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

net/publication/262882582

An Optimal Reliability and Maintainability Design of a Searching

System

Article  in  Communication in Statistics- Simulation and Computation · April 2014

DOI: 10.1080/03610918.2013.815771

CITATIONS READS

4 2,081

3 authors, including:

Won Young Yun Goeun Park

Pusan National University Korea Environmental Industry and Technology Institute
169 PUBLICATIONS   1,933 CITATIONS    13 PUBLICATIONS   65 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Won Young Yun on 10 December 2014.

The user has requested enhancement of the downloaded file.

This article was downloaded by: [Pusan National University Library ]
On: 09 December 2014, At: 23:11
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Communications in Statistics - Simulation

and Computation
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/lssp20

An Optimal Reliability and

Maintainability Design of a Searching
System
a a a
Won Young Yun , Goeun Park & Young Jin Han
a
Department of Industrial Engineering , Pusan National University ,
Busan , Korea
Accepted author version posted online: 07 Dec 2013.Published
online: 14 Apr 2014.

To cite this article: Won Young Yun , Goeun Park & Young Jin Han (2014) An Optimal Reliability
and Maintainability Design of a Searching System, Communications in Statistics - Simulation and
Computation, 43:8, 1959-1978, DOI: 10.1080/03610918.2013.815771

To link to this article: http://dx.doi.org/10.1080/03610918.2013.815771

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
 Communications in Statistics—Simulation and ComputationR , 43: 1959–1978, 2014
Copyright © Taylor & Francis Group, LLC
ISSN: 0361-0918 print / 1532-4141 online
DOI: 10.1080/03610918.2013.815771

An Optimal Reliability and Maintainability Design

of a Searching System

WON YOUNG YUN, GOEUN PARK, AND YOUNG JIN HAN

Department of Industrial Engineering, Pusan National University, Busan, Korea
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

In this article, we deal with an optimal reliability and maintainability design problem of a
searching system with complex structures. The system availability and life cycle cost are
used as optimization criteria and estimated by simulation. We want to determine MTBF
(Mean Time between Failures) and MTTR (Mean Time to Repair) for all components
and ALDT (Administrative and Logistics Delay Times) of the searching system in order
to minimize the life cycle cost and to satisfy the target system availability. A hybrid
genetic algorithm with a heuristic method is proposed to find near-optimal solutions
and compared with a general genetic algorithm.

Keywords Reliability; Maintainability; Availability; Simulation; Genetic algorithm;

Searching system.

Mathematics Subject Classification Primary 90B25; Secondary 68020.

1. Introduction
Recently, weapon systems such as aircraft, battle tanks, and ships have become more
complex and their technologies more advanced because it requires performing a lot of
missions during their life cycle. As a result, the operational and maintenance costs of
weapon systems have been increasing gradually. For performing missions successfully,
highly reliable weapon systems need to be developed as well as effectively and efficiently
maintained periodically. Also, the system performances need to be evaluated accurately and
system availability and life cycle cost are often used as key measures because two measures
are influenced by system operation characteristics, maintenance, and logistic activities.
Additionally, it is important to design an optimal reliability and maintainability for weapon
systems from the beginning phase of the system design and development. In realistic
models, it is difficult to obtain reliability and maintainability solutions of components in
weapon systems by analytical method because the systems consist of many components
and system reliability structure may be complicated. Also, the time to failure of each
component may follow different failure distribution. In this situation, simulation can be
a good alternative because it can describe real system behavior based on the operational
scenario and provides various statistics about the system behavior or critical components
from a reliability/availability point of view.

Received December 31, 2012; Accepted June 7, 2013

Address correspondence to Won Young Yun, Department of Industrial Engineering, Pusan
National University, San 30, Jangjeon-dong, Geumjeong-gu, Busan, South Korea; E-mail:
[email protected]

1959
 1960 Yun et al.

Landers et al. (1991) analyzed the mission reliability of complex systems. In order to
estimate the mission reliability and the performance capabilities of the voice communica-
tion system in the F17 fighter aircraft, a discrete-event simulation model was proposed.
Yanez et al. (1997) proposed a functional diagram model to describe complex system struc-
tures, and a simulation tool was developed to evaluate reliability and availability of a mine
hunter system. Yun et al. (2008) proposed an object-oriented simulation model to estimate
reliability, availability, and maintainability of multi-function systems with complex struc-
tures. Chung (2009) developed a simulation system for evaluating operational policies of
multi-indenture multi-echelon (MIME) systems. The proposed simulation model was de-
veloped using the object-oriented method and explained through various diagrams such as
class diagram and state transition diagram. For complex large-scale systems such as rolling
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

