Computers and Mathematics with Applications 61 (2011) 2388–2400

Contents lists available at ScienceDirect

Computers and Mathematics with Applications

journal homepage: www.elsevier.com/locate/camwa

A volume flexible economic production lot-sizing problem with

imperfect quality and random machine failure in fuzzy-stochastic
environment
Debasis Das a , Arindam Roy b , Samarjit Kar a,∗
a
Department of Mathematics, National Institute of Technology, Durgapur, W.B, Pin-713209, India
b
Department of Computer Science, Prabhat Kumar College, Contai, Purba Medinipur, W.B, Pin-721401, India

article info abstract

Article history: An ‘‘economic production lot size’’ (EPLS) model for an item with imperfect quality
Received 20 October 2009 is developed by considering random machine failure. Breakdown of the manufacturing
Received in revised form 10 February 2011 machines is taken into account by considering its failure rate to be random (continuous).
Accepted 10 February 2011
The production rate is treated as a decision variable. It is assumed that some defective
units are produced during the production process. Machine breakdown resulting in idle
Keywords:
time of the respective machine which leads to additional cost for loss of manpower is
Production lot-sizing
Imperfect quality
taken into account. It is assumed that the production of the imperfect quality units is a
Random machine failure random variable and all these units are treated as scrap items that are completely wasted.
Maintenance time The models have been formulated as profit maximization problems in stochastic and
fuzzy-stochastic environments by considering some inventory parameters as imprecise in
nature. In a fuzzy-stochastic environment, using interval arithmetic technique, the interval
objective function has been transformed into an equivalent deterministic multi-objective
problem. Finally, multi-objective problem is solved by Global Criteria Method (GCM).
Stochastic and fuzzy-stochastic problems and their significant features are illustrated by
numerical examples. Using the result of the stochastic model, sensitivity of the nearer
optimal solution due to changes of some key parameters are analysed.
© 2011 Elsevier Ltd. All rights reserved.

1. Introduction

During the past few decades, attempt has been made to make inventory control models more realistic with real life
industrial problems. In most of the classical economic production quantity (EPQ) model, it is assumed that items produced
are of perfect quality and quality control of the product generally is not considered. However, in a production system, it is
quite natural that a machine cannot produce all items perfect during a whole production period.
Salameh and Jaber [1] extended the classical economic order quantity EOQ model by considering imperfect quality items
while using EOQ formulae. Later, Cardenas-Barron [2]corrected a mistake in the final formula of Salameh and Jaber’s model.
Goyal and Cardenas-Barron [3] then reconsidered the work of Salameh and Jaber and presented a practical approach for
determining the optimal lot size. Hayek and Salameh [4] derived an optimal operating policy for the finite production
model under the effect of reworking of imperfect quality items and assuming that all the defective items are repairable.
Chiu [5] examined an EPQ model with scrap items and the reworking of repairable items. Wee et al. [6] developed an optimal
inventory model for items with imperfect quality and shortages. Chung et al. [7] proposed a new inventory model with two

∗ Corresponding author. Tel.: +91 9434453186.

E-mail addresses: [email protected] (D. Das), [email protected] (A. Roy), [email protected] (S. Kar).

0898-1221/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.camwa.2011.02.015
 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400 2389

