You are on page 1of 18

Study on Rolling Stock Maintenance Strategy and Spares Parts Management

Yung-Hsiang Cheng 1, Ann Shawing Yang 2 , Hou-Lei Tsao 3


1
National Kaohsiung First University of Science and Technology, Kaohsiung City, Taiwan,
2
Shu-Te University, Kaohsiung County, Taiwan, 3 National Kaohsiung Fist University of
Science and Technology, Kaohsiung City, Taiwan

Abstract

The purpose of this paper was to present a method for rolling stock’s maintenance strategy
selection that allows for the consideration of important interactions among decisions levels and
criteria. The methodology adopts Analytic Network Process (ANP) for this evaluation to decide the
possible ratio between preventive maintenance and corrective maintenance that can induce
possible spares parts quantities and replacement interval of component of rolling stock. The
empirical result indicates preventive maintenance should be much more emphasized than
corrective maintenance. Safety is the most crucial factor for rolling stock maintenance strategy
selection.

Key words: rolling stock maintenance, maintenance strategy, spares part

Introduction

The maintenance of rolling stock can be categorized in two types: failure based maintenance
(corrective maintenance) and life based maintenance (preventive Maintenance). The time interval
at which the preventive maintenance could be scheduled is dependent on both the life distribution
of the components and the total cost involved in the maintenance activity. However, the corrective
maintenance cannot be avoided when a random failure of a component occurs. The total cost of the
maintenance depends on p ercentages in performing preventive maintenance and corrective
maintenance.

In addition, contrary to maintenance strategy selection in manufacturing industries, the


performance of rolling stock’s maintenance will have great influence on passenger‘s safety and
comfort on board. Consequently, various combination strategies between preventive maintenance
and corrective maintenance will affect railway system safety, passenger comfort and total operation
cost. Railway system operator is therefore required to have a complete overall thought to build
rolling stock maintenance strategy to achieve optimal system operation performance. How to build
a complete and sustainable maintenance strategy will have immense influence on system
operators “railway companies”, system safety supervisor “government” and system users
“passengers”.

Therefore, this study first examines rolling stock maintenance strategy through multiple criteria
decision-making by expert choice. The stock of spare parts is strongly dependent on maintenance
strategy. In the second step, spare parts quantities and replacement intervals estimation were
conducted based on the maintenance strategy chosen by experts in the first step.

The selection of suitable maintenance strategy is very complicated, because the operator needs to
consider the non-metric variables (safety, passenger comfort) and metric variables (maintenance
cost, inventory cost, shortage cost) simultaneously to decide final strategy. Therefore, this study
adopts ANP method that could consider jointly non-metric and metric variables all together.

This study will provide an expert decision method to consider various strategic combinations of
rolling stock preventive maintenance and corrective maintenance and then decide the spares parts
and replacement interval. This fruit of this study could serve as a reference for railway system
operator in adjusting maintenance strategies.

Literature review
There were plentiful studies on manufacturing system maintenance and replacement problems
(McCall (1963), Barlow, and Proshan (1965, 1975), Pierskalla and Voelker(1976), Osaki and
Nakagawa (1976), Sherif and Smith (1981), Jardine and Buzacott (1985), Valdez-Flores and
Feldman (1989), Cho and Parlar (1991), Jensen (1995), Dekker (1996), Pham and Wang (1996),
Van Der Duyn Schouten (1996),and Dekker et al. (1997)). Thousands of maintenance and
replacement models have been created. Most previous researches on maintenance model are
formulated by total cost consideration (Ruhul Sarker, amanul Haque, 2000, Won Young Yun, Luis
Ferreira, 2003)

In comparison with manufacturing system maintenance, the studies on rolling stock maintenance
were relative rare. Dipark Chaudhuri and P.V. Suresh (1995) developed an algorithm for
determining the best type of maintenance, period length and replacement policy using fuzzy set
theory. But this study did not consider the safety for model formulation which is the crucial factor in
the rolling stock maintenance.

