components are subjected to significant number of cyclic loadings. Making
several root causes, such as sudden overloads, improper engine operating and maintenance, or by fatigue as mentioned by Fonte et al.
[2]. The fatigue occurrence is due to the cyclic loadings with stress levels lower than yield or ultimate strength of material.


Corresponding author at: Department of Mechanical & Materials Engineering, Faculty of Engineering & Built Environment, Universiti Kebangsaan Malaysia, 43600
Bangi, Selangor, Malaysia.
E-mail address: shahrum@ukm.edu.my (S. Abdullah).

http://dx.doi.org/10.1016/j.engfailanal.2017.06.023
Received 21 November 2016; Received in revised form 6 June 2017; Accepted 14 June 2017
Available online 15 June 2017
1350-6307/ © 2017 Elsevier Ltd. All rights reserved.
S.S.K. Singh et al. Engineering Failure Analysis 80 (2017) 464–471

Making the design of crankshafts is vital, to ensure that component is robust and will be able to operate with a high degree of
safety and reliability throughout the lifetime of the engine. But yet, the main reason for failure of crankshafts is due to fatigue under
fluctuating loads [3]. Particularly over the recent decade, the failure analysis of the crankshaft was carried out in terms of forensic
engineering, whereby Ktari et al. [4] investigated the failure using Scanning Electron Microscope (SEM), fractography, tensile and
hardness tests. This was further enhanced by Becerra et al. [5] and Fonte et al. [6] through chemical analysis, x-ray analysis (EDAX)
and visual examination techniques. This illustrated that the fatigue failure of the component were seen purely as a deterministic
failure attributed from the material properties, geometric parameters and large interval or missing loads.
An automobile crankshaft must meet strict criteria during production to ensure that the engine is reliable and able to operate for a
long time. However, there is always the possibility of unavoidable component failure, which over time will result from fluctuating
service loading [7]. Not only does the crankshaft nearing or experiencing fatigue failure threaten the performance and correct
functioning of the entire system, but it will also cause other components to operate and perform at suboptimal levels. Defining these
fluctuating load–time histories as a standardised input for calculating the stress response in durability assessment as mentioned by
Wanneburg et al. [8]. This incomplete information of its effective parameters could lead into catastrophic failure. Besides that, the
stochastic consideration of the randomness is dependent on the material properties and geometric shape of the crankshaft. Therefore,
the importance of predicting fatigue failure is unequivocal, and data from real-time monitoring can provide an accurate assessment of
the durability of a component or structure. This is because the fatigue failure was associated with structural integrity; render the
functioning of crankshaft components to become less efficient. As indicated by Liu & Mahadevan [9] and Ling et al. [10] that
monitoring any stochastic fatigue damage modelling under variable amplitude loading provides an important indicator for durability
assessment.
A review of existing assessment methods for detecting fatigue failure is an extremely useful exercise to examine the events that
may lead to catastrophic engine failure due to the condition of the crankshaft. In this study, the application of using the stochastic
method that has been employed for many years is reviewed to establish and affirm greater clarity and understanding of fatigue
behaviour. The objective of this study is to review the elements causing crankshaft fatigue failure from the stochastic viewpoint. It is
currently viewed as a deterministic issue, from the local automotive industries standpoint in Malaysia. So, an understanding of the
stochastic process and failure mechanisms will greatly assist in the method to evaluate the structural integrity of components in
contrast to the deterministic model.

