Vous êtes sur la page 1sur 9

Applied Mathematical Sciences, Vol. 9, 2015, no.

138, 6867 - 6875

HIKARI Ltd, www.m-hikari.com
http://dx.doi.org/10.12988/ams.2015.57503

Minimal Repair-Replacement Model for

a System with Two Types of Failure
Bashir Yusuf 1, Ibrahim Yusuf 2 and Bashir M. Yakasai 3
1

2,3

Department of Mathematical Sciences, Bayero University, Kano, Nigeria

distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract
This paper considers an optimal replacement model for a system subject to two
types of failure. If the failure is of type 1, the system is minimally repaired. In the
case of type 2 failure, the system is minimally repaired with probability p and
replaced with probability q 1 p . A modified random and age replacement
model is considered in which the system is replaced at a planned time T, at a
random working time Y, or at the first non repairable type 2 failure whichever
occurs first. Explicit expression for the expected total cost and the optimal
replacement time T, minimizing the expected total cost are derived analytically.
Finally, numerical example is considered to highlight the characteristics of the
policy.
Keywords: Replacement, Expected cost, Minimal repair, Random working time,
Optimal time

1. Introduction
Most systems deteriorate and subsequently fail due to age and usage. Such
failures have negative effect on revenue, production of defective items and causes
delay in customer services. To reduce the incidences of system failures,
management of organizations is always interested with implementing an
appropriate preventive replacement policy for normal system operation. For this
reason, many researchers developed several optimal replacement policies for
reducing unnecessary high operating costs. Barlow and Hunter  developed a
periodic replacement policy in which the system is replaced at time T, while a mi-

6868

Bashir Yusuf et al.

nimal repair is conducted at any intervening failures. This is the basic minimal
repair policy and has been modified by many researchers in different ways.
Beichelt  developed a repair cost limit policy in which when a unit fails, the
repair cost is estimated and repair is done if the estimated cost is less than a
predetermined limit, otherwise a replacement is carried out. Nakagawa 
considered age replacement in which a unit is replaced at time T, or at a number
of N failures, whichever occurs first and undergoes minimal repair at failure. Sheu
and Griffith  developed a generalized age replacement policy with age
dependent minimal repair and random lead time and derived the explicit
expression of the average cost per unit time based on the stochastic behavior of
the assumed system which reflects the cost of storing a spare as well as the cost of
the system down time. Mamabolo and Beichelt  considered a maintenance
policy where each failure is removed by a minimal repair and on the first failure
after a given system age T, an unplanned replacement is carried out. Zhao and
Nakagawa  considered age and periodic replacement last models with working
cycles, where a unit is replaced before failure at a total operating time T, or at a
random working cycle whichever occurs last. Ali and ibrahim  presented a
sructural dependence replacement model for two units parallel system in which
both units operate simultataneously and the system is subjected to two types of
failures. The type I failure is minor and occur with the failure of a sigle unit and is
rectified by a minimal repair while the type II failure is catastroph in which both
units failed and the system is replaced. Suguira et al  considered a replacement
policy in which the system is replaced at time T or at random working time Y and
undergoes minimal repair at failure. Chen  considered age and random
replacement policy in which the system is replaced at time T, Y or at failure,
whichever occurs first. Nakagawa  considered a system in which corrective
replacement is performed at failure and the system is preventively replaced at a
total operating time T. Aven and Castro  considered a replacement model in
which the system is subject to two types of failure. If the failure is of type 1, the
system is minimally repaired. In the case of
type 2 failure, the system is
minimally repaired with probability p and replaced with probability q=1-p. The
system is replaced at a planned time T or at a non repairable type 2 failure
whichever occurs first. Chang  considered a preventive maintenance policy
with two types of failure. The type I failure (repairable) occurs with probability q
and is rectified by a minimal repair, whereas the type II failure (non-repairable)
occurs with probability p 1 q is removed by a corrective replacement. The
system is replaced at a random working time Y, or at a planned time T or at a first
type II failure which ever occurs first.
This paper extends the works of  and  by considering a system subject to
two types of failure. The type 1 failure (repairable) is removed by a minimal
repair. If the failure is of type 2, the system is rectified by a minimal repair with
probability p and replaced with probability q=1-p. A random and age replacement
model is considered in which the system is preventively replaced at a random
working time or at a planned time T or is correctively replaced at a non repairable
type 2 failure whichever occurs first. Explicit expression for the expected total cost is

6869

derived analytically and the optimal replacement time which minimizes it is

obtained and computed numerically.