warehouses and imperfect quality. Jaber et al. [8] presented the concept of entropy cost to extended a new model under the
assumptions of perfect and imperfect quality. Jaber et al. [9] extended Salameh and Jaber [1] and assumed the percentage
of defective per lot reduces according to a learning curve.
Nowadays, with the advent of multinationals, there is a stiff competition in the market and the management implements
flexible manufacturing systems (FMS) for improving production efficiency. Volume flexibility that is capable of adjusting
the production rate with variability of demand in the market is one of the important components in FMS. Volume flexibility
helps to reduce production rate to avoid rapid accumulation of inventories. Obviously, the machine production rate is a
decision variable in the case of a FMS and then the unit production cost becomes a function of production rate. Khouja and
Mehrez [10] and Khouja [11] extended the EPLS model to an imperfect production process with a flexible production rate.
Moon et al. [12], Gallego [13], etc. extended the EPLS model with flexible production rate by considering constant demand.
Sana et al. [14] developed an economic manufacturing model in an imperfect production system where the defective items
are sold in a reduced price.
When some inventory parameters are fuzzy in nature, its objective function also becomes fuzzy. After the introduction
of fuzzy set theory in 1965 by Zadeh, extensive research work has been done on defuzzification of fuzzy numbers. Among
these techniques Centroid Method [15], Weighted Average Method [16], Graded Mean Value Method [17], Nearest Interval
Approximation Method [18], Graded Mean Integration Value [19], etc., have drawn more attention. All these techniques
replace the fuzzy parameters by their nearest crisp number/interval and the reduced crisp objective function is optimized.
In reality, a machine cannot work smoothly forever because its spare parts will breakdown sooner or later. It may become
out of order during its working time and there is a mean time between its failures/breakdowns. During a breakdown
period, demand in the system persists, but there is no production. When inventory level becomes less than the demand,
the management unit is rendered fully idle. In this paper we consider the idle time of the machine which leads to an
additional cost for the loss of man-hours. We take the time between successive breakdowns of the machines to be random
and the maintenance time is also considered to be random. Here, models have been formulated as profit maximization
problems in stochastic and fuzzy-stochastic environments. In a fuzzy-stochastic environment the model is transferred into
multi-objective problem and solved by Global Criteria Method (GCM). In order to illustrate the solution method numerical
examples are provided. Sensitivity of the decision variable and total expected profit is examined to check how far the output
of the model is affected by changes or errors in its input parameters.

2. Preliminaries

This section provides an introduction to fuzzy number and interval arithmetic.

A of real number R with membership function µ

Definition 1 (Fuzzy Number). A fuzzy subset  A : R → [0, 1] is called a fuzzy
number if
(a) 
A is normal, i.e. there exist an element x0 such that  µA (x0 ) = 1;
A is convex, i.e. µ
(b)  A (λ x1 + (1 − λ) x 2 ) ≥ µ A ( x 1 ) µA (x2 ) for all x1 , x2 ∈ R and λ ∈ [0, 1];
(c) µ

A is upper semi-continuous; and
(d) supp( A) is bounded, here supp( A) = {x ∈ R : µ A (x) > 0}.

Example 1 (Triangular Fuzzy Number). Triangular fuzzy number (TFN) A is an example of fuzzy number with the membership
function µ
A (x), a continuous mapping µ
A : R → [0, 1] and is defined by

for x < a1

 0
 x − a1

for a1 ≤ x < a2


a −a

µA (x) = a2 − x1
3
for a2 ≤ x < a3



 − a2
a

 3
0 for x ≥ a3 .

A is denoted by (a1 , a2 , a3 ), where a1 , a2 , a3 ∈ R.

Generally, 

2.1. α -cut of fuzzy number

The α -cut of a fuzzy number is a crisp set which is defined as [

A]α = {x ∈ R : µ
A (x) ≥ α}. According to the definition of
fuzzy number it is seen that α -cut is a non-empty bounded closed interval, it can be denoted by
A]α = [AL (α), AR (α)].
[
AL (α) and AR (α) are the lower and upper bounds of the closed interval, where
AL (α) = inf{x ∈ R : µA (x) ≥ α},
and
AR (α) = sup{x ∈ R : µ
A (x) ≥ α}.
2390 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400

2.2. Arithmetic of interval number

Throughout this section lower case letters denote real numbers and upper case letter denote closed intervals. The set of
all positive real numbers is denoted by R+ and R0 = R+ {0}. An order pair of brackets defines an interval A = [aL , aR ] =
{a : aL ≤ a ≤ aR , a ∈ R0 } where aL and aR are, respectively, left and right limits of A.

Definition 2. Let ∗ ∈ {+, −, ., /} be a binary operation on the set of positive real numbers. If A and B are closed intervals
then A ∗ B = {a ∗ b : a ∈ A, b ∈ B} defines a binary operation on the set of closed intervals. In the case of division, it is
assumed that 0 ̸∈ B.
The operations on intervals used here may be explicitly calculated from the above definition as
A + B = [aL , aR ] + [bL , bR ] = [aL + bL , aR + bR ],
A − B = [aL , aR ] − [bL , bR ] = [aL − bR , aR − bL ],
and
A.B = [aL , aR ].[bL , bR ] = [aL bL , aR bR ],
for all aL , aR , bL , bR in R0 ;
 
A [ aL , aR ] aL aR
= = , ,
B [bL , bR ] bR bL

where 0 ̸∈ B, 0 ≤ aL ≤ aR and 0 ≤ bL ≤ bR and

[kaL , kaR ],

for k ≥ 0
kA =
(kaR , kaL ), for k < 0,

where k is a real number.