2. Overview of crankshaft fatigue failure

The rotation speed of vehicle crankshafts is generally 3000–5000 rpm depending on the engine type. The operating parameters
require a revolution of 720° to complete a 4-stroke cycle, which consists of intake, compression, power, and exhaust strokes. The
highest load on the piston is experienced during compression and power cycles. During these stages, heat generation results from the
ignition of the compressed air-fuel mixture or from diesel internal combustion. The massive load forces the piston downwards until
reaching the dead centre position. The crankshaft is subjected to maximum loading pressure on its journal bearings during these
cycles [11]. Similarly cylinder pressure and piston-connecting rod inertia result in direct loading pressure applied on crankpins and
main journals. The result is cyclic bending and steady torsional on the main journals, and reverse bending on as a result of loading
pressure being transmitted from the pistons and connecting rods [12,13]. The rotating bending and steady torsion exist on the main
journals. The crankpins experience reverse bending only due to the translation movement, where the torsion effect does not exist.
Crankshaft design incorporates many safety factors to avoid exceeding the fatigue strength of the material.
However, high cyclic loading and local stress concentrations can cause component failure, even when fatigue strength values are
not exceeded, and the inherent safety margins of the component are not compromised. Experimental analysis was initially performed
in 1970 by Jensen [14], who applied strain gauges along the journal and fillet web regions to accurately determine the crankshaft
fatigue strength based on bending and torsion loads. The equivalent fatigue strength of each test section was assessed and was plotted
using the probabilistic stress-life curve [15].
The components in an engine are a contributory factor in crankshaft durability where fatigue failure is contingent upon their
resilience. A study on material properties to identify the critical zone based on the loading sequence was conducted in 1997 by Taylor
et al., applying the finite element method [16–17]. The initial crack region was found to occur at the fillet region where cracks spread
in a plane normal to the maximum principal stress direction across the journal bearing, at an approximate 45° angle to the mid-plane
from where the oil seal is located. The interaction results from the oil seal acting as a notch and succumbing to high-stress levels, as
shown in Fig. 1.
Over the recent decade, various experimental techniques such as chemical, mechanical, microstructural and visual inspection was
used to determine the probable causes of premature crankshaft failure [18–23]. The fractures were caused mainly by the resonance
from natural vibrations from the load fluctuations related to the frequency of the shaft rotational speed. The experimental analysis
concentrated more on forensic engineering as it deals with the component that has already failed. Likewise, neglecting the im-
portance of durability analysis in terms of structural integrity. In fatigue life prediction, the fluctuating load–time histories is used for
calculating the stress response based on the materials fatigue properties and geometric shape of the crankshaft.
Fatigue failure generally begin at the crankpins close to the web fillets and spread along a fracture plane at a 45° angle to the
rotational axis in a plane normal to the maximum principal stress direction. Becerra et al. [24] observed that the crack initial at the
crank-pin web region in a plane at around 45° with respect to the rotational axis with typical beachmarks for fatigue failure. This
illustrates that the magnitude and direction of the principal stresses is between the bending and torsion, which inline failure of
rotating shafts due to bending, torsion and effects of multiaxial loading. According to Fonte et al. [23], the torsional load on the

465
S.S.K. Singh et al. Engineering Failure Analysis 80 (2017) 464–471

Fig. 1. Illustration of failure modes (i.) Bending: crack propagation takes place from A to B in a normal plane (ii.) Torsion: In torsion cracks initiate at A and A′ and
grow at approximately 45° to the mid plane [17].

crankshaft can be neglected for the load analysis as it is considered less than 10% of the bending load. This is because of maximum
loads takes place during combustion, hence the acting force on the crankshaft is purely a bending load nature due to the direction of
the force which is exactly at the centre of the crank radius. By considering torsional vibration, the fatigue failure can be related to the
vibration modes at the natural frequencies of the shaft. Therefore, the actual torsional vibration problem can occur due to the
variation of parameters such as those of the vibration system, including mass of inertia and stiffness is based on the which design
[25]. Contributory factors to crankshaft failure which may be deemed insignificant and overlooked in many studies are poor design,
incorrect assembly and shaft misalignment, poor engine operation, inappropriate geometries of fillets on the crankpin webs and main
journals or lubricating holes, and the result of abnormal vibrations and resonance [26]. The chronological overview of the focus of
studies on fatigue failure is shown in Fig. 2.

3. Comparison of stochastic and deterministic failure for the crankshaft

Experimental investigations are frequently conducted on the fatigue behaviour of engineering materials to validate the designs,
guide development, and predict service life. The deterministic approach to quantifying the fatigue behaviour of materials has resulted
from a misalignment between the theory and the actual practice [27]. Testing a single specimen can yield an evaluation of the
material parameters, but using multiple specimens leads to observing loading value results that randomly fluctuate. This is un-
avoidable as the inherent behaviour of materials varies, due to the random nature of process defects, material and structural errors,
and loadings under operating conditions.
It is important that the stochastic approach is accounted for in conjunction with fatigue life prediction under variable amplitude
loading. Therefore, appropriate modelling and analysis techniques are necessary for safer and more reliable designs of mechanical
components. The results obtained from the deterministic and stochastic models can be compared against a specific set of input values,
which originate from similar parametric conditions. The inherent nature of the stochastic framework implies that different results will
occur each time for a set of values, despite the parametric conditions remaining constant [28]. Deterministic results have another
basis, and the differences in fatigue damage assessment can be summarised accordingly in Table 1.