stock systems, power plants, a simulation method is used to assess their performances and
Park et al. (2011) dealt with a preventive maintenance (PM) scheduling problem for a
rolling stock system in the railway system, especially KTX (Korea Train eXpress) and a
commercial RAM simulation S/W, AvSim was used to describe the operational scenario
of KTX and estimate the system availability and life cycle cost. Borogonove et al. (2000)
studied maintenance strategies for power plants and used Monte Carlo simulation to model
the system operation with imperfect repair for evaluating the system maintenance strategies.
However, simulation methods give only various statistics of system performance mea-
sures and it needs appropriate optimization methods to find optimal solutions for the
optimization problem. The reliability and maintainability design problem is one of impor-
tant optimization and engineering issues for which a lot of studies exist (refer Kuo et al.,
2001). The reliability and maintainability of components in the system are often optimized
by using meta-heuristic algorithms when simulation methods are used and Gen and Kim
(1999) reviewed genetic algorithms which were applied to solve various reliability design
problems such as the reliability design problem of redundant systems. Hsieh et al. (1998)
considered a reliability design problem in series, series-parallel, and bridge systems and
a genetic algorithm was proposed to simultaneously determine the reliability of compo-
nents and the redundancy allocation for maximizing system reliability. Gupta et al. (2009)
studied a redundancy allocation problem in a series system with interval values of relia-
bility of components and proposed a genetic algorithm to find the near-optimal solution
for maximizing the system reliability. Moghaddam et al. (2008) dealt with a redundancy
allocation problem in series-parallel systems. A genetic algorithm was proposed to find the
best redundancy strategy for subsystems maximizing the system reliability.
In this article, we deal with a reliability and maintainability design problem for a
searching system and want to determine MTBF and MTTR of all components and ALDT
of the searching system optimally. Firstly, we analyze the searching system such as the
physical structure and reliability structure between components. The operational scenario
of the searching system is also analyzed to build the simulation model for estimating
the system availability and the life cycle cost. Next, a hybrid genetic algorithm with a
heuristic method is proposed to find near-optimal design alternatives for reliability and
maintainability of components in the system and numerical results are studied to compare
the proposed hybrid genetic algorithm with the general genetic algorithm. This article is
organized as follows: In Section 2, we introduce a searching system as a research target
system and show the reliability structures in the searching system. A simulation model is
proposed to estimate the system performance measures in Section 3 and a simulation-based
optimization procedure with a hybrid genetic algorithm is explained in Section 4. Next,
numerical examples are studied to investigate the effect of model parameters to optimal
 Optimal Ram Design of a Searching System 1961

solutions and compare the hybrid genetic algorithm with the general genetic algorithm in
Section 5. Finally, Section 6 concludes the article.

2. Reliability Structure of a Searching System

A searching system is developed for sound propagation to navigate, communicate, and de-
tecting objects, such as other vessels on or under the water and consists of many components
with complex structure. Among components in the searching system, sensors are one of the
most important components to detect marine objects and sensors in the searching system
are classified into passive sensors and active sensors. The passive sensors are essentially
used to listen for the sound made by vessels and always work to detect noise from marine
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

objects because it does not emit its own pulse. While the active sensors usually do not work
because their own pulse of sound to detect the marine objects is omitted and it results that
they can be exposed to the other searching systems. Thus, active sensors only work when
necessary but otherwise is kept in standby during the mission period.
A searching system performs a lot of missions and one or more functions are required to
carry out a mission. Many different components are also required to support these functions
simultaneously and they can have various reliability structures. In order to keep high system
availability, there are some components with more complex reliability structures such as
standby and k-out-of-n structures in the searching system (Rausand and Høyland, 2004).
For example, electronic cabinets are also one of the most important components to carry out
search functions successfully because they support all sensors in the searching system. Thus,
electronic cabinets need to be replaced by reserved components immediately at failures.
Figure 1 shows an example of a standby structure between sensors and electronic
cabinets in a searching system. Both sensor A and electronic cabinet 1 are needed for
function 2, and sensor G and electronic cabinet 3 are needed for function 4. If an electronic
cabinet among cabinet 1 and 3 fails (the right case of Figure 1), the remaining electronic
cabinet covers two sensors together until the failed electronic cabinet is repaired.
The passive and active sensors in the searching system also consist of several com-
ponents with k-out-of-n structure. Figure 2 shows an example of the k-out-of-n structure
between components in the sensor A and the sensor B fails when at least three components
(k = 3) are failed among n components.
The system availability and life cycle cost are key measures to evaluate the system
performances and often used as optimization criteria. The system availability is influenced
by reliability and maintainability of components in the system, and it is important to

Figure 1. Standby structure between sensors and electronic cabinets (Example) 49 × 17 mm (300 ×
300 DPI).
 1962 Yun et al.

