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

net/publication/245571559

Reliability Analysis and Preventive Maintenance for Fatigue Life of End Beam
for Uncovered Freight Car

Article  in  Transactions of the Korean Society of Mechanical Engineers A · March 2005

DOI: 10.3795/KSME-A.2005.29.3.495

CITATIONS READS

4 230

5 authors, including:

Seok Heum Baek

DNDE Inc. South Korea
36 PUBLICATIONS   341 CITATIONS   

SEE PROFILE

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

Shape Optimization of Impeller Blades for Bidirectional Axial Flow Pump View project

All content following this page was uploaded by Seok Heum Baek on 02 November 2020.

The user has requested enhancement of the downloaded file.

 Journal of
Mechanical
Science and
Technology
Journal of Mechanical Science and Technology 23 (2009) 1225~1233
www.springerlink.com/content/1738-494x
DOI 10.1007/s12206-008-0901-z

Reliability design of preventive maintenance scheduling for

cumulative fatigue damage†
Seok-Heum Baek1, Seok-Swoo Cho2, Hyun-Su Kim3* and Won-Sik Joo3
1
School of Mechanical Engineering, Dong-A University, Busan, 604-714, Korea
2
Department of Vehicle Engineering, Kangwon National University, Kangwon-do, 245-711, Korea
3
Department of Mechanical Engineering, Dong-A University, Busan, 604-714, Korea

(Manuscript Received March 12, 2007; Revised February 14, 2008; Accepted September 1, 2008)
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Abstract

As the cumulative running times of a locomotive truck increases, degradation such as fatigue, wear, and deterioration
occur. Particularly the container train and uncovered freight train, their maintenance cost during their lifetime is three
times more than the manufacturing cost. Generally, for the freight train, corrective maintenance to repair a bad part
after a breakdown is not adapted; however, preventive maintenance that fixes a bad part before a breakdown is. There-
fore, it is important and necessary to establish a system of optimal preventive maintenance and exact maintenance pe-
riod. This study attempts to propose a preventive maintenance procedure that predicts a repair period using reliability
function and instantaneous failure rate based on fatigue test and load history data. Additionally, this method is applied
to the end beam of an uncovered freight train, which is the brake part, and its usefulness is examined and analyzed.

Keywords: Preventive maintenance; Cumulative fatigue damage; Miner’s rule; Maximum likelihood estimator; Kaplan-Meier estimator
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

objectives.
1. Introduction
According to the time of maintenance executed,
The performance of locomotive truck frames de- maintenance is usually divided into two major catego-
creases as the cumulative running time increases. Its ries: corrective maintenance (CM) and preventive
causes are expected to be fatigue, wear, and deteriora- maintenance (PM) [6-11]. The former corresponds to
tion. New inspection and repair methods are needed the actions that occur after the system breaks down,
for their safety because freight trains require higher while the latter corresponds to the actions that take
speed and longer running time. In the case of con- place while the system is operating. Most investiga-
tainer and uncovered freight trains, the maintenance tions on preventive maintenance have been limited by
cost is about three times more than their manufactur- regular inspection time or components management.
ing cost during their lifetime [1, 2]. Therefore, it is Since the performance test or endurance test of freight
important to establish an optimal preventive mainte- trains is achieved in proving railroad, its reliability
nance schedule to minimize the total cost with the estimation is very difficult to obtain.
desired or specified levels of operational safety and Reliability estimation function can be predicted by
reliability [3-5]. It can also maximize the availability probability density function, cumulative distribution
of the components and optimize some other specified function, or hazard rate function, which is a function
†This paper was recommended for publication in revised form by Associate of replacement time (or fatigue crack initiation time)
Editor Chongdu Cho based on the maintenance records. The preventive
*
Corresponding author. Tel.: +82 51 2007651, Fax.: +82 51 2007656
maintenance of freight trains is very useful because
E-mail address: [email protected]
© KSME & Springer 2009 the demand prediction of their parts through reliabil-
1226 S.-H. Baek et al. / Journal of Mechanical Science and Technology 23 (2009) 1225~1233

ity estimation allows the acquisition of spare items.