2. Model formulation and assumptions

We consider a preventive replacement model for a system in which repairs and
replacements take place according to the following scheme:
1. A new system with failure time X begins to operate at time 0. When X has the
general distribution F (t ) and probability density function f(t ). A preventive
replacement is planned to be conducted when the system reaches age T.
2. The system failure at time t can be of two types: a type 1 failure (minor) which
is repairable occurs with intensity function r1 (t ) and cumulative failure intensity
t

function h1 (t ) r1 (u)du . In the case of type 1 Failure, the system is always

0

corrected by a minimal repair. The type 2 failure occurs with intensity function
t

r2 (t ) and cumulative failure intensity function h2 (t ) r2 (u)du . In the case of

0

type 2 failure, the system is minimally repaired with probability p and replaced by
a new one with a probability q 1 p , where 0 p 1 .
3. It is assumed that Y is a random working time of the system with a general
distribution G(t ) and does not take into account any actual failure (i.e. Y is
independent of X). It will be necessary to replace a system at random times as its
working times in cases where the working time is large (Nakagawa, 2005, p.245).
Another preventive replacement is conducted at the completion of the working
time.
4. In summary, the system is replaced at time T or Y or immediately after non
repairable type 2 failure, whichever occurs first, where T is constant and Y is a
random variable with distribution G(t ) .
5. The minimal repair cost associated with type 1 and type 2 failures are c1 and c m
respectively. The preventive replacement cost due to age T and due to random
working time are cT and cY respectively. The corrective replacement cost due to
type 2 failure is c Z . It is assumed that cZ cT and cZ cY .
6. After a replacement, the system is as good as new and the replacement time is
negligible.
The problem is to determine a replacement time T which balances the costs of
unplanned repairs/replacements and the cost of planned replacement.
Following Ross , we define the total expected cost per unit time as follows

C (T )

E (cost per cycle)

E (length of cycle)

(1)

6870

Consider a replacement cycle defined by the time interval between replacements

of the system caused by non repairable type 2 failures or by a planned
replacement at time T or Y . Let Z be the waiting time until the first type 2 failure
resulting in replacement and hence is independent of Y. From Beichelt  the
survivor function of Z can be expressed as

F (t ) P(Z t ) e(1 p ) h2 (t ) , t 0.

(2)

0

failures that occur in [0, t].

The probability that the system is replaced at age T is

(3)

The probability that it is replaced at random working time is

T

P(Y T, Y Z) F (t ) d G(t )

(4)

And the probability that it is replaced at the first non repairable type 2 failures is
T

(5)

T

TF (T ) G(T) tF (t )dG(t ) tG(t )dF (t ) F (t )G(t )dt

(6)

Let C (T ) denote the total expected cost, then the total expected cost is given by,

C (T )

0
T

c1 F (t ) G(t)r1 (t )dt cm F (t ) G(t) p r2 (t )dt cZ G(t )dF (t )

F (t ) G(t)dt

cY F (t )dG (t ) cT F (T ) G(T)
0

(7)

F (t )G (t )dt

Now using that,

d ( F (t ) G(t)) 1 F (T )G(T )

We obtain
T

F (T ) G(T) 1 F (t )dG(t ) G(t )dF (t )

and we can write the expected total cost as
C (T )

F (t )G (t )dt

(8)

Minimal repair-replacement model for a system

6871

3. Optimization
The problem is to find a value of T that minimizes C (T ) given by (8). Let T be
an optimal value of T .
The derivative of C (T ) is
C '(T )

(cY cT ) F (T ) dG(T) [c1r1 (T ) (cm p (cZ cT )(1 p))r2 (T ) C (T )]F(T) G(T)

(9)

F (t ) G(t)dt
0

At T , C '(T ) 0 ,
(cY cT ) r(T) c1r1 (T) (cm p (cZ cT )(1 p))r2 (T) C (T ) 0
Where,
dG (T )
r (T )
G(T )
The second derivative of C (T ) is
H (T )
C ''(T ) T
,
( F (t ) G(t)dt ) 2

