4th Reading

September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014

International Journal of Reliability, Quality and Safety Engineering
Vol. 19, No. 3 (2012) 1250014 (16 pages)
c World Scientic Publishing Company
DOI: 10.1142/S0218539312500143
OPTIMIZATION OF MAINTENANCE SCHEDULING
OF SHIP BORNE MACHINERY FOR IMPROVED
RELIABILITY AND REDUCED COST
AJIT KUMAR VERMA
,
, A. SRIVIDYA

,
COMMANDER ANIL RANA

and SANJAY K. KHATTRI

Stord/Haugesund University College, Haugesund, Norway

Indian Institute of Technology, Bombay, India

akvmanas@gmail.com
Received 20 April 2012
Revised 19 May 2012
Published 12 September 2012
Ships have a wide variety of machinery available onboard that is crucial for her
sustenance at sea for prolonged durations. The machinery can be grouped into vari-
ous plants, such as propulsion plant, air conditioning plants, power generation plants,
etc., each having its own specic function. The plants in turn are composed of various
systems which further comprise various types of machinery. There are redundancies built
in at the plant level, as well as at the system and at machinery level, so as to improve
the reliability of the ship as a whole. Planning of maintenance schedule, specically for
tasks which can only be undertaken in an ashore repair yard is a daunting task for the
maintenance managers. The paper presents a NSGA-II (nondominated sorting genetic
algorithm) based multi-objective optimization approach to arrive at an optimum main-
tenance plan for the vast variety of machinery in order to improve the average reliability
of ships operations at sea at minimum cost. The paper presents the advantages that
can accrue from introducing short maintenance periods for a select group of machinery,
within the constraints of mandatory operational time, over the method of following a
common maintenance interval for all the machinery. The problem function in hand is
nonlinear, multi-modal and multi-objective in nature. The search spaces for the prob-
lem is noncontinuous and the (multiple) variables, such as time interval for maintenance,
serial number of equipment, number of minor maintenance actions, etc., are uncoded real
parameters.
Keywords: Multi-objective optimization; TBPM (time-based preventive maintenance);
CBPM (condition-based preventive maintenance); NSGA-II.
1. Introduction
Machinery present onboard warships are organizationally similar to that present
in a large industrial set up, however, there is one very important dierence; the
ships need to depend on ashore-based repair facilities for a large portion of her
maintenance requirements. Though some minor maintenance can always be carried
out by her crew at sea itself, majority of maintenance on her machinery needs to
1250014-1
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
A. K. Verma et al.
be carried out ashore for which the ship has to return to harbor thereby incurring
huge opportunity costs. Moreover, untimely failure of her machinery at sea can
jeopardize her entire mission and therefore reliability of her machinery for mission
accomplishment is of utmost importance. In view of the above factors, it is impor-
tant that the maintenance schedule for a ships machinery is prepared with extreme
care and caution so as to balance her reliability needs and maintenance costs. The
problem for maintenance scheduling for ships machinery is compounded by the
presence of a vast variety of equipment organized into various systems and plants
each with its own set of redundancies. It is therefore important that the problem
in hand is solved keeping the organizational set up of the ships machinery in the
backdrop.
A large amount of literature is available dealing with the problems of mainte-
nance optimization, however not many of them have covered maintenance schedul-
ing of vast number of machinery available on a single platform, such as a warship.
From Barlow and Hunter
1
in 1960, till date, there have been many models and case
studies on optimization of PM intervals. References 15 are few such examples.
Authors such as Dekker,
6,7
Scarf
8
and Horenbeek et al.
9
have collectively surveyed
more than 300 works on the subject. It has been brought out by them that most of
the research work on the subject has been carried out at an individual component
or equipment level. Studies are often used only to demonstrate the applicability of
a developed model, rather than nding an optimal solution to a specic problem
of interest to a practitioner. Another limitation perceived in literature is that most
of the models focus on only one optimization criterion, making multi-objective
optimization models an unexplored are of maintenance optimization. As far as
maintenance planning/scheduling for ship borne machinery is concerned, Perakis