2.3. Formulation of the multi-objective problem

We define a general non-linear objective function with some parameters as interval valued numbers is as
k n
γij
Maximize Z (x) =
∑ ∏
Ci xj (1)
i=1 j =1

i = 1, 2, . . . , k, j = 1, 2, . . . , n
where Ci = [CLi , CRi ].
Now, we exhibit the formulation of the original problem (1) as a multi-objective non-linear problem. Since the objective
function Z (x) involves some parameters represented by intervals, it is natural that the solution set of (1) should be defined
by preference relations between intervals [cf. [20,21] etc.]. Now, from Eq. (1), following interval arithmetic, the right and
left limits ZR (x), ZL (x) of the interval objective function Z (x) respectively may be elicited as
k n
− ∏ γij
ZR (x) = CRi xj (2)
i=1 j =1

k n
− ∏ γij
ZL (x) = CLi xj (3)
i=1 j =1
[ ]
1
ZC (x) = ZR (x) + ZL (x) . (4)
2
Thus the problem in Eq. (1) is transformed into
[ ]
Maximize ZL (x), ZC (x) . (5)

3. Global Criteria Method (GCM)

Global Criteria Method (GCM) is one of the useful technique for solving a multi-objective optimization problem. This
method generates a set of Pareto optimal solutions and use some additional criteria or rule to select one particular Pareto
 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400 2391

optimal solution as the solution of the multi-objective optimization problem. Global Criteria Method (GCM) needs initial
reference point in order to start the solution procedure. With a given reference point multi-objective multiple criteria
decision making problems can be solved by locating the alternative(s) or decision(s) closest to the reference (ideal) point.
Thus the problem becomes how to measure distance to the reference point. Let us consider a multi-objective constrained
optimization problem as:
 
Maximize f1 (x), f2 (x), . . . , fk (x)

subject to gj (x) ≤ 0, j = 1, 2, . . . , m
x = (x1 , x2 , . . . , xn )T , where some or all of the functional fi (x) and gj (x) are non-linear.

GCM measures this distance by using Minkowski’s Lp metric. The Lp metric defines distance between two points, f and
  1p
∗ p
i=1 fi (x) − fi , p ≥ 1.
∑k  
f ∗ (the reference point) in k-dimension space as: 

Hence, the Lp -problem is

  1p
k  p
fi (x) − f ∗ 
∑
Minimize 
i (6)
i=1

subject to gj (x) ≤ 0, j = 1, 2, . . . , m
x = (x1 , x2 , . . . , xn )T ∈ X .

From the definition of the ideal objective vector f ∗ , it is known that

fi (x) ≥ fi∗ , (i = 1, 2, . . . , k).

This is why no absolute values are needed if the global ideal objective vector is known. Unfortunately, because of
incommensurability among the objectives, it is impossible to directly use the above family of distance function. Thus, it
is necessary to normalize the distance family by using reference points to remove the effects of the incommensurability.
Yu and Zeleny [22] normalized the distance family of Eq. (6) by using reference points. The problem is then to solve the
following auxiliary problem:
  1p
k 
 fi (x)−fi∗ p
∑ 
Minimize  f∗  (7)
i
i =1

subject to gj (x) ≤ 0, j = 1, 2, . . . , m
x = (x1 , x2 , . . . , xn )T ∈ X .

The value chosen for p reflects the way of achieving a compromise in minimizing the weighted sum in deviations of criteria
from their respective ideal solutions. Boychuk and Ovchinnikov [23] suggested p = 1 which is the most credible situation.
Since Lp problem (7) is an increasing function of the corresponding problem without the exponent 1p an usual value of p is 2.
When p = 2, Eq. (7) becomes a convex programming problem. This convex programming problem can be solved by using any
method of finding a local extrema such as pattern search techniques, gradient method and other non-linear programming
techniques. The solution of Lp problem (7) (where 1 ≤ p < ∞) is Pareto optimal.