In spare parts consideration, Chelbi and Ait-Kadi (2001) proposed a jointly optimal periodic
replacement and spare parts provisioning strategy, the performance of this strategy was evaluated
in terms of total average cost per time unit over an infinite horizon. Yun and Ferreira (2003)
described the development of a simulation model to assess the inventory requirements of
alternative rail sleeper replacement strategies. The main aim of the model is to determine the
optimal replacement strategy, given replacement costs and resultant train operating cost benefits.

The replacement cost consists of the fixed cost and variable cost proportional to the number of units
replaced. A finite horizon is considered and total expected cost is a criterion for comparing the
proposed policies. But they all have no accurate mentioning about the spare parts that the quantity
must prepare in the unit time, but about this part in Almeida (2001) in the research has considered
this point. Almeida (2001) presented multi-criteria decision models for two maintenance problems:
repair contract selection and spares provisioning. In the repair contract problem the model
incorporates consequences modeled through a multi-attribute utility function. The consequences
consist of contract cost and system performance, represented by the system interruption time. Two
criteria (risk and cost) are combined through a multi-attribute utility function in the spares
provisioning decision model.

Chaudhri and Suresh (1995), Cassady et al. (1998), and Nakagawa (1989) conducted their
researches on principles of maintenance cost minimization to seek best replacement cycle. They
considered maintenance costs including preventive maintenance costs and facility damage
maintenance costs.Huang et al. (1995) and Sarker and Haque (2000) suggested system damage
rates will increase with amortization. When damage maintenance costs are superior of preventive
maintenance costs, a suitable preventive maintenance period will minimize total maintenance costs.
This study follows related literatures’ assumption to apply Weibull distribution on rolling stock
component amortization to obtain a more realistic result.

According to the literatures reviewed, most researches regarding maintenance s trategy and
replacement policy concentrated on model development which formulated by total cost
consideration. The rolling stock maintenance model formulation should avoid merely consider the
cost-oriented model. Thus, in this present study we incorporate the safety as the crucial factor and
adopt expert decision model (Analytic Network Process, ANP) method to select the appropriate
maintenance strategy of rolling stock.

Research Design and Methodology


This study applies expert decision methodology to obtain the appropriate maintenance strategy
followed by related spare parts inventory estimation to reach rolling stock’s component
replacement interval. Expert questionnaires are conducted in two phases. Phase 1 interviews
maintenance staffs on site to establish questionnaire framework for Phase 2. Phase 2 interviews
maintenance managers and applies ANP method to obtain the appropriate rolling stock
maintenance strategy and to decide he weight for the evaluation factors. In addition, multiple utility
function assumptions are applied to obtain spare parts estimation values to comprehend preventive
and corrective maintenance ratio. Finally, the Weibull distribution is applied to assume component
life cycle to obtain optimal replacement interval and cost differences between preventive and
corrective maintenances. Figure 1.1 presents analysis procedure.

Best maintenance strategy obtained


through ANP technique by expert decision

The ratio between preventive maintenance


and corrective maintenance is decided

Spare part quantity and replacement interval


estimation through Weibull distribution

Figure 1 Research analysis procedure

Chang (2002) applied AHP methodology to conduct investigation analysis. A drawback of AHP is
the assumption of independent condition that is in contrast of actual situation. Therefore, many
latest researches apply ANP methodology. The ANP methodology is applied in areas including
priority ranking, substitution production, most suited selection, decision demand, resource
allocation, maximization, performance evaluation, forecasts and risk evaluation. Lee and Kim
(2000) suggested an improved Information system (IS) project selection methodology which reflect
interdependencies among evaluation criteria and candidate projects using analytic network
process (ANP) within a zero-one goal programming (ZOGP) model. But the ANP not only uses in
the appraisal aspect, also may apply in the management domain. Wolfslehner et al. (2005)
compared two different multi-criteria analysis approaches: the analytic hierarchy process (AHP)
with a hierarchical structure and the analytic network process (ANP) with a network structure.
Comparisons are made for evaluating sustainable management strategies at forest
management-unit level by using a C&I approach based on the Pan-European guidelines for
sustainable forest management (SFM). But this research also because will consider to in the
maintenance criterion, some many criteria are to be dependent, therefore will use the ANP to take
this research the appraisal method.