and Inozu
10
had developed reliability-based models to optimize the winter layup
replacement practices for major components of one and two diesel engines for Great
lake vessels (cargo ships). Although single objective optimization is attractive from
the modeling point of view, this approach does not capture all important aspects
of a real life situation.
2. Peculiarities of the Ship Borne Machinery
There are three main peculiarities of the ship borne machinery vis- a-vis an industrial
plant or any other large machinery installation.
First, as far as the maintenance requirements of the ship borne machinery are
concerned, they can be divided into two major groups: one that requires support
from the repair facilities-based ashore and the other that do not require such support
and can be easily carried out by the sta present onboard the ship. How it eects
the maintenance planning of the ship machinery is that the maintenance that need
to be carried out in harbor calls for the ship to be down (in terms of operational
availability) and this comes at a huge opportunity cost. All the maintenance actions
that can be undertaken at sea by the ships sta do not call for the ship to be put
1250014-2
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
Optimization of Maintenance Scheduling of Ship Borne Machinery
away from her operational zone. At best the ship can remain nonavailable for the
duration of the maintenance time. In view of the above reasons, the maintenance
actions that need support from shore-based facilities are considered to be more
critical than those that can be undertaken by the ships sta. Due to this dierence
in the requirement of shore-based facilities for carrying out maintenance, the two
groups of maintenance actions need to be treated separately and may be optimized
separately.
Second, the equipment where the condition-based PM is followed and where
the maintenance actions can be carried by the ships sta, the maintenance can be
scheduled as and when the condition of the equipment demands. However, for equip-
ment under CBPM where the maintenance actions need to be carried out ashore,
the maintenance manager needs to decide, during the ships every visit to harbor,
whether the maintenance actions can be carried during her stay or deferred till
next visit. This decision can be based on the condition of the equipment monitored
during her current visit to harbor.
Lastly, by virtue of all the machinery borne on the same platform, that is, the
ship itself, it is intuitively clear that the optimal time for maintenance in harbor
would be the one in which, all the machinery need to be attended to at the same
time. It would seem to be sub-optimal to bring the ship to harbor again and again
for carrying out maintenance that falls (TBPM) due on dierent equipment at
dierent times. It is for this reason that the maintenance managers plan what is
called the rets of the ships. These rets are planned in advance and it is the time
when the ship is known to be unavailable for any mission requirement. The ret is
generally the time when most of the equipments are put down time for maintenance.
However, whether these onetime planning of all maintenance jobs are optimal or
not and what is the basis of planning these rets needs some investigation and this
is the topic of this paper.
The problem in hand for deciding maintenance intervals of a large number
of ship machinery is nonlinear, multi-modal and multi-objective in nature. The
search spaces for such problems are noncontinuous and the problem (multiple) vari-
ables are uncoded real parameters. One of the most suitable methods to deal with
these problems is the evolutionary algorithms, such as the NSGA-II (see Refs. 11
and 12).
3. Maintenance Scheduling of Ships Machinery
As has been brought out earlier, a ship consists of a variety of plants each with a
dierent function. There might be redundancies available in the plants that improve
the reliability of the plant. These plants in turn will consist of many equipments.
There could be some equipment which follow a time-based preventive maintenance,
TBPM and there would be others that follow a CBPM. However, by virtue of
being present on the same platform, as that of a ship, we will have to nd out what
is the most optimal time interval T for carrying out maintenance on all these
1250014-3
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
A. K. Verma et al.
T
R
T
Ops
T
A
T
Ops
T
A T
Ops
T
R
Refit time. Major ship equipments are under repair. Ship is down.
T
Ops
Operational time. The ship is operational
T
A
A short maintenance period when a few select equipment are under maintenance.
The ship is down.
T
R
T
Ops T
Ops
T
R
Fig. 1. Operation and maintenance cycle of a ship.