Figure 2. k-out-of-n structure between components of sensor B (Example) 51 × 17 mm (300 × 300

DPI).
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

determine optimally reliability and maintainability of components in the system in order

to satisfy the target system availability. In this article, the life cycle cost includes the
development and repair costs to improve the reliability of components, and the investment
cost to improve the maintainability of components. We want to determine the optimal
MTBF and MTTR of all components and ALDT of the searching system to minimize the
life cycle cost and satisfy the required availability.

3. RAM Simulation for a Searching System

For determining optimal MTBF and MTTR of all components and ALDT of the searching
system that minimize the life cycle cost and satisfy the target system availability, the system
performances of the searching system must be evaluated accurately. Therefore, it is needed
to calculate the system availability and life cycle cost of the system for given alternatives of
reliability and maintainability of components, however, it is not easy to obtain the system
performance measures by analytical methods. Simulation is a very powerful and applicable
tool in such situations. Thus, we use simulation technique to estimate the system availability
and life cycle cost. Assumptions about failure and repair in our simulation model are as
follows:

• Two types of maintenance as corrective maintenance (CM) and PM are considered.

• The failure of a component is divided into two types: repairable and nonrepairable
failures. In a nonrepairable failure case, the component is replaced with a new one.
• Failures of all components are independent and the system can perform only one
mission at a time.
• The reliability structures of components as series, parallel, k-out-of-n, and stand-by
structures are considered.

There are several repair models by the state (age) of the component after repair. The
simulation system uses three repair models known as perfect repair, minimal repair, and
imperfect repair (see Pham and Wang, 1996 for details).

• Perfect repair: The age of the components after repair is as good as new.
• Minimal repair: The age of the components after repair is as bad as old.
• Imperfect repair: There are two kinds of imperfect repair models. The first is where
the state of the system is as good as new with probability p or the same as before with
1 – p after all repairs (Nakagawa, 1979a, 1979b). The second one is an age reduction
 Optimal Ram Design of a Searching System 1963

model where each repair proportionally reduces the age of the components (Brown
et al., 1983).

The developed simulation system is basically based on the discrete-event simulation

model using next-event time advance technique in order to advance the simulation clock
because it is appropriate for those systems whose state changes at discrete points in time.
We also use object-oriented design tools to develop the simulation system. The object-
oriented simulation consists of objects that interact with each other as the simulation
evolves through time. Objects contain data that describe its state at a particular point in
time and have methods describing actions that each object can perform.
We also use the process approach in the simulation system. A process is defined
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

as an object that operates by changing states over time and there are several different
types of process in the simulation system (Garrido, 1998). The process approach has
several elements as the simulation clock, the event list, the timing routine, the initialization
routine, the library routine, and report routine to execute simulation. For an example,
simulation begins at time 0 and the system state and the event list is initialized by the
initialization routine. Next, the simulation main routine invokes the timing routine for
determining the next event type and the event routine for updating the state of objects.
Figure 3 shows the logical relationships among routines in simulation (Law and Kelton,
2000).

Figure 3. Flow of discrete-event simulation system using the next-event time advance approach 84 ×
65 mm (300 × 300 DPI).
 1964 Yun et al.

Based on an object-oriented design, we define objects as missions, functions, and

components etc. and an object has three states; RUN, DOWN, and IDLE states. There are
four common methods for objects: “Activate” to run the object, “DeActivate” to stop the
object, “ObjectDown” to inform the failure of the object, and “ObjectFixed” to inform the
repair completion of the object failed to other objects.

• In the mission state case, the state of the current mission changes from RUN to
IDLE if STOP event occurs because system statistics are only related to the current
mission and the statistics of the mission completed do not need to be collected. Thus,
we assume that the system perform the next mission immediately after passing a
specified period of the current mission even if the state of the current mission is
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

DOWN.
• In the function state case, the states of functions RUN and DOWN depend directly on
the state of the current mission, RUN or DOWN while IDLE depends on whether the
next mission requires the function or not. For an example, the state of a function that
does not need to perform the next mission changes from RUN to IDLE. Otherwise,
its state is RUN.
• In the component state case, the state of the component after repair depends on
whether functions that support the current mission require the component or not.
The state of the available component changes from IDLE to RUN if at least a function
requires the component. Otherwise, its state keeps IDLE.