3.1 The analytic network process


Hierarchical models (for example, Analytic Hierarchy Process, AHP), premising independent
elements, face certain limitations when the complexity of decision problems increases and
interactions among criteria and sub-criteria are not implicitly covered. Different approaches have
been proposed to consider interaction and dependence among elements. ( Bernhard Wolfslehner,
Harald Vacik, Manfred J. Lexer, 2005)

ANP model building requires the definition of elements and their assignment to clusters, and a
definition of their relationships (i.e., the connections between them indicating the flow of influence
between the elements). ANP is founded on ratio scale measurement and pair-wise comparisons of
elements to derive priorities of selected alternatives. In addition, relations among criteria and
sub-criteria are included in evaluations, allowing dependencies both within a cluster (inner
dependence) and between clusters (outer dependence) (Saaty, 2001). Pair-wise comparison has
two goals, one for weighting the clusters (i.e., criteria) and the other for estimating the direction and
importance of influences between elements, numerically pictured as ratio scales in a so-called
supermatrix.

Mathematically, an ANP model is implemented following a three-step supermatrix calculation


(Saaty, 2001). In the first step, the unweighted supermatrix is created directly from all local priorities
derived from pairwise comparisons among elements influencing each other. The elements within
each cluster are compared with respect to influencing elements outside the cluster.

This also yields an eigenvector of influence of all clusters on each cluster (Saaty, 1999). In the
second step, the weighted supermatrix is calculated by multiplying the values of the unweighted
supermatrix with their affiliated cluster weights. By normalizing the weighted supermatrix, it is made
column stochastic. In the third and final step, the limit super matrix is processed by raising the entire
super matrix to powers until convergence in terms of a limes.( Bernhard Wolfslehner, Harald Vacik,
Manfred J. Lexer, 2005)

3.2 Spare parts maintenance model


This study follows Almeida (2001) spare parts maintenance model with few adjustments. A
multi-criteria decision model U(c,α) allows the quantification of spare provisioning for a single item
taking into account the total spare cost (C) and the risk of item non-supply ( α ). That is
Max qU( α ,C) (1)
Item reliability is assumed to follow an exponential probability distribution with a mean failure rate
λ . The system maintainability is also assumed to follow an exponential probability distribution with
a known mean time to repair (MTTR).

Total spare cost C depends on the unit cost C1 and the number of spare q
C = qC1 (2)
The probability of the provisioning shortage α , when the number of failures x > q is assumed to
follow a Poisson distribution, given the assumption of exponential distribution for reliability. Thus
α = Pr{ x > q } = 1 – Pr{ x ≤ q }
− λMTTR
q
e (λMTTR) j
= 1− ∑ (3)
j=0 j!
q
(λMTTR ) j
where ψ = ∑
j =o j!
As before, first one-dimensional utility functions are obtained for U(α) and U(C), and then a
multiattribute utility function U( α ,C) is obtained. This multiattribute utility function is assumed to be
additive
U( α ,C) = K1U( α ) + K2U(C) (4)
Substituting (2) and (3) into (4)

− λMTTRψ
U (α , C ) = K1U [1 − e ] + K 2U ( qC1 ) (5)

If (5) is applied to (1), the optimum solution can de obtained.

3.3 Replacement interval estimation


3.3.1 Weibull distribution
This study relies on Weibull distribution for parts amortization assumption through parameter
changes of Weibull distribution. Developed frequency intensity function is characterized in multiple
variations and suitable for upward or downward product failure description. It is widely applied in
reliability life analysis and consists of 2-parameter Weibull distribution with a frequency intensity
function expressed as:

F ( x ) = 1 − exp[−( x θ ) β ] , where x ≥ 0 θ >0 β >0

Frequency intensity function is:


β x β −1
exp[-( x ) ] , where x ≥ 0
β
f(x)= ( ) θ ≥0 β ≥0
θ θ θ
A3-parameter distribution cumulative intensity function is:
F ( x ) = 1 − exp[ −( x − γ ) β ] , where x ≥ γ θ >0 β >0 γ ≥0
θ

In addition, x represents random variable of Weilbull distribution. On reliability life analysis, units are
measured in time; θ represents scale parameter β represents shape parameter. The different
between 3-parameter and 2-parameter Weibull distribution is location parameter represented by γ .
Weilbull distribution frequency intensity function is mainly affected by scale parameter and shape
parameter.