equipment. Such a time period for maintenance on the ship equipment is called
Ret in naval parlance.
The search for such an optimal time schedule seems to be a simple optimization
problem, but it is not. The reason, as will be shown with an example, are the
advantages that accrue from following a staggered, but grouped time schedules for
some, but not all the equipment. Figure 1 above would provide a much needed
clarication.
In Fig. 1 above T
R
is the scheduled ret of the ship, wherein the majority or
all of the equipment will be brought under maintenance. The T
Ops
is the operational
time for the ship where she fullls all her mission commitments. In the rst cycle
of the gure, the maintenance managers nd out the optimal time for ret T
R
. The
objectives are cost rate and average reliability during the operational time of the
ship. This is the general trend amongst the maintenance managers. What has been
observed in the research work is that, if the operational time of a ship is given to
be say not less than T
Ops
(a constraint) then by bringing up a small select group of
equipment under maintenance both the above stated objectives can be improved.
Such short maintenance intervals may be called short maintenance period (MP)
and these can be scheduled in between the ret periods of the ship. Though in
Fig. 1 it appears that the bottom cycle has a shorter operational time T
Ops
this
may not be the case.
The strategy shown in Fig. 1 however requires solution of a multi-objective
optimization problem with multiple variables. The variables for the optimization
problem can be listed down as follows:
Selection of equipment that need to be maintained during MP (maintenance
period).
Scheduling of the maintenance period T
mp
after completion of a ret.
Total number of MPs (maintenance periods required between the ret period).
Scheduling of ret period T
rp
after completion of the previous ret.
1250014-4
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
Optimization of Maintenance Scheduling of Ship Borne Machinery
Among the other variables are the MP and/or the ret duration, however, these
are values that are generally given and can be considered to be constraints of the
multi-objective problem being discussed.
3.1. Example
The above approach to the problem can be demonstrated taking an example of a
ship system. We consider a group of three main plants of a ship comprising 13 equip-
ments. The power generation plant has a redundancy. The life distribution and its
assumed parameters are listed in Table 1. The search is for an optimal maintenance
plan for the entire ship such that the ship is available for operations at least for
a minimum period of T
Ops
at the maximum average reliability AvgRelb and
at the minimum maintenance cost rate Costrate. Maintenance actions covered
are only replacement of the equipment, except for the turbines, which based on the
level of deterioration, undergo replacement of its worn down components.
Table 1. Parameters of distribution of the 13 equipment example.
S. Equipment/ Distribution Scale Shape Cost Manpower Maintenance
No. component parameter parameter required time
1 Plant 1
Ancillary system
Weibull 150 1.109 1200 3 2
2 Plant 1
Distribution
system
Weibull 465 1.3 4000 2 2
3 Plant 1 Turbine Gamma 0.089 0.97 550 3 1
4 Plant 1
Ancillary system
Weibull 150 1.109 1200 3 2
5 Plant 1
Distribution
system
Weibull 465 1.3 4000 2 2
6 Plant 1 Turbine Gamma 0.089 0.97 550 3 1
7 Propulsion
plant ASD
system
Weibull 907 2.9 19,000 2 2.5
8 Propulsion
plant Turbine
Gamma 0.075 0.52 5500 4 2.5
9 Propulsion
plant
Ancillary system
Weibull 95 1.1 650 3 3
10 Propulsion
plant Shaft
line system
Weibull 200 1.43 1100 2 2
11 AC plant
compressor
Weibull 340 2.109 1200 4 1
12 AC plant
Ventilation
system
Weibull 139 1.02 1100 4 2.5
13 AC plant
Cooling pump
Gamma 0.045 0.82 1200 2 1
1250014-5
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
A. K. Verma et al.
Fast Attack Craft Machinery
Power Generation Plant AC plant Propulsion plant
Plant 2 Plant 1
Ancillary
system-
TBPM
Distribution
system-
TBPM
Turbine-
CBPM
Ancillary
system-
TBPM
Distribution
system-
TBPM
Turbine-
CBPM
ASD
system-
TBPM
Turbine-
CBPM
Ancillary
system-
TBPM
Shaft line
system-
TBPM
Compresso
r-TBPM
Ventilation
system -
TBPM
Cooling
ppCBPM
1
2
3
4
5
6
7
8
9
10
11
12
13
Fig. 2. Schematic list Group of three plants of an attack craft.
Initially we consider the case without any MP (or short maintenance periods).
In this case the search is for a common time where all the equipment would be
maintained together. The problem equation can be written as:
OBJECTIVES
min
T
[Costrate(T)]
max
T
[AverageRe lb(T)]
s.t.
costrate(T) =
Exp cost of failure (T) + C
PM
Re lb(T)
_
T
0
Re lb(t)dt
, (1)
AverageRe lb(T) =
_
T
0
Re lb(t)dt
T
, (2)
where
Re lb(T) =
i=n