The simulation system provides a module to build the operational scenario consisting
of missions for the searching system. The operational scenario is a sequence of missions
and its length is the sum of the length of missions included in the scenario. The length of
missions can be constant or random variables.
The procedure to build the operational scenario in the simulation system is as fol-
lows; first, we define components, functions, and missions and give information on re-
liability and maintainability of each component such as failure and repair distributions,
repair model, etc. Second, we define relationships between components, functions, and
missions. For example, we should specify components required for each function and
functions necessary to perform each mission and can consider several reliability struc-
tures as series, parallel, k-out-of-n, and standby structures in these relationships between
three units. Finally, we define the length of each mission and build an operational sce-
nario that is a sequence of several missions. In particular, the operational scenario can be
a predetermined or random sequence of missions. In a predetermined sequence case, the
system performs missions enrolled in the operational scenario sequentially. In a random
sequence case, the next mission can be determined randomly and the transition probabil-
ity matrix (discrete-time Markov chain) can be used to model the random sequence of
missions.

4. Reliability and Maintainability Design of a Searching System

4.1. Reliability and Maintainability Design Procedure

We consider MTBF and MTTR as measures of the reliability and maintainability of compo-
nents in the system, respectively. We optimize one more supportability parameter, ALDT,
such as waiting times for spare parts in the depot, times to prepare maintenance actions,
and so on.
 Optimal Ram Design of a Searching System 1965
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

Figure 4. The optimization procedure for reliability and maintainability of the searching system
49 × 29 mm (300 × 300 DPI).

In order to determine optimal MTBF and MTTR values of all components and ALDT
value of the searching system, the simulation-based optimization procedure with a hybrid
genetic algorithm for an optimal reliability and maintainability design of the searching
system is proposed as following Fig. 4. First, we input reliability and maintainability data
of all components after building the simulation model for the operational scenario of the
searching system, and set the target system availability. Second, alternatives for MTBF and
MTTR values of components and ALDT of the searching system are generated through
a hybrid genetic algorithm with a heuristic method and the system availability and life
cycle cost are estimated by simulation. If the current solution satisfies the target system
availability, the current solution is the global best solution. Otherwise, it goes back to the
second step in order to generate alternative solutions.
Figure 5 describes the flow of a hybrid genetic algorithm which combines a genetic al-
gorithm with a heuristic method after mutation to improve design alternatives. In subsection
4.3, we explain steps in the hybrid genetic algorithm in detail.

4.2. A General Genetic Algorithm

For finding near-optimal design alternatives for each MTBF and MTTR of all components
and ALDT of the searching system, a genetic algorithm is proposed and Table 1 shows
notation for the genetic algorithm.
In the proposed genetic algorithm, chromosomes are represented by integer strings and
the total length of a string is equal to three times the total number of components in the
system. Three genes in a chromosome mean MTBF and MTTR of a component and ALDT
of the searching system, and Fig. 6 shows an example of the solution representation.
The objective function evaluated by simulation is the life cycle cost with the penalty
cost by using Eq. (1). The life cycle cost consists of the development, investment, repair, and
penalty costs. The development and repair costs are related to the MTBF improvement of
components and the investment costs is related to the MTBF improvement of components
and ALDT improvement of the searching system. The penalty cost in Eq. (3) is added
to the chromosome of which the system availability does not satisfy the target system
 1966 Yun et al.
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

Figure 5. A hybrid genetic algorithm 71 × 49 mm (300 × 300 DPI).

availability.
n   
Fi
Min TCj = RCi × + DCi + (ICMTTRi + ICALDT i ) + PEj (1)
i=1
L

Table 1
Notation for a general genetic algorithm

Notation Definition
i Index of component (i = 1, . . . , I)
j Index of alternative (j = 1, . . . , J)
l Index of generation (l = 1, . . . , L)
e Allowable gap
TCj Life cycle cost of alternative j
RCi Repair cost of component i
Fi Number of repair for component i
L System life cycle
DCi Development cost of component i
ICMTTRi Investment cost for MTTR of component i
ICALDTi Investment cost for ALDT of component i
AT Target availability
Aj System availability of alternative j
PEj Penalty cost of alternative j
 Optimal Ram Design of a Searching System 1967
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

Figure 6. Solution representation 53 × 22 mm (300 × 300 DPI).

Subject to AT − ε ≤ Aj ≤ AT + ε (2)
P Ej = (l + 1) × {TCj × (AT − Aj )} (3)

There are three genetic operators in the genetic algorithm that exist as crossover,
mutation, and selection operations in order to keep the balance between exploration and
exploitation in search space. For generating new chromosomes by exchanging genes of two
chromosomes, the two-cut-point crossover is used as shown in Figure 7.
The mutation operation selects one gene of a chromosome with mutation rate and
replaces MTBF and MTTR of the component and ALDT of the searching system depending
on whether the target system availability is satisfied. Figure 8 shows how a new chromosome
is created from a parent chromosome through the mutation operation (pM = 0.3).
We apply a deterministic adaptation for mutation ratio. The deterministic adaptation
takes place if the value of a strategy parameter as mutation ratio is changed by some
deterministic rule (Gen et al., 2008). Equation (4) shows mutation ratio of the current
generation if we assume the mutation ratio will decrease from 0.5 to 0.2 as the number of

Figure 7. Crossover (Example) 99 × 44 mm (300 × 300 DPI).