4. Assumptions and notations

The Mathematical model in this paper is developed on the basis of following assumptions and notations:
Assumptions:
1. Demand rate is constant.
2. Production rate is considered as a decision variable.
3. Machine breakdowns occur randomly during a production period.
4. Idle time is taken into account.
5. Maintenance time of the machine is also random.
6. Shortages are allowed during the maintenance time.
7. Unit production cost is considered as a function of production rate.
8. Proportion of defective items is random.
9. All defective items are assumed to be scrap items.
2392 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400

Notations:
1. q(u) = On hand inventory of the perfect item at time u, u ≥ 0.
2. t = Time between two consecutive breakdowns (a random variable).
3. τ = Maintenance time (a random variable).
4. P = Production rate (a decision variable).
5. λ = Mean time between two consecutive breakdowns.
6.
f (t ) = Probability density function (p.d.f) of t
1 −t
= e λ , t ≥ 0.
λ
7. µ = Mean repair time of the machine.
8.
φ(τ ) = p.d.f of τ
1 −τ
= e µ , τ ≥ 0.
µ
9. x = Defective rate (a random variable).
10. cm = Maintenance cost per unit time.
11. Id = Idle cost of the machine per unit time.
12. Cs = Shortage cost per unit quantity per unit time.
13. C1 = Holding cost per unit quantity per unit time.
14.
ψ(x) = p.d.f of x.
1
= , a ≤ x ≤ b.
( b − a)
15. D = Demand rate.
16. s = Selling price per unit.
17. η(P ) = Unit production cost = r + g
P α1
+ v P α2 , where r , g , v, α1 and α2 are all positive constants.
18. TP = Total profit.
19. ETP = Expected total profit.

5. Mathematical model

In the proposed model, we consider a manufacturing system for an item with some defective units. In this inventory
model we consider defective rate, breakdown time of machine and machine maintenance time as random variables. The
demand rate is uniform. The production cost per unit item η(P ) = r + P α1 + v P α2 , where r , g , v, α1 and α2 are all positive
g

constants. This cost is based on the following factors:

(i) The material cost r per unit item is fixed.
(ii) As the production rate increases, some costs like labour and energy costs are equally distributed over a large number
g
of units. Hence per unit production cost P α1 decreases as the production rate increases.
α2
(iii) The third term v P associated with tool/die cost is proportional to the production rate.

5.1. Formulation

The governing differential equation of the system is given by,

dq(t )
= (1 − x)P − D, t ≥ 0 and q(0) = 0. (8)
dt
The solution of the differential equation (8) is q(t ) = {(1 − x)P − D}t , t ≥ 0.
Here we also conclude that the idle time of the machine due to break down
q(t )

0,
 if ≥τ
td = D
τ − q(t ) ,

if
q(t )
< τ.
D D
 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400 2393

Expected shortage cost is given by,

∫ b ∫ ∫  
∞ ∞
q(t )
 
ESC = Cs D τ− φ(τ )dτ f (t )dt ψ(x)dx
q(t ) D
a 0 D
∫ b ∫   
∞ ∫ ∞
q(t )
 
= Cs D τ− φ(τ )dτ f (t )dt ψ(x)dx
{(1−x)P −D}t D
a 0 D
 
D µ2 λ(P − D) + Dµ  λ(P − D) + Dµ
  
− b .
 
= Cs D log  − a − log  (9)
λP (b − a) λP λP
Expected cost per break down during idle time is given by,
∫ b ∫ ∫  
∞ ∞
q(t )
 
EIC = Id τ− φ(τ )dτ f (t )dt ψ(x)dx
q(t ) D
a 0 D
∫ b ∫ ∫  
∞ ∞
q(t )
 
= Id τ− φ(τ )dτ f (t )dt ψ(x)dx
{(1−x)P −D}t D
a 0 D
 
D µ2  λ(P − D) + Dµ  λ(P − D) + Dµ
  
− b .
 