4. Development of stochastic fatigue failure

The rainflow counting technique was developed by Endo in 1968 as reported by Johannesson [29] to assess fatigue failure. The
method, which has been the subject of much research, calculates the number of cycles and uses the local maxima and minima.
Currently, the automotive industry applies this method in relation to dynamic phenomena to predict fatigue damage under service
loading [30]. The rainflow counting technique relies on the local minima and maxima reaching the same level, and then examining a
small downward or upward modulation at a stage either forward or backwards in each cycle, as shown in Eq. (1)

⎧ rainflow cycle with ⎫


NKrfc (i, j ) = # minimum 〈i and maximum 〉 j
⎨ ⎬
⎩ for xk t = 0, 1, …, K ⎭ (1)

where the minimum is less than i, and the maximum is greater than j, it will provide a rise in the interval [i, j].
To overcome the shortcomings of even ideal laboratory testing, modelling fatigue failure under random loading will use the
amount of scatter associated with crack growth under stochastic dynamic service loads. The high computational power required in

466
S.S.K. Singh et al. Engineering Failure Analysis 80 (2017) 464–471

2010’s - now 1. Simulation through finite element to determine the


fatigue vibration for durability assessment
2. Dynamic lumped model analysis
3. Effects of multiaxial loading to determine the main
effects of failure
4. Forensic investigation using mechanical, chemical and
metallurgical analysis

2000’s Investigated failure of the crankshaft in terms of forensics


using:

1. Chemical analysis
2. Mechanical analysis
3. SEM to study the micro mechanism of failure
4. Metallurgical analysis
5. Visual inspection
6. Fractrography
1990’s
’s 1. Simulation through finite element to determine the
critical regions with high stress levels
2. Influences of forged steel and ductile cast iron the life of
the crankshaft

1970’s Experimental analysis to determine toughness of


crankshaft.
Fig. 2. Chronological timeline of crankshaft fatigue failure studies.

calculating fatigue damage results in a cycle-by-cycle method [31]. Using the linear damage accumulation concept based on cyclic
loading enables fatigue to be assessed where the local strain is transformed into the damage domain, according to Eq. (2).

k
n
D= ∑ Ni
i=1 i (2)

where D = fatigue damage of the material; ni = number of applied loading cycles to failure to at a given load level; Ni = number of
cycles to failure at the load level; and Ci = material properties.
Several approaches analysing multiaxial failure by Carpenteri et al. [32], including empirical models, the average stress criterion

Table 1
Difference between the stochastic and deterministic processes in fatigue damage assessment.

Input Stochastic Deterministic

Mechanical model (loading sequence, Random process Able to compute the evolution in time of the stresses in each
material properties, and geometric point of a component from the prescribed geometry, material,
properties) and loading
Description of loading applied Loading is influenced by environmental effects, Loading with no influence on environmental effects. i.e.
i.e. variations of Young's modulus or Poisson variations of Young's modulus or Poisson ratio
ratio
Uncertainties in sub-factors Size, effect, and surface finish can be modelled The effects of sub-factors are considered minimal and at times
through the probability density function negligible
Extraction and counting of cycles Uses the rainflow cycle based on the scatter of Uses the rainflow cycle based on the load history
the specimen test data

467
S.S.K. Singh et al. Engineering Failure Analysis 80 (2017) 464–471

Probabilistic framework

Uncertainties in Random Data Manufacturing


model loading analysis defaults

Geometric parameters: Cyclic a) Number of cycles a) Surface treatment


a) Length loading to failure b) Surface finish
b) Fillet radius b) Statistical analysis c) Environmental
conditions
Material properties: d) Model errors
a) Young’s modulus
b) Poisson ratio
c) Endurance strength
d) Modulus of rigidity
e) Yield strength
Fig. 3. Development of the probabilistic framework.

of the critical plane technique, and the damage mechanics approach. Even for very low amplitudes of cyclic stress, partial damage
occurs albeit the material is consistently below the yield stress level. Small damage increments that are produced accumulate and
eventually result in failure.
In the current study, the chosen method is to calculate fatigue damage using a scalar damage variable. This is due to the ap-
plicability to assess fatigue life. Based on max–min stresses, the random fatigue load sequence using real measured loads is essential,
as variable service loads have a significant effect on the assessment of damage [33]. The assessment is vital in relation to both the
safety and reliability, and among the automobile components, where the crankshaft is of primary importance. A fatigue assessment is
difficult to perform, particularly in the case of cyclic or random multiaxial loading histories on an automobile crankshaft. This is
because during the whole phenomenon there is variation and damage accumulation that depends on all the components of the stress
tensor. The effect of fatigue damage on the life of the crankshaft components and their structural integrity under uniaxial or mul-
tiaxial random stress histories with variable amplitude loadings is relevant to the consideration of materials.