This study discusses the preventive maintenance
method based on actual load history. Fatigue life is
estimated by the probability distribution of life data
from the durability analysis process. Furthermore, the
reliability function and hazard rate function are evalu-
ated by the Kaplan-Meier method [12-14]. Finally,
this method is applied to the fatigue damage problem
of the end beam which is a brake part of an uncovered (a) (b)
freight train to examine the usefulness of preventive Fig. 1. Rainflow cycle counting.
maintenance.
and histogram of the cycle ranges. For the infinite life
2. Cumulative Fatigue Damage Evaluation design for very high mean stresses, the Buch mean
2.1 Definition of cumulative fatigue damage stress correction is selected. Palmgren-Miner rule is
expressed as follows. Failure is expected to occur if
The end beam of a freight train is subjected to vari-
able amplitude service loading. The measurements of n1 n n n
the service stress (or strain) histories are required to D= + 1 + 1 +L= ∑ i ≥1 (1)
N f1 N f1 N f1 i N fi
obtain general information on service loading and to
determine the fatigue life of the end beam. Signal
processing uses cycle counting algorithms to extract where ni is the number of applied cycles and N fi is
stress-strain hysteresis loops quickly and accurately. the number of cycles to failure at a specified stress
In this study, the rainflow cycle counting [15-18] is amplitude σ i , respectively.
used as a signal processing method for fatigue analy- Since in many cases the Palmgren-Miner rule leads
sis. to non-conservative life predictions, the linear dam-
Fig. 1 shows the procedure for the cycle counting age rule associated with the critical damage sum D,
method as demonstrated by Downing and Socie [15]. which is other than 1, has been proposed in many
design codes for fatigue damage assessment of struc-
[STEP1] Consider the following sequence of peaks/ tures subjected to variable amplitude loading. In this
valleys. The notation uses point A as the most recent study, the critical cumulative damage value of D is 1
data point, point B as the previous point and so on in Eq. (1).
(Fig. 1(a)).
2.2 Fatigue reliability assessment
[STEP2] A cycle is closed because the range from A
to B is greater than the range from B to C and is rep- Reliability assessment prevents products from pre-
resented by the range from B to C. mature failure. There are two parameter estimation
methods for statistical data according to life distribu-
Range A to B ＞ Range B to C tion assumption: parametric distribution analysis and
nonparametric analysis. In the case that fatigue life
[STEP3] Fig. 1 (b) shows a new cycle. Previously, data is based on the Weibull distribution, shape pa-
the range from A to B is greater than that from B to C; rameter m and scale parameter η are determined by
therefore, B to C is a cone cycle. This procedure is maximizing the loglikelihood function. This method
repeated until no more cycles are closed at this point. is called maximum likelihood estimation [14, 21].
The maximum likelihood estimator is determined as
The cumulative fatigue damage rule for the actual follows:
running environment is necessary for estimating the
safety of railway vehicles. The extracted cycle pro- r
⎛ n ⎞
duces stress amplitude σ a and mean stress. Cumula- ∑ ln t i ⎜ ∑ ln ti ln ti
m
⎟ 1
i =1
=⎜ n
i =1
⎟− (2)
tive damage D and number of cycles to fractures N r ⎜ ⎟ m
⎜ ∑ ln ti
m

are determined using the Palmgren-Miner rule [19, 20] ⎟

⎝ i=1 ⎠
 S.-H. Baek et al. / Journal of Mechanical Science and Technology 23 (2009) 1225~1233 1227

1/ m
⎛ n m⎞
⎜ ∑ ln ti ⎟
η = ⎜ i =1 ⎟ (3)
⎜ r ⎟
⎜⎜ ⎟⎟
⎝ ⎠

where r is the number of data observed. Reliability

function, cumulative distribution function, and prob-
ability density function are defined to perform reli-
ability estimation. Their functions for the Weibull Fig. 2. Apparatus for Schenck-type bending fatigue testing.
probability distribution can be expressed as Eqs. (4)