= Id log 
 − a − log 
  (10)
λP (b − a) λP λP
Expected selling price is given by,
∫ b ∫
∞

ESP = s (1 − x)P tf (t )dt ψ(x)dx
a 0
 
(a + b)
= sP λ 1 − . (11)
2

Expected holding cost is given by,

∫ b  ∫  ∫ b ∫ 
C1 ∞
C1 {(1 − x)P − D}2 ∞
EHC = (1 − x)P − D tf (t )dt ψ(x)dx + tf (t )dt ψ(x)dx
2 a 0 2 a D 0
 
λC1 (a + b) λC1
[ ]
= (P − D) − P + (P − D − aP )3 − (P − D − bP )3 . (12)
2 2 6PD(b − a)

Expected maintenance cost is given by,

∫ b ∫ ∞ ∫ ∞  
EMC = cm τ φ(τ )dτ f (t )dt ψ(x)dx
a 0 0

= cm µ. (13)
Total expected production cost is given by,
∫ ∞
EPC = η(P )P tf (t )dt
0

= η(P )P λ. (14)
So, expected total profit per breakdown is,
ETP = ESP − EPC − EHC − ESC − EIC − EMC
     
(a + b) λC1 (a + b) g α2
= sP λ 1 − − (P − D) − P − r+ + vP Pλ
2 2 2 P α1

λ C1
[ ]
− (P − D − aP ) − (P − D − bP )
3 3
6PD(b − a)
  
D µ2  λ(P − D) + Dµ  λ(P − D) + Dµ
 
− (Cs D + Id ) − b − cm µ.
 
log 
 − a − log 
  (15)
λP (b − a) λP λP
2394 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400

239.8
239.6
239.4

Expected total profit

239.2
239
238.8
238.6
238.4
238.2
75 80 85 90 95
Production rate

Fig. 1. Expected total profit versus production rate.

5.2. Stochastic model

Hence the above problem can be defined as,

max ETP (P ) (16)
P > 0.
Taking P as the decision variable, the necessary condition for the maximum of the expected profit is
   
d(ETP ) ( a + b)
 
g α2 α 2 −1
= sλ 1 − − r+ + vP λ − P λ −g α1 P −α1 −1
+ vα2 P
dP 2 P α1
 
λC1 (a + b) λ C1
[ ]
− 1− + ( P − D − aP )3
− ( P − D − bP )3
2 2 6P 2 D(b − a)

λC1
[ ]
− 3(P − D − aP ) .(1 − a) − 3(P − D − bP ) .(1 − b)
2 2
6PD(b − a)
 
 λ(P − D) + DµD µ2  λ(P − D) + Dµ
  
+ (Cs D + Id ) 2
 
log  − a − log  − b
λP (b − a) λP λP
 
D2 µ2 (λ − µ) 1 1
− (Cs D + Id ) 2 3 −
λ P (b − a)  λ(P −D)+Dµ − a  λ(P −D)+Dµ − b
 
λP λP

= 0. (17)
The sufficient condition for the maximum of ETP is

d2 (ETP )
< 0. (18)
dP 2
The expression for ETP being highly complicated, it is not possible to examine its concavity analytically. Here we generate
a graph of the function for some parametric values. The graph is shown in Fig. 1, which have a clear maximum point. The
Eq. (17) being highly non-linear is solved by a numerical method by using Generalized Reduced Gradient Method (Global
Criteria Method) for given parametric values. If the solution P ∗ of this equation satisfies the condition (18) then P ∗ becomes
the optimal solution and the corresponding value of ETP ∗ is the optimal expected total profit (see Tables 3–5).

5.3. Fuzzy-stochastic model

In reality holding cost and shortage cost may be imprecise i.e. vaguely defined in some situations. So we take C1 and Cs as
fuzzy numbers, i.e. as C1 and C
s . Also the unit cost parameters, r, g, v may be consider as fuzzy numbers, i.e. as  r ,
g ,
v . In this
model we consider C1 = (C11 , C12 , C13 ), C
s = (Cs1 , Cs2 , Cs3 ),
r = (r1 , r2 , r3 ),
g = (g1 , g2 , g3 ), 
v = (v1 , v2 , v3 ). Using interval
for the fuzzy numbers mentioned in Section 2, the multi-objective inventory problem is given by
[ ]
max ETP αL , ETP αC (19)
 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400 2395

where
[ ]
1
ETP αC = ETP αL + ETP αR , α ∈ [0, 1].
2
   
( a + b) λ ( a + b)