4.1. The stochastic framework in fatigue failure assessment

In this study, as the fatigue failure is considered to be stochastic to model the process of modelling will use random data that is
generated from the same parameters. The nature of randomness of fatigue failure crankshaft fails is a result of the following [28]:
1) Uncertainties in the model, such as the geometric parameters and material properties;
2) Randomness in loading vector intervals;
3) Rainflow data analysis replicating variable amplitude loadings; and
4) Faults that result from the component manufacturer.
These 4 stochastic aspects can be further enhanced as shown in Fig. 3 by considering the randomness as load-time history which in
return is dependent on the material properties and geometric shape of the crankshaft.
Uncertainties in respect of material properties have their origin in the manufacturing process, model choice and parameters,
modelling mechanism, data measurement and numerical evaluations. The variable amplitudes based on load-time history serves as an
input for mentioned uncertainties in assessing fatigue damage using the rainflow matrix. This approach is appropriate as in reality,
structures are subjected to service loadings based on the given probabilistic framework. Hence, making the stochastic approach as the
representative of the actual conditions.
The theory of stochastic processes advocates using a mathematical tool to quantifying variable loading. It uses a probabilistic
approach consisting of measures that define a sample space and a time function assigned to each outcome [34]. More recently the
extensive study of stochastic modelling under variable amplitude loading has proposed various models, each incorporating the
probabilistic technique. The stochastic models for fatigue damage has evolved in the direction of more feasible computational
mathematical methods using the Markov process, rather than lengthier and heavy computational techniques of Karhunen–Loeve,
Weiner integral, and Rayleigh techniques [35–37]. This is due to the fact that the Markov approach has the capability of generating
random but yet near similar data with lesser loading intervals that provides accurate relationship between experimental and si-
mulation data as shown by Bocchini et al. [38]. As probability relates to the physical condition of components, the reliability of
random fatigue crack growth under variable amplitude loading can model the effects of the loading sequence and correlate this to the
effects on the structure.

468
S.S.K. Singh et al. Engineering Failure Analysis 80 (2017) 464–471

4.2. Stochastic loading characteristics in fatigue failure

To understand the uncertainty in fatigue failure analysis, research using random or variable amplitude loading has been ex-
tensively explored. According to Sonsino [39], in 1939, Gassner first mentioned variable amplitude loading in conjunction with
procedures he was developing to represent load amplitude. The loading amplitude change is also referred to as spectrum or random
assignment. Experiments will typically use the information collected from tests using service load–time histories based on automobile
components. Fatigue damage testing can form a load block by comparing the amplitudes that lead to random fatigue damage and this
will effect of the random amplitude fatigue test.
Over the last decade, the use of random processes has been investigated, to assess the effects on crack growth life dispersion by the
history length and extreme values [40]. These types of models are based on random processes and observed load series which are
derived from the measurements which are representative of such processes. Under random loading, an assessment of the fatigue life
of notched specimens will mostly be based on the nominal stress approach. Notch analysis is also referred to as the Miner's damage
calculation or the local strain approach. It is a common practice in several fields involving fatigue design, to use standardised load-
time histories or load sequences to adequately understand the impact of variable amplitude loads on a structure. This type of test can
be beneficial for material selection and structural design, as the standardised load sequences are useful to compare the structural
integrity and prediction models associated with fatigue crack.
To aid material selection and structural design, comparing load sequences in tests helps to demonstrate structural integrity and
validate fatigue crack prediction models. The comparative tests are based on Gaussian random processes, which are continued until
failure takes place. This is attributed to sensitivity to extreme values and length of the load history, based on a typical stress–time
series. The process does not involve real and measured load data, so there is no statistical validation content in the observed load
sequences as it makes use of random processes. The use of the probabilistic model results in scatter in the loads that are relevant to
the operational behaviour of automobile components.