i=1
R
i
(T)
_
_
1
_
_
1
j=n

j=1
R
j
(T)
_
_

_
1
k=n

k=1
R
k
(T)
_
_
_
, (3)
1250014-6
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
Optimization of Maintenance Scheduling of Ship Borne Machinery
Exp cost of failure (T) =
n

j=1
_
T
0
f
j
(x)
n

i=j
R
i
(x)
_
1

R
k
(x)
_
(Cost
j
+ N
Rj
T
Rj
Cost
R
+C
Oprn
( T
Rj
+ T
RH
+ T
LD
))dx, (4)
where j and k are equipment belonging to the plants which have a hot redundancy
with each other and
N
Ri
= number of people required for maintenance of equipment i,
T
Ri
= Time required for maintenance of equipment i,
T
RH
= Time required for ship to reach harbor,
T
LD
= Logistics delay time,
age
i
= age of equipment i measured from completion of previous ret
since all equipment ,
Cost
i
= i equipment cost,
C
Opm
= cost of opportunity,
Cost
R
= Cost of repair action per time per repair crew,
f
i
() = pdf of lifetime for equipment i,
F
i
() = cdf of lifetime for equipment i,
R
i
() = reliability of equipment i,
C
PM
=
n

j=1
(Cost
j
+ N
Rj
T
Rj
Cost
R
).
Given an assumption of cost of opportunity of Rs 250,000/- and a cost of repair
of Rs 3500/- per man per day and the cost of equipment and manpower as shown
in Table 1, we get the plots as shown in Fig. 3. The range of multiple objectives
lay to the left of the optimal point T = 30 days. At this point the cost rate is
0.5478*10
5
and the average reliability during the operational duration is 0.7610.
Any time to the left of T = 30 is also an acceptable point since the average
reliability remains higher than 0.7610 but at a higher cost rate. At any time to the
Fig. 3. Plot of Cost rate and Average reliability.
1250014-7
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
A. K. Verma et al.
right of T = 30 days we approach a zone of lower average reliability at a higher
cost rate and hence all time points to the right of T = 30 are clearly sub-optimal.
We now consider the case when we can introduce some minor/short MPs or
maintenance periods in between the ret schedules. The problem equation can now
be written as given in the next section.
3.2. Example of multi-objective optimization
of maintenance interval scheduling
We now allow some MP in between the ret periods and nd out what are the
eects on the multiple objectives of average reliability and maintenance cost rate.
The acceptance of minor maintenance intervals in between the ret period gives
rise to multiple variables such as selection of equipment to be included in the
minor maintenance period, time interval for carrying out such minor maintenance,
time interval for scheduling the ret of the ship and number of maintenance period
allowed to be carried out before the ret is carried out. The multi-objective problem
for the present case can be written as follows:
OBJECTIVES
min
M[]
[Costrate(M[n
t
, n
amp
, E[], T])]
max
M[]
[Average Re lb(M[n
t
, n
amp
, E[], T])],
s.t.
(5)
costrate(T) =
Exp cost of failure (M[n
t
, n
amp
, E[], T])
+PMcost(M[n
t
, n
amp
, E[], T])

n
n
amp=1
_
_
(T
Ops
+n
t
)(n
amp
)
(T
Ops
+n
t
)(n
amp
1)
Re lb(t)dt
_
+
_
_
TT
Ramp
(T
Ops
+n
t
)(n)
Re lb(t)dt
_,
(6)
Average Re lb(M[n
t
, n
amp
, E[], T]) =

n
n
amp=1
_
_
(T
Ops
+n
t
).(n
amp
)
(T
Ops
+n
t
)(n
amp
1)
Re lb(t)dt
_
[T
Ops
+ n
t
] n
+
_
_
TT
Ramp
(T
Ops
+n
t
)(n)
Re lb(t)dt
_
(T T
Ramp
) [(T
Ops
+ n
t
) (n)]
, (7)
where
Re lb(x) =
i=N

i=1
R
i
(age
i
+ x)
R
i
(age
i
)

_
_
1
_
_
1
j=N

j=1
R
j
(age
j
+ x)
R
j
(age
j
)
_
_

_
1
k=N

k=1
R
k
(age
k
+ x)
R
k
(age
k
)
__
age
i
=0iE[]
, (8)
1250014-8
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
Optimization of Maintenance Scheduling of Ship Borne Machinery
where j and k are equipments belonging to the plants which have a hot redundancy
with each other
PMcost(M[n
t
, n
amp
, E[], T])
=
_
_
n

n
amp
=1
N

j=1
[Cost
j
+ N
Rj
T
Rj
Cost
R
] Re lb((T
Ops
+ n
t
) n
amp
)
_
_
jE[]
+
N

i=1
[Cost
i
+ N
Ri
T
Ri
Cost
R
] Re lb((T T
Ramp
(T
Ops
+ n
t
) n),
(9)
Exp cost of failure (M[n
t
, n
amp
, E[], T])
=
_
_
n

n
amp
=1
N

j=1
_
(T
Ops
+n
t
)(n
amp
)
(T
Ops
+n
t
)(n
amp
1)
f
j
(age
j
+ x)
R
j
(age
j
)
N

i=j
R
i
(age
i
+ x)
R
i
(age
i
)

_
1

R
k
(age
k
+ x)
R
k
(age
k
)
__
age
i
=0iE[]
[Cost
j
+ N
Rj
T
Rj
Cost
R
+ C
Oprn
(T
Rj
+ T
RH
+ T
LD
)]dx
+
_
_
N

j=1
_
T
(T
Ops
+n
t
)n
f
j
(age
j
+ x)
R
j
(age
j
)
N

i=j
R
i
(age
i
+ x)
R
i
(age
i
)