-
to (6), respectively. 300
T-type welding, R= 1
Fracture
250 No Fracture

Stress amplitude ∆s / 2 ,MPa

⎡⎛ t ⎞ m ⎤ 200
⎡1 8 ⎤
1 .64σ (log N ) = 1 .64 ⎢ ∑ {log N i − (αˆ 1 + βˆ1 log S i )} 2 ⎥
1/ 2

R (t ) = exp ⎢⎜ − ⎟ ⎥
⎣ 6 i =1
(4) = 0.405
⎦

⎢⎣⎝ η ⎠ ⎥⎦ 150
P = 95%
50%
5%
⎡⎛ t ⎞ m ⎤ 100
F (t ) = 1 − exp ⎢⎜ − ⎟ ⎥ (5)
⎢⎣⎝ η ⎠ ⎥⎦ 50
52.8

Regression equation of inclination for P=50% reliability ;

m −1 ⎡⎛ t ⎞ m ⎤
m⎛ t ⎞ 0 log N = αˆ + βˆ S ± 1.64σˆ (log N )
exp ⎢⎜ − ⎟ ⎥
1 1

f (t ) = ⎜ ⎟ (6)
η ⎝η ⎠ ⎢⎣⎝ η ⎠ ⎥⎦ 4
10 10
5
10
6
10
7

Number of cycles to failure

Furthermore, the hazard rate function (instantane- Fig. 3. P-S-N curve for SS400 steel.
ous failure rate function) is the ratio of probability
3. Freight train preventive maintenance sched-
density function to reliability function. It is expressed
uling
as Eq. (7).
3.1 P-S-N curve of SS400 steel
m −1
m⎛ t ⎞ The material of the end beam for an uncovered
h(t ) = ⎜ ⎟ (7)
η ⎝η ⎠ freight train is SS400 steel. Its fatigue test was per-
formed by the Schenck-type twisting and bending
The Kaplan-Meier estimator [12, 14] for fatigue fatigue testing machine under the maximum bending
data with censored observations is used in this study. moment of 4 kg·m, frequency of 1,800 rpm, and
It is compared with the median that corresponds to stress ratio of -1 (Fig. 2). The static and dynamic test
more than 50% of the failure probability. The Kaplan- results of the T-type welded specimens are introduced
Meier estimator of hazard rate function and reliability in the references [2]. See references for further details.
function are expressed as Eqs. (8) and (9), respec- A P-S-N (probabilistic-stress-life) curve can be ob-
tively. tained from JSME S002.
Fig. 3 shows the P-S-N curve for SS400 steel. The
⎛ dj ⎞ relationship between stress amplitude and fatigue life
R (t ) = ∏ (1 − h j (t ) ) = ∏ ⎜1 −
⎜ nj
⎟
⎟
(8) is given in Eq. (10).
j∈t j < t j∈t j < t ⎝ ⎠
dj log N = 6.728 − 0.0094∆s / 2 ± 0.405 (10)
h j (t ) = (9)
nj
The mean of the fatigue limit by the JSME statisti-
cal S-N testing method was 52.8 MPa.
where n is the total number of units, nj and dj denote
the number of subjects that survive just before time tj
3.2 Estimation of loading history
and the number of failure occurring at time tj , respec-
tively. It is very difficult to change the size of the end beam
1228 S.-H. Baek et al. / Journal of Mechanical Science and Technology 23 (2009) 1225~1233

Fig. 4. Photograph of a fractured end beam.

Fig. 5. Attachment location and number of strain gauges.