α
ETP L = sP λ 1 − − C13 − α(C13 − C12 ) (P − D) − P
2 2 2
 
{g3 − α(g3 − g2 )}
  
α2
− r3 − α(r3 − r2 ) + + v3 − α(v3 − v2 ) P Pλ
P α1
λ
 [ ]
− C13 − α(C13 − C12 )
(P − D − aP )3 − (P − D − bP )3
6PD(b − a)
 
D µ2  λ(P − D) + Dµ
[  ] 
− Cs3 − α(Cs3 − Cs2 ) D + Id

log  − a
λP (b − a) λP

 λ(P − D) + Dµ

− b − cm µ

− log 
λP
and
   
( a + b) λ (a + b)

α
ETP R = sP λ 1 − − C11 + α(C12 − C11 ) (P − D) − P
2 2 2
  
{g1 + α(g2 − g1 )}
 
− r1 + α(r2 − r1 ) + α
+ v1 + α(v2 − v1 ) P α2 P λ
P 1

λ
 [ ]
− C11 + α(C12 − C11 ) (P − D − aP )3 − (P − D − bP )3
6PD(b − a)
 
Dµ2  λ(P − D) + Dµ
[  ] 
− Cs1 + α(Cs2 − Cs1 ) D + Id

log 
 − a
λP (b − a) λP

 λ(P − D) + Dµ

− b − cm µ.

− log 
λP

6. Numerical illustration

6.1. Stochastic model

The problem (16) is now solved with the help of a computer program using the following parametric value:
s = $9, C1 = $1.0, a = 0, b = 0.1, D = 20, λ = 1.2, µ = 0.15, cm = $25, Cs = $1.9, Id = $5, α1 = 1.34, α2 =
0.9, v = 0.06, g = 200, r = 2 in appropriate units.
The optimal value of P along with maximum expected total profit are calculated and the corresponding values are
P ∗ = 85.248 and ETP ∗ = 239.565.

6.2. Sensitivity analysis

For the given numerical example mentioned in Section 6.1, sensitivity analysis is performed to study the effect of changes
of different parameters like production cost, mean life time between two consecutive breakdowns, mean repair time on
production rate and maximum expected profit of the system.

6.2.1. Sensitivity of expected total profit w.r.t. α1 and α2 of volume flexibility

Sensitivity analysis is performed for the maximum expected total profit with respect to the different values of the
parameters α1 and α2 and results are plotted in Fig. 2 (three dimension linear graph using STATISTICA).
In is observed that for fixed value of α2 , expected total profit increases with α1 , and for fixed of α1 , it decreases as α2
increases.

6.2.2. Sensitivity of unit production cost w.r.t. α1 and α2

Sensitivity analysis is performed for the unit production cost with respect to the different values of the parameters α1
and α2 and results are plotted in Fig. 3 (three dimension linear graph using STATISTICA).
2396 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400

Fig. 2. Expected total profit versus α1 and α2 .

6.2
6.0
5.8
5.6
5.4
n(P)

5.2
5.0
4.8
4.6
4.4
4.2
8
0.9 1.
6 42
0.9 1.
4 40
0.9 1.
2 38
0.9 1.
0 36
0.9 1.
34
8
0.8 1.
32
6
0.8 1.
30
4 1.
0.8 28
2 1.
0.8 26

Fig. 3. Unit production cost versus α1 and α2 .

It is observed that for fixed value of α1 , as α2 increases, unit production cost increases and for fixed of α2 , it decreases
with increase of α1 .

6.2.3. Sensitivity of unit production cost w.r.t. production rate

Sensitivity analysis is performed for the unit production cost with respect to the different values of the production rate,
P and results are plotted in Fig. 4.
From the above figure, we see that unit cost is minimum when rate of production, P = 44.658 (cf. Fig. 4). But, when P
is considered as a decision variable, maximum expected total profit is obtained for higher production rate (85.248). Hence,
one more common belief that lowest unit cost gives maximum profit is not true in the case of present inventory model.
 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400 2397

5.09

5.085

Unit production cost

5.08

5.075

5.07

5.065

5.06
35 40 45 50 55
Production rate

Fig. 4. Unit production cost versus production rate.