3.3.2 Weibull parameter estimation


Montanari (1997) conducted his study through the application of graphical analysis to estimate
shape parameter and scale parameter. Weibull Probability Paper is a relatively easy and frequently
applied graphical analysis. Its advantages are simple and rapid. Weibull Probability Paper’s
principle is to conduct linear transformation of failure function of Weibull distribution into a 1st
degree linear function with life data indicated. This linear function’s slope is shape parameter
estimated value. The distance of linear parameter could estimate value of scale parameter. If spare
parts lack utilization time or pre-heating time, location parameter is set at 0. Linear transformation
of Weibull failure function is as follows:

x
−( ) β
1 − F ( x) = e θ

We set natural logarithm in this equation and obtained the following


x x
ln(1 − F ( x )) = −( ) β , − ln(1 − F ( x)) = ( ) β
θ θ
We set natural logarithm again:
x
ln[ − ln(1 − F ( x ))] = β ⋅ ln( ) = β ⋅ ln x − β ⋅ ln θ
θ

Make Y = ln[ − ln(1 − F ( x))]


X = ln x
Rewrite above function into a straight function, we have:
Y = β ⋅ X − β ⋅ ln θ

Fothergill (1990) suggested the application of cumulative frequency graphic analysis when
Monte-Carlo simulation method evaluation possesses 2-parameter Weibull distribution. We take Pi
as the estimation value for F(x)
i − 0.3
Pi =
n + 0. 4
where, Pi represents the ith observed cumulative frequency value of n numbers of samples.
Therefore, this study applies F(x) for value estimation followed by frequency graphic analysis
expressed as Y = ln[-ln(1-F(x))], X = ln x t o obtain shape and scale parameters of Weibull
distribution. Spare parts replacement interval forecasting is conducted by following Huang et al.
(1995).

Empirical result analysis

4.1 Appropriate maintenance strategy selection


This study applies expert questionnaire survey to select an appropriate rolling maintenance
strategy. Accordingly, this study needs to find most suited evaluation factors affecting maintenance
strategy selection. Theses evaluation factors selected will be an evaluation framework of ANP
questionnaire. There are three different maintenance strategy were provided according to various
combination between preventive maintenance and corrective maintenance.

Previous literature review to comprehend


rolling stock maintenance strategy
evaluation factors and design 1st
questionnaire

Executive factor analysis through 1st


phase questionnaire and develop 2nd
phase ANP questionnaire

Apply ANP technique to analyze 2nd


phase questionnaire results to select an
appropriate maintenance strategy and
to find weight for the evaluation factors
and sub-factors

Figure 4.1 Questionnaire investigation process

4.2 Questionnaire design and investigation


This study adopts a two-phases questionnaire design. First phase questionnaire is designed by
following Shu (1999) who has considered some specific factors affecting maintenance strategy
selection. Interviews are conducted on site with the maintenance staffs working for various railway
operators including conventional railway system, mass rapid transit system, and high-speed rail
system. This study applies factor analysis technique to conduct data reduction and summarization
and extract important evaluation factors to select the most suitable rolling stock maintenance
strategy.

4.2.1 Factor analysis result


This study adopts the KMO test and Bartlett test to examine the appropriateness of 13 sub-factors
on rolling stock maintenance for factor analysis. Table 4.1 shows a KMO value greater than 0.7 and
a Bartlett test value at 635.797. This indicates the appropriateness of factor analysis application in
rolling stock maintenance related issues. Rolling stock maintenance factor analysis result evidence
total explainable variance at 71.143% and extract three factors.