_
1

R
k
(age
k
+ x)
R
k
(age
k
)
__
age
i
=0iE[]
[Cost
j
+ N
Rj
T
Rj
Cost
R
+ C
Oprn
( T
Rj
+ T
RH
+ T
LD
)]dx, (10)
where
M[] = set of selected variables that aect the cost rate and average reliability
of the ship,
T
Ops
= minimum operational time the ship needs to be operational after a
ret or maintenance period,
n
t
= time after T
Ops
when the MP is scheduled,
n
amp
= number of MPs that are scheduled before the ret of ship falls due,
E[] = set of equipments selected to undergo maintenance during MP,
T = time (after the previous ret) when the ships ret falls due,
T
Ramp
= cumulative maintenance time for equipment chosen for maintenance
during MP,
N
Ri
= number of people required for maintenance of equipment i,
T
Ri
= time required for maintenance of equipment i,
T
RH
= time required for ship to reach harbor,
1250014-9
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
A. K. Verma et al.
T
LD
= logistics delay time,
age
i
= age of equipment i measured from completion of previous ret.
Since all equipment E[] are maintained at every MP, age for such
equipment =0 after every MP,
Cost
i
= i equipment cost; C
Opm
= cost of opportunity,
Cost
R
= cost of repair action per time per repair crew,
f
i
() = pdf of lifetime for equipment i,
F
i
() = cdf of lifetime for equipment i,
R
i
() = reliability of equipment i.
4. Use of NSGA II Elitist Genetic Algorithm Program
The above noncontinuous, discrete, multi-objective, multi-variable optimization
problem can be approached using the elitist NSGAII program as explained earlier.
A chromosome for the GA population is shown in Fig. 4 below. The chromosome
contains all the multi-variables required for evaluating the health (or magnitude of
its objectives).
The rst 13 slots of the chromosome are for the equipment which in this case
are a total of 13. The slots can contain either a 1 or a 0 showing whether that
specic equipment is selected to undergo replacement/repair during the MP or not
respectively. The 14th slot shows the number of MP (short maintenance period)
selected by the chromosome. A 0 in this position would mean that there are no
MPs for equipment and therefore the number 0 or 1 in any of the rst 13 slots
would then be irrelevant.
The 15th slot shows the time after the mandatory T
Ops
period when the MP is
being scheduled. Since time is continuous, this slot may have innite choices. We
therefore choose a small time interval t the multiples of which in integers can
be shown in this 15th slot. A 0 in this slot would therefore mean that the MP
has been scheduled right after completion of the mandatory T
Ops
period. It has
been brought out before that the T
Ops
period is the minimum time the ship should
E E E n
a
n
t
T
13 Slots for 13 equipment
This slot decides the number of MPs
Time after T
Ops
when the MP is scheduled
Time when the refit is scheduled. The time is
counted from the end of the previous refit
Fig. 4. A chromosome used for solving the NSGAII-based MOOP.
1250014-10
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
Optimization of Maintenance Scheduling of Ship Borne Machinery
remain operational after maintenance to meet her operational commitments. The
16th slot shows the time when the ret is scheduled. The time shown here is counted
from the elapse of the previous ret period.
5. Assumptions and Result
We make the following assumptions for the example under discussion:
Maintenance actions considered are only replacement of equipment.
All maintenance actions require assistance of repair crew present ashore.
All maintenance actions are considered sequential in nature. In actual practice
simultaneous repair actions can be undertaken resulting in reduced downtime
due to maintenance actions.
Time required for maintenance is considered deterministic in nature.
The failure of all the equipments is statistically independent in nature.
The equipment which is being monitored for wear or deterioration follows a non-
stationary gamma wear process. All other equipments follow a Weibull process