(10)

(11)

where
T

H (T ) F (t)G(t)dt[(cY cT )(F(T) dG(T))' (c1r1' (T ) (cm p (cZ cT )(1 p))r2' (T )

0

C ' (T ))](F(T) G(T)) (c1 r1 (T ) (cm p (cZ cT )(1 p))r2 (T ) C(T))(F(T) G(T))'

(F(T) G(T))2 ((cY cT ) r(T) (c1r1 (T ) (cm p (cZ cT )(1 p))r2 (T ) C (T ))

At T , the second derivative becomes

C ''(T)

F (t ) G(t)dt

If (cy cT )( F (T ) dG(T))' [(c1r1 (T ) (cm p (cZ cT ))r2 (T )

C(T))(F(T) G(T))]' 0
Then C ''(T ) 0 and the T obtained from (9) minimizes of C (T ) .

(12)

Some replacement models are special cases of this model. They are demonstrated
as follows

Case 1. If r1 (t ) 0 . This case is considered by Chang (2014, p.187), in

which the system is replaced at time T, Y or at a non repairable type II failure
whichever occurs first and undergoes minimal repair with probability q at a
repairable type I failure. If we set r1 (t ) 0 in (8), the expected total cost is

6872

Bashir Yusuf et al.

C1 (T )

cT (cY cT ) F (t )dG (t )
0

F (t )G (t )dt
T

0

(13)

F (t )G (t )dt

Case 2. If p 0 and r1 (t ) 0 . This case is considered by Chen et al (2010),

in which the system is replaced at T, Y or at complete failure whichever occurs
first. If we set p 0 and r1 (t ) in (8), then the expected total cost is
T

C2 (T )

cT (cY cT ) F (t )dG(t ) (cZ cT ) G(t )dF (t )

0

(14)

F (t )G (t )dt

Case 3. If G(t ) 0 . This is the case considered by Aven and Castro (2008,
p.513), in which the type 1 failures are corrected by a minimal repair and the type
2 failure is rectified by a minimal repair with probability p and replaced with
probability q 1 p. The system is replaced at T, or at a non repairable type 2
failure whichever occurs first. and undergoes minimal repair at a repairable type 1
failure. If we set G(t ) 0 in (8), the expected total cost is

C3 (T )

c1 r1 (t ) F (t ) dt c2 p r2 t F (t )dt c2,r F (t ) cr F (t )
T

(15)

F (t )dt

Case 4. If G(t ) 0 , p 0 and r1 (t ) 0 .This is the standard age

replacement (SAR) considered by Barlow and Proschan (1965, p.87) and
Nakagawa(2005, p.72), in which corrective replacement is done immediately after
failure, and preventive replacement is done before failure at a total operating time
T. If we set G(t ) 0 , p 0 and r1 0 in (8), the expected total cost is

C4 (T )

cT (cZ cT ) F (T )

(16)

F (t)dt

Case 5. If F (t ) 0 , p 1 and r2 (t ) 0 . This is the case considered by

Suguira et al (2004) in which the system is replaced at time T or Y whichever
occurs first and undergoes minimal repair at complete failure. If we set F (t ) 0 ,
r2 (t ) 0 and p 1 in (8), the expected cost rate is

6873

C5 (T )

cT (cY cT ) G(T) c m G(t ) r (t ) dt

0

(17)

G (t )dt

Where, r (t ) r1 (t ) and cm c1 .

4. Numerical example
Let the failure distribution of the system has a gamma distribution of order 2 i.e.
f (t ) 2te t , t 0 , 0 and the random working distribution has exponential
distribution G(t ) 1 e t . c1 60 , cm 115 , cT 200 , cY 350 , cZ 1000 ,

1 5 , 2 1 .
The intensity functions of the processes are given by
ri (t )

i 2t
, i 0 , t 0 , i 1, 2.
1 i t

(18)

The failure intensities of the processes are given by

25t
t
and r2 (t )
r1 t
1 5t
1 t

(19)

Table 1 Optimal replacement time and minimum expected cost rate for
different values of p and