Table 4.1 KMO and Bartlett’s Test

Kaiser-Meyer-Olkin Measure of Sampling Adequacy. .799


Bartlett’s Test of Approx. Chi-Square 635.797
Sphericity Df 78
Sig. .000

As for the reliability analysis result, this study conducts a Cronbach’s α value to examine rolling
stock maintenance questionnaire reliability to assure all measured factors are highly consistency.
Table shows all Cronbach’s α individual are above 0.7. Table 4.2 presents factor analysis result.

It is suggested to reduce the sub-factors considered in applying ANP technique. Therefore, this
study groups highly correlated sub-factors into one factor through correlation analysis. Empirical
Results show correlation between rolling stock shut-down time and maintenance time reduction is
at 0.626. Therefore, we group these two sub-factors into one factor and rename as maintenance
cost and shut-down time reduction. Correlation between worker and staff safety and passenger
safety is at 0.643. Therefore, we group these two sub-factors into one factor and rename as worker
and passenger safety assurance. Correlation amongst maintenance cost reduction, staff efficiency
improvement, and staff work assignment are at 0.643 and 0.673 respectively. Correlation between
staff efficiency and work assignment improvement reach as high as 0.780. Therefore, we group
these three sub-factors into one factor and rename as staff work efficiency increases. Correlation
between sudden incident occurrence reduction and railway failure rate improvement is at 0.704.
Therefore, we group these two sub-factors into one factor and rename as rolling stock failure rate
reduction.
Table 4.2 Empirical result of factor analysis to extract evaluation factors affecting selection the

suitable maintenance strategy


Factor Sub-factor Factor Eigenvalue % of Cumulative Cronbach’s α
loading Variance variance
%
Maintain high 0.615
quality maintenance
Quality Maintain 0.735
and appropriate
Efficiency available spare 6.357 48.898 48.898 0.889
parts
Staff work efficiency 0.643
improvement
Staff work 0.668
assignment and
preparation
improvement
Reduce the impact 0.696
in case of
emergency
Railway car 0.814
out-of-service rate
improvement
Maintain rolling 0.436
Cost and stock in good
Reliability condition
Reduce rolling stock 0.812 1.814 13.957 62.855 0.820
shut-down time
Reduce 0.894
maintenance time
Reduce 0.721
maintenance cost
Assure staff and 0.874
Safety personnel safety
Assure railway 0.832 1.077 8.288 71.143 0.810
system safety
Assure passenger 0.693
safety

4.2.2 ANP questionnaire design and empirical result

The 2nd phase questionnaire design is based on factor analysis result of 1 st phase questionnaire.
The 2nd phase questionnaire is constructed on three main evaluation factors with eight sub-factors
and three selection alternatives. This study refers to Taipei Mass Rapid Transit System on
preventive maintenance and corrective maintenance percentages for selection possible
alternatives.
The following presents eight reorganized sub-factors categorized into three main factors.
(1) Factor 1: Cost and quality
This factor includes four sub-factors such as high quality maintenance, staff work efficiency
increase, appropriate usable spare parts maintenance and rolling stock failure rate reduction
(2) Factor 2: rolling stock availability
This factor includes two sub-factors such as facility maintenance in good condition and
maintenance cost and shut down time reduction
(3) Factor 3: safety
This factor includes two sub-factors such as worker and passenger safety assurance and
railway car safety assurance.
We took the model 321 metro rolling stock current collecting shoe as a component for analysis.
Because it is crucial for MRT’s daily operation. In addition, three alternatives are proposed:
Alternative A : the ratio between PM and CM is 7:3
Alternative B : the ratio between PM and CM is 1:1
Alternative C : the ratio between PM and CM is 3:7

Table 4.3 Evaluation factors and sub factors as the basis for ANP method questionnaire

Factor Sub-factor
High quality maintenance

Quality and Increase staff work efficiency


Efficiency Maintain appropriate usable
spare parts
Reduce rolling stock failure rate
Maintain rolling stock in good
Cost and Reliability
condition
Reduce maintenance cost and
shut-down time
Assure worker and passenger
Safety
safety
Assure rolling stock safety
Table 4.4 maintenance strategy expectation index