It is very difficult to change the size of the end

beam because it is built into the lower part of the
truck frame. Its structural redesign cannot be per-
formed to prevent fatigue fracture in the viewpoint of
safety and economy. Therefore, if the fatigue proper-
ties of the end beam are understood, planned preven- Fig. 6. Strain time-history in 1,500 seconds.
tive maintenance interval can be determined.
Fig. 4 shows the end beam supporting brake system principal stress direction. A one axis strain gauge
and its fatigue failure site. Fatigue crack in the end (KFG-5-128) was installed on the end beam before
beam was initiated at the welded gusset plate edge loading, and the test track was the Donghae-Jecheon
and was propagated in the vertical direction of run- section. The load history in the test track was meas-
ning direction. The load history through the structure ured at 16.7 m/s over 1,500 seconds from start to
test of the truck frame was measured to predict fa- braking. Fig. 6 shows the results of the test series with
tigue damage. The load history in the running section the load history based on the six strain gauges. The
included the start, acceleration, and brake sections. distribution of the stress range at each location of
The running section was the place where fatigue fail- strain gauges is listed in Table 1.
ure of the end beam occurred. FE-safe [22], a commercial fatigue analysis pro-
Fig. 5 shows the attachment locations of the six gram, was used to calculate the fatigue life of the end
strain gauges with a direction perpendicular to the beam. Palmgren-Miner rule was used as the fatigue
fatigue crack. In this study, we assumed that the di- cumulative damage rule. The first counting data for
rection perpendicular to the crack propagation is the the stress level was determined within the confidence
 S.-H. Baek et al. / Journal of Mechanical Science and Technology 23 (2009) 1225~1233 1229

Table 1. Results of the strain gauge measurement.

Stress
G1 G2 G3 G4 G5 G6
(MPa)
Min. 0.51 36.26 36.75 70.65 31.14 3.41
Max. 7.74 48.26 51.62 94.14 72.02 6.96
Mean 3.71 8.47 8.83 13.28 0.13 0.4

Table 2. Expected fatigue life and cumulative fatigue damage

at each measured location.

Location
Modified Palmgren-Miner rule
number
Fig. 7. Result of the rainflow cycle counting histogram.
Mean stress
Life Damage Year
correction
G1 Unlimited 0 Unlimited
G2 983,700 1.02×10-6 46.79
G3 360,400 2.77×10-6 17.14
None
G4 209,800 4.77×10-6 7.98
G5 5,146,000 1.94×10-7 244.78
G6 170×106 5.88×10-9 8,086
G1 23.2×106 4.31×10-8 4,833
G2 35,350 2.83×10-5 7.36
G3 36,550 2.74×10-5 7.61
Goodman
G4 14,010 7.14×10-5 2.92
G5 50,780 1.97×10-5 10.58
G6 11.5×106 8.69×10-8 2,397

Fig. 8. Result of the rainflow cycle counting histogram.

interval of the P-S-N curve by a correction method
for the curve that considers stresses under fatigue life expectancy is 7.36 years.
limit.
Fig. 7 shows the cycle range-mean histogram lo-
3.3 Reliability analysis
cated at G4. Fig. 8 shows the 2-dimension contour
line of the fatigue damage for this case. It should be Reliability estimation for life data was imple-
noted that although the high amplitude stress cycle mented using the statistical analysis software
has low frequency, fatigue damage is relatively large. MINITAB R14 [23] to predict the probabilistic fa-
When a freight train is broken, high amplitude stress tigue life of the end beam.
cycle can occur. Therefore, the breaking load influ- Fig. 9 shows the goodness-of-fit test of the life data.
ences the fracture of the end beam. Weibull distribution, normal distribution, and log-
The stress cycle with a mean stress correction pro- normal distribution were used as the candidate statis-
duces a shorter fatigue life than the stress cycle with tical distributions of the life data [24, 25]. The Ander-
no mean stress correction. Table 2 shows the fatigue son-Darling value that evaluates the difference be-
life and damage at all strain gauge locations using the tween test data and approximated data was selected as
modified Palmgren-Miner rules. the test statistics. The locations of G1 and G6 had
The stress cycles and fatigue damage at the G2 lo- infinite fatigue life; therefore, they did not influence
cation were 35,350 cycles and 2.83×10-5, respectively. the fatigue life of the end beam. Random sample data
Fatigue life at the location was expected to be were generated for four life data except those above
53,025,120 seconds because the stress history was two. The goodness-of-fit test and reliability analysis
measured during 1,500 seconds. If we suppose that were performed for those. Fig. 9 (b) is the result of
the running speed of a freight train is 16.7 m/s to con- the goodness-of-fit test for random data. Among the-
vert time life to cycle life, fatigue life corresponds to ses distributions, samples could be approximated by
883,750,000 m. Furthermore, if we consider that the the Weibull distribution because the test statistics of
design life of a freight train is 25 years (3×109 m), its the population parameter was 0.808.
1230 S.-H. Baek et al. / Journal of Mechanical Science and Technology 23 (2009) 1225~1233

(a) Experiment data

(b) Random data

Fig. 9. Result of the goodness-of-fit test.