for failure.
The equipments which have been selected for maintenance during MP remain
xed for every MP. In practice, we may consider to choose the equipment that
can be maintained during every MP. The chromosome for solving the GA problem
in this case would be three-dimensional in nature.
The crossover probability for the example selected above is 0.8 and the mutation
probability is 0.01. A random population of 40 chromosomes brings out interesting
results after just four generations, as shown below through 2 out of the total 40
solutions.
Soln 1 = [1 0 1 1 0 1 0 1 1 1 1 0 1 2 0 145];
Cost rate = 0.4647*10
5
; Average reliability =0.6738.
Soln 2 = [1 0 1 0 0 1 1 1 0 1 1 1 1 3 6 225];
Cost rate =0.3993*10
5
; Average reliability =0.6275.
The above two solutions bring out two such nondominated solutions. The rst
solution shows that if nine equipments are selected to undergo 2 MPs exactly after
completion of T
Ops
of 30 days and the ret is scheduled after every 145 days the
cost rate drops to 0.4647*10
5
down from 0.5478*10
5
(that happens when no MP is
selected). The average reliability of course drops from 0.7610 to 0.6738.
The second solution also brings out another nondominated solution. When nine
equipments are selected to undergo 3 MPs, 12 days after T
Ops
of 30 days (or after
every 42 days) and if the ret is scheduled after every 225 days, the cost rate drops
further to 0.3993*10
5
but the average reliability also drops to 0.6275.
When the T
Ops
becomes 20 days we see improvement in both the objectives
(Fig. 5). This shows that it is not optimal to put down all the equipments together
1250014-11
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
A. K. Verma et al.
Fig. 5. Set of optimal solutions obtained from NSGA II program.
for maintenance during a ret without keeping any short maintenance period
MP in between the ret periods. The number of MPs and the scheduling of the
MPs play a vital part in improving the cost rate and average reliability of the ship
during the operation phase in between the ret periods. The NSGA II program
routines for the example were written in Matlab 7. The objective values and the
elite solutions are placed in Tables 2 and 3 below.
Table 2. Objective values for a population of 40 after 20 generations.
Solution no. 1 2 3 4 5 6 7 8 9 10
Cost rate 0.4815 0.2592 0.2861 0.2585 0.2449 0.1967 0.2426 0.4815 0.1418 0.2224
1-Avg
Reliab 0.1785 0.2257 0.1943 0.2418 0.2587 0.4259 0.2643 0.1785 0.5476 0.3083
Solution no. 11 12 13 14 15 16 17 18 19 20
Cost rate 0.1927 0.4239 0.2426 0.2585 0.1967 0.2283 0.4815 0.1927 0.1418 0.1967
1-Avg
Reliab 0.4426 0.1789 0.2643 0.2418 0.4259 0.2932 0.1785 0.4426 0.5476 0.4259
Solution no. 21 22 23 24 25 26 27 28 29 30
Cost rate 0.2449 0.1913 0.1967 0.1927 0.2794 0.4815 0.1927 0.2886 0.1913 0.2738
1-Avg
Reliab 0.2587 0.4864 0.4259 0.4426 0.205 0.1785 0.4426 0.194 0.4864 0.2051
Solution no. 31 32 33 34 35 36 37 38 39 40
Cost rate 0.2224 0.2738 0.1458 0.1999 0.2283 0.201 0.2149 0.2338 0.1988 0.4815
1-Avg
Reliab 0.3083 0.2051 0.4872 0.3977 0.2932 0.3968 0.3252 0.2723 0.4112 0.1785
1250014-12
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
Optimization of Maintenance Scheduling of Ship Borne Machinery
Table 3. 40 solutions of the MOOP obtained through NSGAII.
Soln no. 1 [1 1 0 1 1 1 0 1 1 1 0 0 1 2 1 101]
Soln no. 2 [1 1 0 0 0 0 1 0 0 1 0 0 1 5 1 183]
Soln no. 3 [0 1 0 1 1 1 1 1 0 1 0 0 1 5 1 209]
Soln no. 4 [0 1 1 1 1 1 1 0 0 1 0 0 1 5 2 211]
Soln no. 5 [0 1 1 1 0 1 1 0 0 1 0 0 1 5 2 209]
Soln no. 6 [0 1 0 0 0 0 0 1 0 0 1 0 1 5 13 312]
Soln no. 7 [1 1 1 1 0 1 0 0 0 1 0 0 1 5 2 209]
Soln no. 8 [1 1 0 1 1 1 0 1 1 1 0 0 1 2 1 101]
Soln no. 9 [0 0 0 0 0 0 1 0 0 0 1 0 1 5 11 312]
Soln no. 10 [0 1 1 0 0 0 1 0 0 1 0 0 1 5 2 211]
Soln no. 11 [0 0 0 0 0 0 0 1 0 0 1 0 1 5 13 312]
Soln no. 12 [1 1 0 1 0 1 0 1 1 1 0 0 1 5 1 209]
Soln no. 13 [1 1 1 1 0 1 0 0 0 1 0 0 1 5 2 209]
Soln no. 14 [0 1 1 1 1 1 1 0 0 1 0 0 1 5 2 211]
Soln no. 15 [0 1 0 0 0 0 0 1 0 0 1 0 1 5 13 312]
Soln no. 16 [0 1 1 1 0 1 0 0 0 1 0 0 1 5 2 209]