Evaluation factors maintenance strategy weights STi C kj

Cj Weight(Rj) WT1 WT2 WT3 ST1C j ST 2 C j ST3 C j

C1 0.176 0.619 0.244 0.137 0.109 0.043 0.024


C2 0.092 0.521 0.281 0.198 0.048 0.026 0.018
C3 0.294 0.615 0.239 0.146 0.181 0.070 0.043
C4 0.234 0.641 0.230 0.129 0.150 0.054 0.030
C5 0.052 0.615 0.227 0.158 0.032 0.012 0.008
C6 0.048 0.508 0.293 0.199 0.024 0.014 0.010
C7 0.042 0.576 0.268 0.156 0.024 0.011 0.007
C8 0.062 0.629 0.219 0.152 0.039 0.014 0.009

Maintenance strategy expectation index DIi 0.607 0.244 0.149

For evaluation results, Figure 4.2 shows Cost and Reliability, Safety and Quality and Efficiency with
weight values of 0.268, 0.528, and 0.204 respectively. This result fully explains safety as first
priority in railway rolling stock maintenance considerations. Figure 4.2 further shows maintenance
strategy A (0.607) is significantly superior of other two maintenance strategies after calculation of
maintenance strategy expectation index in table 4.2. Therefore, best maintenance strategy A is
preventive maintenance and corrective maintenance with percentages at 7:3 followed by
preventive and corrective maintenance strategy weight percentages at 1:1 (alternative B, weight
value 0.244). Finally, preventive and corrective maintenance strategy weight percentage at
3:7(alternative C, weight value 0.149). This result explains a majority of maintenance work is
focused on preventive maintenance for rolling stock maintenance. It is dangerous to cease railway
system operation for the reason of component failure of rolling stock. It seems that a
preventive-oriented maintenance strategy could probably assure rolling stock safety. In addition,
this study applies questionnaire analysis result of ANP for estimating spare parts of component and
replacement interval.
Maintain rolling stock in good condition (0.176)
Cost and Reliability
(0.268)
Reduced maintenance cost and shut-down time
PM CM=7 3
( 0.092)
( 0.607)
Rolling stock maintenance strategy

Assure workers and the passenger’s safety (0.294)

Safety
(0.528) Assure rolling stock safety (0.234)
PM CM=1 1
(0.244)

High quality maintenance (0.052)

Quality and Efficiency


Increase staff work efficiency (0.048)
(0.204) PM CM=3 7
(0.149)
Maintain appropriate usable spare parts (0.042)

Reduce rolling stock failure rate (0.062)

Figure 4.2 ANP analysis framework and weight value of factors and sub-factors
4.3 Spare Parts Estimation
Through ANP methodology, railway rolling stock’s preventive maintenance and corrective
maintenance ratio is obtained 7 3 and applied to this ratio to estimate needed spare parts quantities
for of Taipei MRT’s rolling stock model 321 current collecting shoe. The multiple utility function is
applied to estimate maximum efficiency of spare parts.

Supply shortage possibility α


Assumed reliability follows a Poisson distribution. Random 50 current collecting shoes data
provided by Taipei Mass Rapid Transit obtains MTTR at 12.78 months. Therefore, supply shortages
possibility is:
α = P r{ x > q } = 1 – Pr{ x ≤ q }
q
e − λMTTR (λMTTR) j
= 1− ∑
j=0 j!
e −λ λx
where λ is unknown, and f ( x) = in a Poisson distribution. Therefore, this study applies
x!
λ as estimation value for λMTTR . It is set MTTR at 12.78 months, we obtain λ at 0.939
failures/year. With known λ value, we can obtain supply shortage possibility of individual spare
parts quantities. When q = 5, obtained α value is 0.00062.