0.1
p

0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9

T
1.05
1.15
1.25
1.30
1.50
1.60
1.85
2.15
2.75

0.3

C (T )
646.6
624.3
601.1
576.9
551.3
524.3
495.5
464.6
430.3

T
1.15
1.20
1.30
1.40
1.60
1.80
2.05
2.50
3.40

0.5

C (T )
689.2
668.8
643.5
619.1
593.4
566.4
537.2
505.8
471.3

T
1.20
1.30
1.40
1.50
1.70
2.00
2.30
2.95
4.40

0.7

C (T )
732.6
710.2
686.8
662.5
636.7
609.4
580.4
649.0
514.6

T
1.25
1.40
1.50
1.70
1.90
2.25
2.75
3.70
4.95

0.9

C (T )
777.0
754.6
731.3
706.9
681.3
654.2
625.4
594.5
561.4

T
1.35
1.50
1.60
1.85
2.25
2.55
3.40
4.60
4.95

C (T )
822.5
800.2
777.1
752.9
727.5
700.9
672.7
643.0
611.8

It can be seen from Table1 that, the value T which minimizes C (T ) for different
values of p exists as predicted by our results. Also, we can see that when the
minimal repair probability p increases, the minimum expected cost rate decreases
and the optimal replacement time increases. This is intuitive, as lower probability
of replacement leads to lower expected total cost and increases the replacement
schedule.

6874

Bashir Yusuf et al.

5. Conclusion
In this paper, age and random preventive maintenance policy for a system with
two types of failures is presented. Explicit expression for the total expected cost
and the optimal replacement time are derived analytically. Numerical example is
given to highlight the characteristic of the policy.

References


R.E. Barlow, and C. Hunter, Optimal preventive maintenance policies,

Operations Research, 8 (1960), 90-100.
http://dx.doi.org/10.1287/opre.8.1.90



F. Beichelt, A replacement policy based on limits for repair cost rate, IEEE
Transactions on Reliability, 31 (1982), 401-403.
http://dx.doi.org/10.1109/tr.1982.5221393



T. Nakagawa, Periodic inspection policy with preventive maintenance, Naval

Research Logistics Quaterly, 31 (1984), 33-40.
http://dx.doi.org/10.1002/nav.3800310106



S. Sheu, and W.S. Griffith, Optimal age-replacement policy with age

dependent minimal repair and random lead time, IEEE Tansactions on
Reliability, 50 (2001), no, 3, 302-309.
http://dx.doi.org/10.1109/24.974128



R.M. Mamabolo, and F. Beichelt, Maintenance policies with minimal repair,

Economic Quality Control, 19 (2004), no. 2, 143-166.
http://dx.doi.org/10.1515/eqc.2004.143



X. Zhao, and T. Nakagawa, Optimization of replacement first or last in

reliability theory, European Journal of Operational Research, 223 (2012), 141149. http://dx.doi.org/10.1016/j.ejor.2012.05.035



U. Ali, and Ibrahim Yusuf, Structural dependence replacement model for

parallel system of two units, Nigerian Journal of Basic and Applied Sciences,
20 (2012), no. 4, 324-326.



T. M. Suguira, Optimal random replacement policies, Tenth ISSAT

International Conference on Reliability and Quality Design, (2004), pp. 99103.



6875

M. M. Chen, Random and age replacement policies, International Journal of

Reliability, Quality and Safety Engineering, 17 (2010), no. 1, 27-39.
http://dx.doi.org/10.1142/s0218539310003652



T. Nakagawa, Maintenance Theory of Reliability, Springer, London, 2005.

http://dx.doi.org/10.1007/1-84628-221-7



T. Aven, and I.T. Castro, A minimal repair replacement model two types of
failure and a safety constraint, European Journal of Operations Research, 188
(2008), 506-515. http://dx.doi.org/10.1016/j.ejor.2007.04.038



Chang, C.C, Optimum preventive maintenance policies for systems subject to

random working times, replacement, and minimal repair, Computers and
industrial engineering, 67 (2014), 185-194.
http://dx.doi.org/10.1016/j.cie.2013.11.011



1970.



F. Beichelt, A unifying treatment of replacement policies with minimal repair,

Naval Research Logistics, 40 (1993), 51-67. http://dx.doi.org/10.1002/15206750(199302)40:1<51::aid-nav3220400104>3.0.co;2-v