100
Parametric
reliability function
95% lower bound
80
95% upper bound
Reliability function, %

60

40

20

0 2 4 6 8 10 12 14
Fatigue life, year

(a) PDF (b) CDF (c) HF

Fig. 10. Reliability analysis of fatigue life using parametric method.

Fig. 10(a) shows the reliability function of the

Weibull distribution with a 95% confidence interval.
Fig. 10(b) shows cumulative distribution function
of the Weibull distribution with 95% confidence in-
terval. Fig. 10(c) shows the hazard rate function that
increases according to time. This expresses the in-
creasing failure rate (IFR) due to system wear and
fatigue. Fig. 11 shows the cumulative distribution
function with a 95% confidence interval on the
Weibull distribution probability paper. Since all data
fell within the 95% confidence interval, life data
could be approximated by the Weibull distribution. Its
shape parameter and scale parameter were 2.92, 7.24,
Fig. 11. Two-parameter Weibull plots of fatigue life data. respectively, and its median was 6.386 years. Table 3
 S.-H. Baek et al. / Journal of Mechanical Science and Technology 23 (2009) 1225~1233 1231

Table 3. Estimation of the two-parameter Weibull distribu- Table 4. Percentile and hazard function for cumulative failure
tion. probability.

95% Normal Cumulative failure

Standard confidence interval Percentile (year) Hazard function
Item Estimate probability (%)
error
Lower Upper
1 1.49824 0.01685
Shape 2 1.90291 0.02939
2.92021 0.606582 1.94359 4.38757
parameter 3 2.19018 0.03964
Scale 4 2.42119 0.04871
7.23963 0.675148 6.03026 8.69155
parameter 5 2.61811 0.057
Mean (MTTF) 6.45737 0.622304 5.34594 7.79986 6 2.79179 0.06476
7 2.94845 0.07211
Standard 8 3.09205 0.07914
2.40430 0.414068 1.71552 3.36963
deviation 9 3.22525 0.08591
Median 6.38570 0.667379 5.20293 7.83734 10 3.34997 0.09247
First quartile 4.72526 0.697779 3.53777 6.31136
Third quartile 8.09643 0.717753 6.80509 9.63281 summarizes the statistical characteristics of the fa-
Interquartile tigue life of the end beam. Furthermore, the probabil-
3.37117 0.588058 2.39496 4.74528
range ity distribution of life data was estimated by the Kap-
lan-Meier method to compare the parametric method
with the nonparametric method.
Fig. 12 and Fig. 13 show the hazard rate function
and reliability function, respectively. Accordingly, as
the hazard rate function increases, the reliability func-
tion decreases. Mean life was estimated to be more
than 1×103 year due to the effect of the locations G1
and G6: however, the median was 7.61 years. Com-
pared with the parametric method, this result showed
a big difference in the mean but a small difference in
the median (6.386 years).

3.4 Preventive maintenance periods

Failure of the end beam influences the time sched-

Fig. 12. Hazard function for fatigue life using Kaplan-Meier ule of freight trains and causes economic loss. There-
method. fore, it is very important to achieve preventive main-
tenance before failure accidents.
Preventive maintenance of existing freight trains
can be divided by 90-day inspections, 30-month in-
spections, and 60-month inspections. Important parts
such as the end beam and others are considered to be
part of the 30-month inspections.
Table 4 shows the percentile and hazard rate func-
tion of the cumulative hazard probability for the dis-
tribution. In Table 4, the percentile expressed in years
is the fatigue life that corresponds to the cumulative
failure probability. Table 4 presents the assurance
period fewer than the 10% failure probability to esti-
mate the assurance period or assurance expense. Ran-
dom failure probability was very high due to the local
Fig. 13. Reliability function for fatigue life using Kaplan- life distribution of the end beam.
Meier method. Therefore, the derating design is necessary to pre-
1232 S.-H. Baek et al. / Journal of Mechanical Science and Technology 23 (2009) 1225~1233