4.3.1 Multiple utility function


Multiple utility function is applied to obtain maximum utility value after the calculation of α value.
The multiple utility function is expressed as follows:
U( α ,C) = K1U( α ) + K2U©

− λMTTRψ
= K 1U [1 − e ] + K 2U ( qC1 )

where U (α ) = exp(− A1α )


U (C ) = exp(− A2 C )
In current collecting shoe cost estimation, due to various purchase volumes of Taipei Mass Rapid
Transit, current collecting shoe costs vary in the range between NT$2000 and NT$4000. This study
assumes current collecting shoe cost as U.S$10.

A1 and A 2 are variables. This study assumes A1 =16 and A 2 =0.002 to obtain maximum utility value.
Because the most suitable strategy derived from ANP result in the first step is to take the ratio
between PM and CM = 7 : 3. We take K1 = 0.7 and K2 = 0.3 to obtain effective value of individual
spare parts. When q=5, the utility value is 0.964
4.3.2 Optimal spare parts estimation

According to the result derived form the multiple utility function, we obtained that when q=5, the
utility value is maximum. Therefore, 5 spare parts need to be prepared. This study follows Almeida
(2001) and assumes a one-to-one facility and spare part composition. Therefore, this study
assumes model 321 of Taipei Mass Rapid Transit adopts the one-to-one car and current collecting
shoe with one rolling stock containing 6 cars. That is, each rolling stock must be equipped with 30
spare parts, 36 rolling stock require 1080 spare parts per year, and 90 spare parts per month on
average.

4.4 Optimal replacement interval


This study examines optimal replacement interval as reference for Taipei Mass Rapid Transit
(MRT) conducting maintenances. Statistical data are obtained from 50 random maintenance record
of past maintenance history of Taipei Mass Rapid Transit. Results found average replacement time
at 12.779 months. This study follows Huang et al. (1995) in assuming preventive and corrective
maintenance’ cost ratios at 1:15. The cost of corrective maintenance is 15 times more than
preventive maintenance.

4.4.1 Parameter estimation


Random replacement data shows Y = ln[ − ln(1 − F ( x))] and X = ln x and obtains 1 st Weibull
distribution failure function linear equation expressed as Y = 7.50X-19.6. Through this linear
equation, obtained Weibull distribution shape parameter β and scale parameter θ are 7.50 and
13.644 respectively. Huang et al. (1995) apply values of these two parameters to obtain optimal
replacement interval.

4.4.2 Replacement interval calculation


This study follows Huang’s et al. (1995) mathematical model T0 * = 0.1 ×T s * × θ to obtain MRT
321 type rolling stock’s collecting current shoe’s optimal replacement interval with T0 * as optimal
replacement time. The θ value is estimated from 50 random replacement data. Replacement time
“ Ts * ” is obtained by following Huang et al. (1995) through the β value and ratio between
corrective and preventive maintenance costs. This study adopts β value of 7.50 and ratio
between corrective and preventive costs at 15 1. Through estimation graph proposed by Huang et
al. (1995), estimated replacement time Ts * is between 5.0 and 6.0. Therefore, optimal
replacement interval is between 6.822 and 8.1864 months.
Conclusion and Suggestion

The purpose of this paper was to present a method for rolling stock’s maintenance strategy
selection that allows for the consideration of important interactions among decisions levels and
criteria. The methodology adopts ANP for this evaluation. Consequently, we use the empirical
result derived from ANP to decide the possible ratio between preventive maintenance and
corrective maintenance that can induce the possible spares parts quantities and replacement
interval of component of rolling stock.

The empirical result based on ANP method on maintenance strategy of rolling stock indicates
preventive maintenance should be much more emphasized than corrective maintenance. This
result is consistent with the studies of maintenance strategy on industrial equipments
(Nakagawa,1989, Huang et al. 1995, Chelbi and Ait-Kadi,2001)

According to the empirical result derived from ANP method, safety is the most crucial factor for
rolling stock maintenance strategy selection. Safety here considers not only passenger safety but
also maintenance mechanic agent safety. The second important factor is to keep high availability of
rolling stock for operation to avoid trains idling in the maintenance site. Maintenance cost and
quality is the third factor affecting rolling stock maintenance strategy choice by the experts. This
result exists essential difference between industrial facility and equipment maintenance and rolling
stock maintenance.