4.3. Importance of assessing low-cycle stochastic fatigue damage

Engineering practice has recently moved its attention to strain-based fatigue analysis as it is acknowledged to be more appropriate
than the conventional stress-based analysis, due to random cyclic strain loading. Hence, the influences of mean stress for fatigue life
assessment under random loads can take place in all loading modes [41]. This includes the deterministic loading history. In return the
reliability can be predicted based on the fatigue damage assessment [42]. The breaking down into component strains of the complete
strain amplitude cycle is an essential step in strain life analysis of cyclic property data. The division includes plastic and elastic strain,
described in Eq. (3). Although high cycle fatigue only manages stress, as low cycle or strain life fatigue testing offers the opportunity
for stress and strain to be calculated from each other [43].

Δε Δεe Δεp
= +
2 2 2 (3)

Low-cycle fatigue and high energy impact loads are considered in the local stress–strain approach to the fatigue life assessment
using the Manson–Coffin equation, as shown in Eq. (4) [43]. This model does not consider the effects of mean stress. If nonzero
normal stress is present, where there are relatively low mean stress levels compared to the cyclic yield stress, there will be a
noticeable effect on the fatigue behaviour of materials. Moreover, the life of a component or structure can be significantly increased
or decreased depending on the mean stresses, including crack initiation and propagation in fatigue loading.
σ ′f
2 b 2 c
εCM = ⎛ ⎞ + ε′⎛ ⎞
E ⎝D⎠ ⎝D⎠ (4)

The Ramberg–Osgood equation is used to convert the obtained stresses to the strains in the local strain life approach. The equation
is applied to the components subjected to cyclic loading based on the material properties of the structure. The total strain technique
uses Eq. (5) to calculate the damage parameters, using the load cycle obtained from the rainflow analysis
1
σ σ n′
εt = + ⎛ ⎞
E ⎝ K′ ⎠ (5)

where K and n are properties of the material.


This is a broadly accepted method to analyse material fatigue behaviour of a structure or component where there is cyclic loading,
and the response is observed regarding the plastic strain amplitude.
The above approaches are all empirically-based estimations of the strain-life fatigue, and test data should be compared to de-
termine the most appropriate model for the material and test conditions. It is essential for accuracy that an estimation method is
entirely appropriate bearing in mind the conditions, as the mean stress effect correction will depend on the number of cycles to
failure.
The Morrow model indicates that a tensile mean stress would reduce the fatigue strength coefficient and is increased by the mean
compressive stress as illustrated through Eq. (6). The Smith, Watson and Topper model is shown in Eq. (7), was modelled by Karakas
[44] to quantify the effects on fatigue behaviour of mean stresses, especially in conjunction with grey cast iron. This model relates the
cyclic strain range and the maximum stress in machine load mean stress, whereby it is used as a correction factor in uniaxial loading

469
S.S.K. Singh et al. Engineering Failure Analysis 80 (2017) 464–471

conditions. This model is applied to parameters of multiaxial loading during a loading cycle, based on the total fatigue strain
amplitude and the maximum tensile stress.
σ ′f − σm 2 b 2 c
εa = ⎛ ⎞ + ε′f ⎛ ⎞
E ⎝D⎠ ⎝D⎠ (6)

(σ ′f )2 2 2b 2 b+c
σmax εa = ⎛ ⎞ + σ ′f ε′f ⎛ ⎞
E ⎝D⎠ ⎝D⎠ (7)

where b = fatigue strength exponent; c = fatigue ductility exponent; σm = mean stress; σ'f = fatigue strength coefficient;
ε'f = fatigue ductility coefficient; Nf = cycle life; and ε'f = strain amplitude.