vent failure during the random failure period. Fur- Des., 12 (4) (1999) 614-621.
thermore, the middle inspection of the end beam [4] C. Marquardt and H. Zenner, Lifetime Calculation
should be performed within 1.5 years for planned under Variable Amplitude Loading with the Appli-
preventive maintenance. cation of Artificial Neural Networks, Int. J. Fatigue,
27 (8) (2005) 920-927.
[5] C. H. Bae, Y. Chu, H. J. Kim, J. H. Lee and M. W.
4. Conclusions
Suh, A Study on Maintenance Reliability Alloca-
This study performed the reliability analysis of fa- tion of Urban Transit Brake System Using Hybrid
tigue accumulation damage in a freight train using the Neuro-Genetic Technique, Journal of Mechanical
damage summation method. The main results are as Science and Technology, 21 (1) (2007) 32-47.
follows. [6] R. Cooke and J. Paulsen, Concepts for Measuring
(1) The fatigue life for the end beam of a freight Maintenance Performance and Methods for Analys-
train is predicted at a variable amplitude loading with ing Competing Failure Modes, Reliability Engineer-
the modified Miner hypothesis of cumulative damage. ing and System Safety, 55 (2) (1997) 135-141.
The longer the running time of the end beam is, the [7] I. Gertsbakh, Reliability Theory: With Application
higher the failure rate. Therefore, preventive mainte- to Preventive Maintenance, Springer-Verlag Berlin
nance is available if a specific failure rate is applied to Heidelberg New York, (2000).
the maintenance schedule. [8] S. E. Rigdon and A. P. Basu, Statistical Methods for
(2) The shape, scale, and median parameters in the the Reliability of Repairable Systems, Wiley, New
approximated Weibull distribution are 2.92, 7.24, and York, (2000).
6.386 years, respectively. The median above is simi- [9] R. K. Mobley, An Introduction to Preventive Main-
lar to the median evaluated by the Kaplan-Meier tenance, Second Edition, Butterworth-Heinemann,
method. (2002).
(3) Due to the different local life distributions of the [10] Y. Zhao, V. Volovoi, M. Waters and D. Mavris, A
end beam, random failure probability is very high. Sequential Approach for Gas Turbine Power Plant
Therefore, middle inspection of the truck frame Preventative Maintenance Scheduling, ASME J.
should be performed within 1.5 years (instantaneous Eng. Gas Turbines Power, 128 (4) (2006) 796-805.
failure rate of 1%) for planned preventive mainte- [11] Y. H. Park and J. Tang, Optimum Replacement
nance. Interval for Mechanical Components Based on Fa-
tigue Reliability, ASME J. Pressure Vessel Technol.,
129 (4) (2007) 683-688.
Acknowledgment
[12] B. W. Turnbull, Nonparametric Estimation of a
This study was supported by research funds from Survivorship Function with Doubly Censored Data,
Dong-A university. Journal of the American Statistical Association, 69
(345) (1974) 169-173.
[13] G. Li, On Nonparametric Likelihood Ratio Esti-
References
mation of Survival Probabilities for Censored Data,
[1] S. H. Baek, K. Y. Lee, S. S. Cho and W. S. Joo, Statistics & Probability Letters, 25 (2) (1995) 95-
Reliability-based Optimization for Fatigue Mainte- 104.
nance Planning of Freight Car, 10th International [14] W. Q. Meeker and L. A. Escobar, Statistical
Conference on the Mechanical Behavior of Materi- Method for Reliability Data, John Wiley & Sons,
als, The Mechanical Behavior of Materials X: Ses- (1998).
sion 14, Korea, (2007) 1369-1372. [15] M. Matsuishi and T. Endo, Fatigue of metals sub-
[2] S. H. Baek, S. S. Cho and W. S. Joo, Fatigue Life jected to varying stress-fatigue lives under random
Prediction Based on the Rainflow Cycle Counting loading. Proc. Kyushu District Meeting, JSEM, Fu-
Method for the End Beam of a Freight Car Bogie, kuoka, Japan, (1968) 37-40.
Int. J. Automotive Technology, 9 (1) (2008) 95-101. [16] S. D. Dowing and D. F. Socie, Simplified Rain-
[3] Z. H. Jiang, L. H. Shu and B. Benhabib, Steady- flow Cycle Counting Algorithms, Int. J. Fatigue, 4
State Reliability Analysis of Repairable Systems (1) (1982) 31-40.
Subject to System Modifications, ASME J. Mech. [17] H. Wang, N. H. Kim and Y. J. Kim, Safety Enve-
 S.-H. Baek et al. / Journal of Mechanical Science and Technology 23 (2009) 1225~1233 1233