Soln no. 17 [1 1 0 1 1 1 0 1 1 1 0 0 1 2 1 101]
Soln no. 18 [0 0 0 0 0 0 0 1 0 0 1 0 1 5 13 312]
Soln no. 19 [0 0 0 0 0 0 1 0 0 0 1 0 1 5 11 312]
Soln no. 20 [0 1 0 0 0 0 0 1 0 0 1 0 1 5 13 312]
Soln no. 21 [0 1 1 1 0 1 1 0 0 1 0 0 1 5 2 209]
Soln no. 22 [0 1 0 1 1 1 0 0 0 1 0 0 1 5 9 312]
Soln no. 23 [0 1 0 0 0 0 0 1 0 0 1 0 1 5 13 312]
Soln no. 24 [0 0 0 0 0 0 0 1 0 0 1 0 1 5 13 312]
Soln no. 25 [0 1 1 1 0 1 1 1 0 1 1 0 1 5 2 218]
Soln no. 26 [1 1 0 1 1 1 0 1 1 1 0 0 1 2 1 101]
Soln no. 27 [0 0 0 0 0 0 0 1 0 0 1 0 1 5 13 312]
Soln no. 28 [1 1 0 1 0 1 1 1 0 1 0 0 1 5 1 209]
Soln no. 29 [0 1 0 1 1 1 0 0 0 1 0 0 1 5 9 312]
Soln no. 30 [1 1 1 1 0 1 0 1 0 1 0 0 1 5 2 209]
Soln no. 31 [0 1 1 0 0 0 1 0 0 1 0 0 1 5 2 211]
Soln no. 32 [1 1 1 1 0 1 0 1 0 1 0 0 1 5 2 209]
Soln no. 33 [0 0 0 0 0 0 1 0 0 0 1 0 1 5 11 275]
Soln no. 34 [0 0 0 0 0 1 0 1 0 0 1 0 1 5 13 286]
Soln no. 35 [0 1 1 1 0 1 0 0 0 1 0 0 1 5 2 209]
Soln no. 36 [0 0 0 0 0 1 0 1 0 0 1 0 1 5 13 284]
Soln no. 37 [0 1 1 0 0 1 0 0 0 1 0 0 1 5 2 209]
Soln no. 38 [0 1 1 0 1 0 1 0 0 1 0 0 1 5 2 209]
Soln no. 39 [0 1 0 0 0 1 0 1 0 0 1 0 1 5 13 312]
Soln no. 40 [1 1 0 1 1 1 0 1 1 1 0 0 1 2 1 101]
1250014-13
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
A. K. Verma et al.
6. Conclusion
The paper has presented a NSGA II-based MOOP approach for maintenance
scheduling of a vast number of ship borne machinery that need the support of
a repair yard present ashore. The objectives of the problem were to maximize the
average reliability of the ship during her operational phase and to minimize the
cost rate, which included the maintenance costs. Using the failure information of
various equipment and depending upon how they are organized into plants and sys-
tems each with its own redundancies, it has been shown through an example that it
is benecial to introduce short MPs for a select group of equipment in between the
ret periods of the ships instead of following a common maintenance interval for all
her machinery. Multiple variables such as the choice of equipment, the number of
MPs required to be scheduled, time interval for scheduling the MPs and the time
interval for scheduling the ret of the ship can be all chosen in accordance with
the solutions arrived at by the solution of the MOOP using NSGA II an elitist
genetic algorithm method.
References
1. R. Barlow and L. Hunter, Optimum preventive maintenance policies, Oper. Res. 8
(1960) 90100.
2. H. Mine and H. Kawai, Preventive replacement of a 1 unit system with a wear out
state, IEEE Trans. Reliab. R-23(1) (1974).
3. T. Nakagawa, Optimum preventive maintenance policies for repairable systems, IEEE
Trans. Reliab. R-26 (1977).
4. K. S. Park, Optimal continuous wear limit replacement under periodic inspections,
IEEE Trans. Reliab. 37(1) (1988) 97102.
5. I. B. Gertsbakh, Models of Preventive Maintenance (North Holland Publishing Com-
pany, 1977).
6. R. Dekker, Applications of maintenance optimization models: A review and analysis,
Reliab. Eng. Syst. Saf. 51(3) (1996) 22940.
7. R. Dekker, On the use of operations research models for maintenance decision making,
Microelectron. Reliab. 35(910) (1995) 13211331.
8. P. A. Scarf, On the application of mathematical models in maintenance, Eur. J. Oper.
Res. 99 493506.
9. V. A. Horenbeek, L. Pintelon and P. Muchiri, Maintenance optimization models and
criteria, in Proc. 1
st
Int. Congress on e-Maintenance, Lulea, Sweden, 2224 June
2010, pp. 513.
10. A. N. Perakis and B. Inozu, Optimal maintenance repair and replacement of Great
Lakes marine diesel, Eur. J. Oper. Res. 55 (1991) 165182.
11. D. E. Goldberg, Genetic Algorithm for search, Optimization and Machine Learning
(Addison-Wesley, Reading, MA, 1989).
12. K. Deb, Multi-Objective Optimization using Evolutionary Algorithms (John Wiley &
Sons, Ltd, 2002).
About the Authors
Ajit Kumar Verma is a Professor in Engineering, Haugesund/Stord University Col-
lege, Haugesund, Norway and has been a Professor (since Feb 2001) with the
1250014-14