4.4. Recent computational advances in assessing stochastic fatigue failure

Computational methods are widely recognised in engineering to measure stochastic structural mechanics [45]. However, there
have been issues raised concerning the accuracy, as stochastic analyses are highly complex compared to deterministic methods.
Procedures based on the Monte Carlo simulation have been particularly under scrutiny, especially the concept of how statistical
samples are generated [46]. Computation enables statistical information of the stochastic model, such as the correlation, distribution,
and spectral density, to be determined. The stochastic algorithm has been applied to many loads to account for the randomness of the
structural responses. It shows a convergence between the numerical and analytical models that was like the stochastic model. This
was supported by comparing the expected discrepancy between model prediction and the statistical analysis, which can identify the
quantity required for a particular model simulation [47].
To estimate fatigue properties, a practical expert system was developed by Jeon [48] to predict the fatigue properties of metallic
materials. But was later added with variable load for fatigue life prediction by Kim [49]. However, it does have limitations and is
restricted to analysing fatigue life under variable loading. Modification of the S–N curve was also developed, to include a damage rule
based on the load interaction effect under variable loading. The methods deployed by the expert system consider fatigue life pre-
diction by analysing the load history and estimating material properties that anticipate fatigue damage. The prediction of the fatigue
life of the material is achieved but however, is restricted to the actual component loading history from the experimental analysis.
The proposed computational methodology is restricted as it focuses on the deterministic approach and fatigue failure is stochastic.
The stochastic modelling technique can follow either probabilistic, deterministic or both approaches and has widespread application
as mentioned by Ossai [50]. The suggestion of a stochastic approach is proposed that can generated random yet near similar data to
the experimental results. The importance of the expert system is its ability through deterministic and stochastic methods, to predict
the fatigue life cycle or imminent fatigue damage. The proposed establishment of an expert system can incorporate data generation,
fatigue damage, and a reliability assessment, leading to a real understanding of structures and components. In the stochastic process,
using relevant continuous time or discrete time enables a transformation of the load data into equal time steps. Statistical root mean
square error and correlation line order can be to test the level of divergence of data in the stochastic model and the accuracy of failure
predictions. This enables accurate prediction in the S–N or ε–N curves compared to the experimental and predicted fatigue life.

5. Conclusion

The fluctuating load–time histories corresponded to a stochastic fatigue failure that lead to structural integrity. But due to missing
data and the sensitivity limitations of the strain gauge in capturing real time loading data for the crankshaft, the stochastic has been
reviewed for fatigue life prediction. As agreed by earlier published works, the experimental studies have been employed in addressing
the fatigue failure issue in terms of forensic engineering, but yet failures continues to occur. The failure of crankshaft is fundamentally
dependent with uncertainty of loading history, material properties and its geometric design. With a need to minimise failures, the
assessment of fluctuating during service loading as it provides a safe life approach (i.e. S-N, ɛ-N) for durability prediction.
This is important as to select an appropriate method to avoid inaccurate estimations. Also, it is defined to be a stochastic failure
depending on fluctuating load as an input parameter which is supported by sub-factors of the materials fatigue properties and
geometric design. The available stochastic methods for improving sensitivity limitations in capturing real time loading data may lead
better durability prediction, due to its accuracy and efficiency. From the expressed views of the authors, the stochastic method
provides a feasible model in improving the fatigue life prediction using fluctuating load–time histories. This helps to motivate the
study for improving the durability through more stochastic modelling and experimental especially in the state of random or fluc-
tuating loading.
This is the platform upon which stochastic fatigue failure can be understood, to avoid incorrect predictive estimates on the
structural integrity of a component of the system.

Acknowledgement

The authors would like to acknowledge the support of the Ministry of Higher Education Malaysia (HLP-KPT.B.600-2/3-
781226085655)for funding this study and the National Automobile Industry for providing access to information to ensure its suc-
cessful conclusion.