lope for Load Tolerance and its Application to Fa- Cho is currently a Professor at the Vehicle Engineering at
tigue Reliability Design, ASME J. Mech. Des., 128 Kangwon National University in Kangwon-do, Korea. Dr.
(4) (2006) 919-927. Cho works on crack growth modeling and composite
[18] C. H. McInnes and P. A. Meehan, Equivalence of design and optimization, and his research interests are in
Four-point and Three-point Rainflow Cycle Count- the areas of structural optimization and inverse and identi-
ing Algorithms, Int. J. Fatigue, 30 (3) (2008) 547- fication problems, and X-ray diffraction, brittle collapse
559. and crack propagation, fatigue fracture phenomena.
[19] M. A. Miner, Cumulative Damage in Fatigue,
Journal of Applied Mechanics, 12 (1945) 159-164. Hyun-Su Kim received a B.S.
[20] J. J. Xiong and R. A. Shenoi, A Reliability-based degree in Mechanical Engineering
Data Treatment System for Actual Load History, from Seoul National University in
Fatigue Fract. Engng. Mater. Struct., 28 (10) 1978. He then went on to receive
(2005) 875-889. his M.S. from KAIST in 1980 and
[21] W. Nelson, Applied Life Data Analysis, John Ph.D. degree from University of
Wiley & Sons, (1982). Iowa in 1989. Dr. Kim is cur-
[22] Fe-safe Software Package Ver. 5, Volume 3- rently a Professor at the Mecha-
Signal Processing Reference Manual, Section 7, 1- nical Engineering at Dong-A University in Busan, Korea.
14, Safe Technology Limited, (2003). His research interests are in the area of high temperature
[23] MINITAB User’s Guide #2, Data Analysis and creep fatigue, bio-engineering, design using the finite
Quality Tools, Minitab Inc, (2000). element method, optimization, and MEMS.
[24] R. B. D'Agostino and M. A. Stephens, Goodness-
of-Fit Techniques, Marcel Dekker, (1986). Won-Sik Joo received a B.S.
[25] E. Castillo, M. Lopez-Aenlle, A. Ramos, A. Fer- degree in Mechanical Engineering
nandez-Canteli, R. Kieselbach and V. Esslinger, from Dong-A University in 1968.
Specimen Length Effect on Parameter Estimation in He then went on to receive his
Modelling Fatigue Strength by Weibull Distribution, M.S. from Dong-A University in
Int. J. Fatigue, 28 (9) (2006) 1047-1058. 1978 and Ph.D. degree from
Kookmin University in 1988. Dr.
Joo is currently a Professor at the
Seok-Heum Baek received a B.S. Mechanical Engineering at Dong-A University in Busan,
and M.S. degree in Mechanical Korea. His research interests are in the area of creep and
Engineering from the Dong-A fatigue in high temperature alloy, fatigue design, and
University in 2001 and 2003, strength evaluation.
respectively. He is currently a
Ph.D. student at the School of
Mechanical Engineering at Dong-
A University in Busan, Korea.
Student Baek works on ceramic composite design and
robust and reliability-based design, and his research inter-
ests are in the areas of trade-off analysis in multicriteria
optimization, design under uncertainty with incomplete
information, and probabilistic design optimization.

Seok-Swoo Cho received a B.S.

degree in Mechanical Engineering
from Dong-A University in 1991.
He then went on to receive his
M.S. from Dong-A University in
1993 and Ph.D. degree from
Dong-A University in 1997. Dr.

View publication stats