 1968 Yun et al.
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

Figure 8. Mutation (Example) 99 × 42 mm (300 × 300 DPI).

generations increases to maxGen.

t
pM = 0.5 − 0.3 (4)
max GEN
where t is the current generation number and maxGen is the maximum generation.
There are some opportunities to lose the best chromosomes from previous generations
when crossover and mutation create a new population. Therefore, a set of the best chromo-
somes from the current generation need to be turned over to the next. Accordingly, we use
the elite selection to keep the best chromosome from parents and offspring.

4.3. A Hybrid Genetic Algorithm with a Heuristic Method

After mutation operation, a heuristic method is used to improve alternatives in the general
genetic algorithm. Table 2 shows notation for the hybrid genetic algorithm with a heuristic
method and the heuristic procedure in the hybrid genetic algorithm is as follows:

Step 1: Select an alternative j randomly. Estimate AM STD and AC of the selected

alternative j. Then, decide whether it will be improved or decreased.
Step 1–1: Generate a random number for an alternative j. (j = 0, 1, 2, . . . , J
(J : Total number of alternatives))
Step 1–2: Select an alternative j based on the generated random number at
“Step 1—1”.
Step 1–3: Estimate AM STD of the selected j by simulation based on the fol-
lowing Eq. (5) then compares it with allowable standard deviation
of availability, AM STDA . (k = 1, 2, . . . , K (K : Total number of
missions))

K
k=1 (Ak − AT )
2
AM STD = (5)
K

Step 1–4: If AM STDA ≥ AM STD , set the selected j as optimal then terminate the
heuristic. Otherwise if AM STDA < AM STD , go to “Step 1—5”.
 Optimal Ram Design of a Searching System 1969

Table 2
Notation for a hybrid genetic algorithm with a heuristic method

Notation Definition
k Index of mission (k = 1, . . . , K)
n Index of component type (n = 1, . . . , N)
CType n Type of component with index n
θ Random number (0 < θ < 1)
AC Estimated system availability of current alternative
Ak Availability of a mission
AM STD Standard deviation of mission availability
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

AM STDA Allowable standard deviation of mission availability

AkC Current availability of mission k
AkI Improved availability of mission k
AkD Decreased availability of mission k
AkI Increased amount in availability of mission k
AkD Decreased amount in availability of mission k
TCC Life cycle cost of a current alternative
TCI Life cycle cost of a improvement alternative
TCD Life cycle cost of a decrease alternative
TCI Increased amount in life cycle cost
TCD Decreased amount in life cycle cost
MTBFMAXi Maximum MTBF of component i
MTTRMINi Minimum MTTR of component i
ALDTMINi Minimum ALDT of component i
MTBFCi , MTTRCi , ALDTCi Current alternatives of component i
MTBFCIi , MTTRCIi , ALDTCIi Improved alternatives of component i
MTBFCDi , MTTRCDi , ALDTCDi Decreased alternatives of component i

Step 1–5: If AT – ε ≤ AC ≤ AT + ε, set the selected j as optimal then terminate

the heuristic. Otherwise if AT – ε > AC or AT + ε < AC , go to “Step
2”.
Step 2: Check whether selected j satisfies AT or not. Then, select a mission for
improvement or decrease.
Step 2–1 If AT − ε > AC , select the longest mission k among in case as AT −
ε > Ak then go to “Step 3”. Otherwise go to “Step 2–2”.
Step 2–2 If AT + ε < AC , select the shortest mission k among in case as AT +
ε < Ak then go to “Step 4”.
Step 3: If AT − ε > Ak , improve MTBFCi , MTTRCi , and ALDTCi of CType n which has
the largest AkI /TCI for selected k at “Step 2”. (n = 1, 2, . . . , N [Total
number of component types])
Step 3–1 Generate θ for MTBFCi , MTTRCi , and ALDTCi of each CType n for
deciding improvement thresholds.
Step 3–2 Obtain MTBFCIi , MTTRCIi , and ALDTCIi of each CType n as follows:

MTBF CIi = MTBF Ci + {θ × (MTBF MAXi − MTBF Ci )} (6)

MTTRCIi = MTTRCi − {θ × (MTTRCi − MTTRMINi )} (7)
ALDT CIi = ALDT Ci − {θ × (ALDT Ci − ALDT MINi )} (8)
 1970 Yun et al.

Step 3–3 Estimate AkI and TCI of each CType n by simulation.