470
S.S.K. Singh et al. Engineering Failure Analysis 80 (2017) 464–471

References

[1] Z. Domazet, F. Luksa, M. Bugarin, Failure of two overhead crane shafts, Eng. Fail. Anal. 44 (2014) 125–135.
[2] M. Fonte, P. Duarte, L. Reis, M. Freitas, V. Infante, Failure mode analysis of two crankshafts of a single cylinder diesel engine, Eng. Fail. Anal. 56 (2015) 185–193.
[3] D.H. Jung, H.J. Kim, P. Young-Shik, A. Gafurov, C. Gue-Cheol, A. Jong-Mo, Reliability prediction of the fatigue life of a crankshaft, J. Mech. Sci. Technol. 23
(2009) 1071–1074.
[4] A. Ktari, N. Haddar, H.F. Ayedi, Fatigue fracture expertise of train engine crankshafts, Eng. Fail. Anal. 8 (3) (2011) 1085–1093.
[5] J.A. Becerra, F.J. Jimenez, M. Torres, D.T. Sanchez, E. Carvajal, Failure analysis of reciprocating compressor crankshafts, Eng. Fail. Anal. 18 (2011) 735–746.
[6] M. Fonte, M. Freitas, Marine main engine crankshaft failure analysis: a case study, Eng. Fail. Anal. 16 (6) (2009) 1940–1947.
[7] M.A. Alfares, A.H. Falah, A.H. Elkholy, Failure analysis of a vehicle engine crankshaft, J. Fail. Anal. Prev. 7 (1) (2007) 12–17.
[8] J. Wannenburg, P.S. Heyns, A.D. Raath, Application of a fatigue equivalent static load methodology for the numerical durability assessment of heavy vehicle
structures, Int. J. Fatigue 31 (10) (2009) 1541–1549.
[9] Y. Liu, S. Mahadevan, Stochastic fatigue damage modeling under variable amplitude loading, Int. J. Fatigue 29 (2007) 1149–1161.
[10] Y. Ling, C. Shantz, S. Mahadevan, S. Sankararaman, Stochastic prediction of fatigue loading using real-time monitoring data, Int. J. Fatigue 33 (7) (2011)
868–879.
[11] F.H. Montazersadgh, A. Fatemi, Dynamic load and stress analysis of a crankshaft, J. Constr. Steel Res. 53 (2007) 1–8.
[12] X. Lei, G. Zhang, S. Xigeng, J. Chen, G. Dong, Simulation on the motion of crankshaft with crack in crankpin-web fillet region, J. Sound Vib. 295 (3–5) (2006)
890–905.
[13] K.S. Choi, J. Pan, Simulations of stress distributions in crankshaft sections under fillet rolling and bending fatigue tests, Int. J. Fatigue 31 (3) (2009) 544–557.
[14] E. Jensen, Crankshaft strength through laboratory testing, SAE Tech. Pap. 700526 (1970).
[15] S. Shimizu, K. Tosha, K. Tsuchiya, New data analysis of probabilistic stress–life (P–S–N) curve and its application for structural materials, Int. J. Fatigue 32
(2010) 565–575.
[16] D. Taylor, A.J. Ciepalowicz, P. Rogers, J. Devlukia, Prediction of fatigue failure in a crankshaft using the technique of crack modelling, Fatigue Fract. Eng. Mater.
Struct. 20 (1997) 13–21.
[17] D. Taylor, A.J. Ciepalowicz, W. Zhou, J. Devlukia, Mixed mode fatigue from stress concentrations: an approach based on equivalent stress intensity, Int. J.
Fatigue 21 (1999) 173–178.
[18] R.K. Pandey, Failure of diesel-engine crankshafts, Eng. Fail. Anal. 10 (2) (2003) 165–175.
[19] W. Changli, C. Zhao, D. Wang, Analysis of an unusual crankshaft failure, Eng. Fail. Anal. 12 (3) (2005) 465–473.
[20] B. Kareem, Evaluation of failures in mechanical crankshafts of automobile based on expert opinion, Eng. Fail. Anal. 3 (2015) 25–33.
[21] P. Spiteri, S. Ho, Y.L. Lee, Assessment of bending fatigue limit for crankshaft sections with inclusion for residual stresses, Int. J. Fatigue 29 (2007) 318–329.
[22] Wei Li, Qing Yan, Jianhua Xue, Analysis of a crankshaft fatigue failure, Eng. Fail. Anal. 55 (2015) 139–147.
[23] M. Fonte, B. Li, L. Reis, M. Freitas, Crankshaft failure analysis of a motor vehicle, Eng. Fail. Anal. 35 (2013) 147–152.
[24] J.A. Becerra Villanueva, E.F. Jiménez, F. Cruz-Peragón, G.M. Torres, A methodology for cracks identification in large crankshafts, Mech. Syst. Signal Process. 25
(8) (2011) 3168–3185.
[25] S.H. Hyung, H.L. Kyoung, H.P. Sung, Parametric study to identify the cause of high torsional vibration of the propulsion shaft in the ship, Eng. Fail. Anal. 59
(2016) 334–346.