This study chooses Taipei MRT’s rolling stock component: model 321 current collecting shoes as
an analytical component to estimate spare parts quantities and optimal replacement interval. The
empirical result indicates each rolling stock must be equipped with 30 spare parts. The optimal
replacement interval is between 6.822 and 8.1864 months. The developed method presented in
this study could be useful for railway operator to select its appropriate rolling stock maintenance
strategy and to decide component’s spare parts quantities and optimal replacement interval.

References

[1] Almeida, A. T. “Multi criteria decision making on maintenance Spares and contracts planning”,
European Journal of Operational Research, pp. 235-241, (2001).
[2] Cassady C. R., L. M. Maillart, R. O. Bowden, B. K. Smith. “Characterization of Optimal
Age-Replacement Policies”, Proceedings Annual Reliability and Maintainability Symposium, pp.
170-175, (1998).
[3] Chang Huan-Kuang “Analytic Time Delay Study of Train Failure Treatment Mode for MRT
System - A Case Study of Metro Taipei”, master thesis, NTUST, Taiwan,( in Mandarin) (2002)
[4] Chaudhuri, D., and Suresh, P. V. “An Algorithm for Maintenance and Replacement Policy
Using Fuzzy Set Theory”, Reliability Engineering and System Safety 50, pp. 79-86, (1995).
[5] Chelbi, A. and D. Ait-Kadi. “Spare provisioning strategy for preventively replaced systems
subjected to random failure”, International Journal of Production Economics, Vol. 74, pp. 183-189,
(2001).
[6] Fothergill, J. C. “Estimating the cumulative probability of failure data points to be plotted on
Weibull and other probability paper”, IEEE Transactions on Electrical Insulation, Vol. 25, No. 3, pp.
489-492. (1990).
[7] J. Huang, C. R. Miller, O. G. Okogbaa. “Optimal Preventive-Replacement Intervals for the
Weibull Life Distribution: Solutions & Applications”, Proceedings Annual Reliability and
Maintainability Symposium, pp. 370-377, (1995).
[8] Lee, J. W., and Soung, H. K., (2000). “Using analytic network process and goal programming
for interdependent information system project selection”, Computers & Operations Research, Vol.
27, pp. 367-382, (2000).
[9] Lie, C. H., and Chun, Y. H. “An Algorithm for Preventive Maintenance Policy”, IEEE
Transaction on Reliability, Vol.R-35, No.1., (1986).
[10] Montanari, G. C., G. Mazzanti, M. Cacciari, and J. C. Fonthergill. “In Search of Convenient
Techniques for Reducing Bias in the Estimation of Weibull Parameters for Uncensored Tests”,
IEEE Transactions on Dielectrics and Electrical Insulation, Vol. 4, No. 3, pp. 306-313, (1997).
[11] Rushul Sarker, Amanul Haque, (2000). “Optimization of maintenance and spare provisioning
policy using simulation”, Applied Mathematical Modeling, pp. 751-760, (2000).
[12] Shu. ”The decision consideration and performance evaluation of maintenance management”,
master thesis, institute of industrial engineering, NTHU, Taiwan ( in Mandarin), (1999).
[13] Su, K. W., S. L. Hwang and T. T. Liu. “Knowledge Architecture and Framework Design for
Preventing Human Error in Maintenance Tasks”, Expert System with Application, 19, pp. 219-228,
(2000).
[14] T. Nakagawa. “A Replacement Policy Maximizing MTTF of a System with Several Spare
Units”, IEEE Transactions Reliability, Vol. 38, No.2, pp. 210-211, (1989).
[15] Wolfslehner, B., Harald, V., Manfred, J. L. “Application of the analytic network process in
multi-criteria analysis of sustainable forest management”, Forest Ecology and Management, Vol.
207, pp. 157-170, (2005).
[16] Yun, W. Y. and Luis, F. “Prediction of the demand of the railway sleepers: A simulation model
for replacement strategies”, International Journal of Production Economics, Vol. 81-82, pp. 589-595,
(2003).

You might also like