Step 3–4 Obtain AkI /TCI based on following Equation (9) of each CType n
by simulation.

AkI AkC − AkI
= (9)
TCI TCC − TCI

Step 3–5 Compare AkI /TCI of each CType n . Then, improve MTBFCi ,
MTTRCi , and ALDTCi as MTBFCIi , MTTRCIi , and ALDTCIi at “Step
3-2” of CType n that has the largest AkI /TCI .
Step 3–6 Estimate AM STD of improved alternative at “Step 3–5” by simulation
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

then compare it with AM STDA .

Step 3–7 If AM STDA ≥ AM STD , set the improved alternative as optimal then ter-
minate the heuristic. Otherwise if AM STDA < AM STD , go to “Step2”.
Step 4: If AT + ε < Ak , decrease MTBFCi , and increase MTTRCi , and ALDTCi of
CType n which has the smallest AkD /TCD for selected k at “Step2”.
Step 4–1 Generate θ for MTBFCi , MTTRCi , and ALDTCi of each CType n for
deciding decrease and increase thresholds.
Step 4–2 Obtain MTBFCDi , MTTRCDi , and ALDTCDi of each CType n as follows:

MTBF CDi = MTBF Ci − {θ × (MTBF Ci − MTBF INi )} (10)

MTTRCDi = MTTRCi + {θ × (MTTRI Ni − MTTRCi )} (11)
ALDT CDi = ALDT Ci + {θ × (ALDT I Ni − ALDT Ci )} (12)

Step 4–3 Estimate AkD and TCD of each CType n by simulation.

Step 4–4 Obtain AkD /TCD based on following Equation (13) of each
CType n by simulation.

AkD AkC − AkD
= (13)
TCD TCC − TCD

Step 4–5 Compare AkD /TCD of each CType n . Then, decrease MTBFCi , and
increase MTTRCi , and ALDTCi as MTBFCDi , MTTRCDi , and ALDTCDi
at “Step 4–2” of CType n which has the smallest AkD /TCD .
Step 4–6 Estimate AM STD of decreased alternative at “Step 4-5” by simulation
then compare it with AM STDA .
Step 4–7 If AM STDA ≥ AM STD , set the decreased alternative as optimal then
terminate the heuristic. Otherwise if AM STDA < AM STD , go to “Step
2”.

5. Numerical Examples
For numerical examples, a searching system that consists of 86 components is considered
and the exponential distribution is used for failure and repair distributions of all components.
We assume that ALDT of the searching system follows a lognormal distribution (3.5,
0.25) and ALDT in wartime is 50% of ALDT in peacetime. The input reliability and
maintainability data of each component are as shown in Table 3.
Table 4 shows information of PM for the searching system.
 Optimal Ram Design of a Searching System 1971

Table 3
Input reliability and maintainability data of components

Repair Repairable
CType n Description MTBF MTTR model probability
CType 1 Components of Sensor A EXP(138) EXP(36) Perfect 1.0
CType 2 Components of Sensor B EXP(146) EXP(44) Perfect 1.0
CType 3 Components of Sensor C EXP(160) EXP(40) Perfect 1.0
CType 4 Components of Sensor D EXP(40) EXP(34) Perfect 1.0
CType 5 Components of Sensor E EXP(152) EXP(44) Perfect 1.0
CType 6 Components of Sensor F EXP(96) EXP(40) Perfect 1.0
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

CType 7 Components of Sensor G EXP(110) EXP(32) Perfect 1.0

CType 8 Electronic Cabinet 1 to 3 EXP(40) EXP(24) Perfect 1.0

Table 5 shows the system structure of the searching system. For instance, we need
sensor F components, which have a standby structure with electronic cabinet 3 and 2 for
Function 3.
All components can be improved up to 60% from the given initial MTBF and MTTR.
In ALDT case, all components can be improved up to 50%. The repair (RCn ), development
(DCn ) and investment (ICMTTRn and ICALDTn ) costs are assumed as quadratic functions of
MTBF, MTTR, and ALDT of all components and an example of components in the sensor
A (CType 1 , n = 1) is given as below:

RC1 = 5600 + 280 (MTTR of CType 1 ) + 1.4(MT T R of CType 1 )2

DC1 = 22000 + 1100 (MTBF of CType 1 ) + 5.5 (MT BF of CType 1 )2
I CMTTR1 = 11200 + 560 (MTTR of CType 1 ) + 2.8 (MT T R of CType 1 )2
I CALDT1 = 1560 + 78 (ALDT of CType 1 ) + 0.39 (ALDT of CType 1 )2

We consider three types of operational scenario for the searching system. Figures 9 and
10 describe operational scenarios during the wartime and peacetime periods, respectively.
In order to perform missions stochastically during the system life cycle, a mission
transition probability matrix is used to select the next mission from the current mission as
shown in Table 6.
The simulation length is 10 years (87,600 hours) and the number of replication is 10.
The parameters for the genetic algorithm are as follows; crossover ratio (0.7), mutation
ratio (changed from 0.5 to 0.2 using deterministic adaptation), population size (100), and
generation size (100). The allowable gap, ε, is 0.005. Allowable standard deviation of
mission availability for the heuristic procedure in the hybrid genetic algorithm, AMk STDA ,
is 0.05.