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
Optimization of Maintenance Scheduling of Ship Borne Machinery
Department of Electrical Engineering (currently) at IIT Bombay (India) with a
research focus in Reliability Engineering and Quality Management. He was the
Director of the International Institute of Information Technology Pune, on lien from
IIT Bombay, from August 2009September 2010. He is also a Guest Professor at
Lulea University of Technology, Sweden. He has supervised/co-supervised 31 PhDs
and 92 Masters theses in the area of Software Reliability, Reliable Computing,
Power Systems Reliability (PSR), Reliability Centered Maintenance (RCM) and
Probabilistic Safety/Risk Assessment (PSA). He has jointly edited books on Reli-
ability Engineering & Quality Management and Quality, Reliability & Information
Technology, and is also an author of books titled Fuzzy Reliability Engineering
Concepts and Applications (Narosa), Reliability and Safety Engineering (Springer)
and Dependability of Networked Computer Based Systems (Springer). He has over
225 publications in various journals (over 100 papers) and conferences. He has been
the Editor-in-Chief of OPSEARCH published by Springer (January 2008January
2011) as well as the Founder Editor-in-Chief of International Journal of Systems
Assurance Engineering and Management (IJSAEM) published by Springer and the
Editor-in-Chief of SRESA Newsletter and Journal of Life Cycle Reliability and
Safety Engineering. He is a recipient of Leadership in Reliability Engineering Educa-
tion & Research award by Society of Reliability Engineering, Quality & Operations
Management.
Professor A. Srividya is a guest Professor at Stord/Haugesund University Col-
lege, Haugesund, Norway on lien from IIT Bombay since July end 2012. She has
supervised/co-supervised 28 PhDs and 50 Masters theses in the area of Struc-
tural Reliability, Reliability Based Optimisation, Simulation Studies for Reliability
Estimation, Quality Benchmarking Studies for Service Industries, Quality Systems
and Accelerated Life Testing. She has jointly edited books on Reliability Engi-
neering and Quality Management, Quality, Reliability and Information Technol-
ogy, and is also author of books titled Fuzzy Reliability Engineering-Concepts and
Applications (Narosa), Reliability and Safety Engineering (Springer), Dependabil-
ity of Networked Computer Based Systems (Springer) and Optimal Maintenance
of Large Engineering Systems (NarosaIn press). She has over 196 publications in
various international and national journals and conferences. She is a recipient of
Leadership in Reliability Engineering Education & Research award by Society of
Reliability Engineering, Quality & Operations Management.
Anil Rana, Commander, Indian Navy, has been working in the maintenance man-
agement eld for the past 20 years. He is presently on the verge of completing his
PhD in the eld of reliability engineering from the prestigious Indian Institute of
Technology, Mumbai with maintenance management as his key research eld. As
an ocer of the Engineering Branch of the Indian Navy, he has had a hand on
experience on maintenance of a wide variety of ship borne equipment and this is
1250014-15
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
4th Reading
September 10, 2012 9:47 WSPC/S0218-5393 122-IJRQSE 1250014
A. K. Verma et al.
reected in the papers he has published on the subject in national and international
journals.
Sanjay K. Khattri received his PhD from the University of Bergen in 2006. He is
currently working as an Associate Professor at the Stord/Haugesund University
College. His research interests include, but not restricted to, scientic computing,
numerical analysis and optimization.
1250014-16
I
n
t
.

J
.

R
e
l
.

Q
u
a
l
.

S
a
f
.

E
n
g
.

2
0
1
2
.
1
9
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
w
o
r
l
d
s
c
i
e
n
t
i
f
i
c
.
c
o
m
b
y

M
r

I
n
d
e
r

M
u
k
h
o
p
a
d
h
y
a
y

o
n

0
9
/
0
2
/
1
4
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.