[26] O. Asi, Failure analysis of a crankshaft made from ductile cast iron, Eng. Fail. Anal. 13 (8) (2006) 1260–1267.
[27] X. Chen, X. Yu, R. Hu, J. Li, Statistical distribution of crankshaft fatigue: experiment and modelling, Eng. Fail. Anal. 4 (2014) 210–220.
[28] B. Sudret, Z. Guédé, Probabilistic assessment of thermal fatigue in nuclear components, Nucl. Eng. Des. 235 (17–19) (2005) 1819–1835.
[29] P. Johannesson, On rainflow cycles and the distribution of the number of interval crossings by a Markov chain, Probab. Eng. Mech. 17 (2002) 123–130.
[30] D. Benasciutti, R. Tovo, On fatigue cycle distribution in non-stationary switching loadings with Markov chain structure, Probab. Eng. Mech. 25 (4) (2006)
406–418.
[31] J.R. Bisping, B. Peterwerth, C. Bleicher, R. Wagener, T. Melz, Fatigue life assessment for large components based on rainflow counted local strains using the
damage domain, Int. J. Fatigue 68 (2014) 150–158.
[32] A. Carpinteri, A. Spagnoli, S. Vantadori, Multiaxial fatigue assessment using a simplified critical plane-based criterion, Int. J. Fatigue 33 (2011) 969–976.
[33] C. Mattrand, J.M. Bourinet, Random load sequence and stochastic crack growth based on measured load data, Eng. Fract. Mech. 78 (2011) 3030–3048.
[34] W. Wagner, Stochastic, analytic and numerical aspects of coagulation processes, Math. Comput. Simul. 62 (2003) 265–275.
[35] D.P. Kihl, S. Sarkani, J.E. Beach, Stochastic fatigue damage accumulation under broadband loadings, Int. J. Fatigue 17 (5) (1995) 321–329.
[36] A. Ray, S. Targilara, A nonlinear stochastic model of fatigue crack dynamics, Probab. Eng. Mech. 12 (1) (1997) 33–40.
[37] H. Shen, J. Lin, E. Mu, Probabilistic model on stochastic fatigue damage, Int. J. Fatigue 22 (2000) 569–572.
[38] P. Bocchini, D. Saydam, D.M. Frangopol, Efficient, accurate, and simple Markov chain model for the life-cycle analysis of bridge groups, Struct. Saf. 40 (2013)
51–67.
[39] C.M. Sonsino, Course of SN-curves especially in the high-cycle fatigue regime with regard to component design and safety, Int. J. Fatigue 29 (2007) 2246–2258.
[40] W. Wu, C. Ni, Statistical aspects of some fatigue crack growth data, Eng. Fract. Mech. 74 (2007) 2952–2963.
[41] A. Plumtree, H.A. Abdel-Raouf, Cyclic stress–strain response and substructure, Int. J. Fatigue 23 (2001) 799–805.
[42] C. Hongxia, C. Yunxia, Y. Zhao, Coupling damage and reliability model of low-cycle fatigue and high energy impact based on the local stress–strain approach,
Chin. J. Aeronaut. 27 (4) (2014) 846–855.
[43] A. Niesłony, C. Dsoki, H. Kaufmann, P. Krug, New method for evaluation of the Manson–Coffin–Basquin and Ramberg–Osgood equations with respect to
compatibility, Int. J. Fatigue 30 (2008) 1967–1977.
[44] O. Karakas, Consideration of mean-stress effects on fatigue life of welded magnesium joints by the application of the Smith–Watson–Topper and reference radius
concepts, Int. J. Fatigue 49 (2013) 1–17.
[45] X.F. Xu, L. Graham-Brady, A stochastic computational method for evaluation of global and local behavior of random elastic media, Comput. Methods Appl. Mech.
Eng. 194 (2005) 4362–4385.
[46] A. Pasanisi, S. Fua, N. Bousquet, Estimating discrete Markov models from various incomplete data schemes, Comput. Stat. Data Anal. 56 (2012) 2609–2625.
[47] J. Mullins, S. Mahadevan, Variable-fidelity model selection for stochastic simulation, Reliab. Eng. Syst. Saf. 131 (2014) 40–52.
[48] W.S. Jeon, J.H. Song, An expert system for estimation of fatigue properties of metallic materials, Int. J. Fatigue 24 (2002) 685–698.
[49] Y.H. Kim, J.H. Song, J.H. Park, An expert system for fatigue life prediction under variable loading, Expert Syst. Appl. 36 (2009) 4996–5008.
[50] C.I. Ossai, B. Boswell, I. Davies, Markov Chain modelling for time evolution of internal pitting corrosion distribution of oil and gas pipelines, Eng. Fail. Anal. 60
(2016) 209–228.

471

También podría gustarte