Table 4
Preventive maintenance for the searching system

PM Type Maintenance place Interval (hour) Duration (hour)

4.5 M Field 3,285 12
28 M Field 20,440 24
 1972 Yun et al.

Table 5
System structure of the searching system
Component
# of Sensor Electronic
Operation Duration components cabinets
type Mission (hour) Function (k-out-of-n) (Main, Sub)
War Mission 1 90 Function 1 None None
Mission 2 117 Function 2 Sensor A (10 out of 14) Electric Cabinets (1, 2)
Sensor B (8 out of 11) Electric Cabinets (1, 2)
Sensor C (16 out of 21) Electric Cabinets (2, 1)
Sensor D (4 out of 7) Electric Cabinets (2, 1)
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

Sensor E (11 out of 16) Electric Cabinets (2, 1)

Function 4 Sensor G (3 out of 6) Electric Cabinets (3, 2)
Mission 3 168 Function 2
Function 4
Mission 4 169 Function 2
Function 4
Mission 5 116 Function 3 Sensor F (6 out of 8) Electric Cabinets (3, 2)
Mission 6 101 Function 2
Mission 7 223 Function 2
Function 3
Mission 8 215 Function 2
Function 4
Mission 9 70 Function 2
Function 3
Mission 10 210 Function 2
Peace Mission 11 143 Function 2
Function 4
Mission 12 128 Function 2
Function 3
Function 4
Mission 13 203 Function 3
Function 4
Mission 14 80 Function 2
Function 3
Function 4
Mission 15 113 Function 5 None None

Figure 9. Operational scenario during wartime period 73 × 7 mm (300 × 300 DPI).

Figure 10. Operational scenario during peacetime period 73 × 7 mm (300 × 300 DPI).
 Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

Table 6
Mission transition probability matrix

M01 M02 M03 M04 M05 M06 M07 M08 M09 M10 M11 M12 M13 M14 M15
M01 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
M02 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
M03 0.00 0.00 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
M04 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
M05 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
M06 0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
M07 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
M08 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00
M09 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00
M10 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.19 0.20 0.20 0.20 0.20
M11 0.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.20 0.20 0.20 0.20
M12 0.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.20 0.00 0.20 0.20 0.20
M13 0.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.20 0.20 0.00 0.20 0.20
M14 0.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.20 0.20 0.20 0.00 0.20
M15 0.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.20 0.20 0.20 0.20 0.00

1973
 1974 Yun et al.
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

Figure 11. Life cycle costs for different system availability target during wartime period 64 × 43 mm
(300 × 300 DPI). (Color figure available online.)

5.1. CASE 1: Change of the System Availability Target

We consider different system availability target as follows; 0.85, 0.90, and 0.95.
Figures 11–13 show that the life cycle cost increases as the system availability target
increases because the higher system availability target requires more repair, development,

Figure 12. Life cycle costs for different system availability target during peacetime period 64 ×
43 mm (300 × 300 DPI). (Color figure available online.)
 Optimal Ram Design of a Searching System 1975
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

Figure 13. Life cycle costs for different system availability target during war/peacetime period 64 ×
43 mm (300 × 300 DPI). (Color figure available online.)

Figure 14. Life cycle costs for different initial scale parameter during wartime period 64 × 43 mm
(300 × 300 DPI). (Color figure available online.)
 1976 Yun et al.
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

Figure 15. Life cycle costs for different initial scale parameter during peacetime period 64 × 43 mm
(300 × 300 DPI). (Color figure available online.)

Figure 16. Life cycle costs for different initial scale parameter during war/peacetime period 64 ×
43 mm (300 × 300 DPI). (Color figure available online.)
 Optimal Ram Design of a Searching System 1977

and investment costs to use highly reliable and maintainable components for satisfying the
system availability target. The hybrid genetic algorithm has 4% lower total life cycle cost
on average than the general genetic algorithm in CASE 1.

5.2. CASE 2: Change of the Scale Parameter

In this case, we change increase% of the initial MTBF scale parameter value of all compo-
nents in the searching system as follows; 0%, 10%, 20%, 30%, and 40%.
Figure 14–16 show that the life cycle cost decreases as the number of failure of
components decreases. The hybrid genetic algorithm has 5% lower total life cycle costs on
average than the general genetic algorithm in this case.
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

6. Conclusion
In this article, we dealt with a reliability and maintainability design problem for a searching
system with complex reliability structures such as standby and k-out-of-n reliability struc-
tures to keep higher system availability. The system availability and life cycle cost is one of
key measures to evaluate the system performance and was used as optimization criteria in
this article. However, it is not easy to obtain them by analytical methods due to the mathe-
matical difficulty and then the simulation method is used to estimate the system availability
and the life cycle cost for evaluating alternative solutions. When developing the simulation
model, an object-oriented design tool and a discrete-event simulation model was used. The
operational scenario of the searching system was also analyzed to express the operation en-
vironment of the searching system through the simulation model. For generating alternative
solutions of MTBF and MTTR values of all components and ALDT of the searching system,
a hybrid genetic algorithm with a heuristic method was developed to improve the quality
of the solutions. Using numerical examples, we compared the performance of a general
genetic algorithm with the hybrid genetic algorithm and the hybrid genetic algorithm has
4% lower life cycle costs on average than the general genetic algorithm. For further stud-
ies, we will consider more complex reliability structures such as consecutive k-out-of-n to
describe relationship between components of the searching system in detail.

Acknowledgments
This research was supported by Basic Science Research Program through the National
Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and
Technology (No.2010-0025084).

References
Borgonovo, E., Marseguerra, M., Zio, E. (2000). A Monte Carlo methodological approach to plant
availability modeling with maintenance, aging and obsolescence. Reliability Engineering and
System Safety 67:61–73.
Brown, J. F., Mahoney, J. F., Sivazlian, B.D. (1983). Hysteresis repair in discounted replacement
problems. IIE Transactions 15:156–165.
Chung, I. H. (2009). Simulation-Based Repair Policy for MIME System. Ph.D Thesis, Pusan National
University, Korea.
Garrido, J. M. (1998). Practical Process Simulation using Object-Oriented Techniques and C++.
Boston, MA: Artech House.
 1978 Yun et al.

Gen, M., Cheng, R., Lin, L. (2008). Network Models and Optimization: Multiobjective Genetic
Algorithm Approach. London: Springer.
Gen, M., Kim, J. R. (1999). GA-based reliability design: State-of-the-art survey. Computers &
Industrial Engineering 37:151–155.
Gupta, R. K., Bhunia, A. K., Roy, D. A. (2009). GA based penalty function technique for solving
constrained redundancy allocation problem of series system with interval valued reliability of
components. Journal of Computational and Applied Mathematics 232:275–284.
Hsieh, Y. C., Chen, T. C., Bricker, D. L. (1998). Genetic algorithms for reliability design problems.
Microelectronics Reliability 38:1599–1605.
Kuo, W., Prasad, V. R., Tillman, F. A., Hwang, C. L. (2001). Optimal Reliability Design: Fundamentals
and Applications. New York: Cambridge University Press.
Landers, T. L., Taha, H. A., King, C.L. (1991). A reliability simulation approach for use in the design
Downloaded by [Pusan National University Library ] at 23:11 09 December 2014

process. IEEE Transactions on Reliability 40:177–181.

Law, A. M., Kelton, W. D. (2000). Simulation Modeling and Analysis. 3rd ed. Ohio, OH: McGraw-
Hill.
Moghaddama, R. T., Safarib, J., Sassanic, F. (2008). Reliability optimization of series-parallel systems
with a choice of redundancy strategies using a genetic algorithm. Reliability Engineering and
System Safety 82:550–556.
Nakagawa, T. (1979a). Optimum policies when preventive maintenance is imperfect. IEEE Transac-
tions on Reliability 28:331–332.
Nakagawa, T. (1979b). Imperfect preventive maintenance. IEEE Transactions on Reliability 28:402.
Park, G., Yun, W.Y., Han, Y.J. (2011). Optimal preventive maintenance intervals of a rolling stock
system. Proceedings of 2011 ICQR2MSE l.1:427–430.
Pham, H., Wang, H. (1996). Imperfect maintenance. European Journal of Operational Research
94:425–438.
Rausand, M., Høyland, A. (2004). System Reliability Theory: Models, Statistical Methods, and
Applications. 2nd Ed. New York: Wiley Interscience.
Yanez, J., Ortuno, T., Vitoriano, B. (1997). A simulation approach to reliability analysis of weapon
systems. European Journal of Operational Research 100:216–224.
Yun, W. Y., Moon, I. K., Kim, G. R. (2008). Simulation-based maintenance support system for
multi-functional complex systems. Production Planning and Control 19:365–378.

View publication stats