Table 1
Sensitivity analysis of ETP and P w.r.t. λ.
λ ETP P

1.00 198.921 85.267

1.05 209.083 85.262
1.10 219.245 85.256
1.15 229.405 85.252
1.20 239.565 85.248
1.25 249.723 85.245
1.30 259.880 85.242
1.35 270.038 85.239

Table 2
Sensitivity analysis of ETP and P w.r.t. µ.
µ ETP P

0.10 240.681 85.228

0.15 239.565 85.248
0.20 238.123 85.282
0.25 236.231 85.325
0.30 235.091 85.373
0.35 233.505 85.429

6.2.4. Sensitivity analysis of expected total profit and production rate w.r.t. λ
Sensitivity is performed for expected total profit and production rate with respect to the different values of the parameter
λ and results are displayed in Table 1.
As expected, with the increase of λ i.e. as the mean time between breakdown increases, the profit increases. Since it does
not have any direct relation with production rate, optimum production rate changes slightly. Production rate remains nearly
constant with increase of λ.

6.2.5. Sensitivity of expected total profit and production rate w.r.t. µ

Sensitivity analysis is performed for expected total profit and production rate with respect to the different values of the
parameter µ and results are displayed in Table 2.
In this case also, as expected, as µ increases i.e. as mean time for maintenance increases, loss of production is more and
hence the expected profit decreases. As the case in λ (cf. Table 1) optimum production rate slightly increases with µ to adjust
the optimum profit.

6.3. Fuzzy-stochastic model

Problem (19) is now solved with the help of a computer program using the above parametric value except fuzzy
parameters.
Experiment No.-1:
C1 = ($0.90, $1, $1.1), C
s = ($1.7, $1.9, $2.1),
r = (1.8, 2, 2.2),
g = (180, 200, 220),
v = (0.05, 0.06, 0.07).
Experiment No.-2: C1 = ($0.90, $1, $1.2), C
s = ($1.7, $1.9, $2.2), 
r = (1.8, 2, 2.3),
g = (180, 200, 225),
v =
(0.05, 0.06, 0.08).
Experiment No.-3:
C1 = ($0.80, $1, $1.1), C
s = ($1.6, $1.9, $2.1), 
r = (1.7, 2, 2.2),
g = (175, 200, 220),
v = (0.04, 0.06, 0.07).
2398 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400

Table 3
Optimal expected profit for different values of α .
α 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

ETP L 166.03 172.46 179.06 185.86 192.85 200.05 207.47 215.12 223.01 231.15 239.56
ETP C 239.56 239.56 239.56 239.56 239.56 239.56 239.56 239.56 239.56 239.56 239.56

Table 4
Optimal expected profit for different values of α .
α 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

ETP L 121.62 130.77 140.38 150.48 161.12 172.35 184.23 196.81 210.18 224.40 239.56
ETP C 206.74 209.79 212.88 216.03 219.23 222.48 225.78 229.14 232.56 236.03 239.56

Table 5
Optimal expected profit for different values of α .
α 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

ETP L 166.03 172.46 179.06 185.86 192.85 200.05 207.47 215.12 223.01 231.15 239.56
ETP C 278.56 274.33 270.18 266.11 262.11 258.19 254.33 250.54 246.82 243.16 239.56

7. Discussion

In this paper stochastic and fuzzy-stochastic models are illustrated with numerical examples by considering imprecise
inventory parameters. Here, exponential distribution with parameters λ and µ for time between two consecutive
breakdowns and maintenance time respectively and uniform distribution for defective rate have been assumed. The
following features are observed from the optimum solution in the numerical example:
1. Production rate of the machine increases with the increase in its mean duration of a breakdown.
2. Production rate of the machine increases in the selling price of the item produce by the machine.
3. Production rate of the machine increases as the idle-time cost of the management decreases.
4. Production rate of the machine increases as the mean time between its successive breakdowns increases.

8. Conclusion

In practical production inventory management, a machine cannot work for ever, due to process deterioration or some
other factors, the generation of imperfect items is inevitable. The idea of this paper is important today as more volume
flexible production systems are being introduced to cope with the fluctuations in the market demands. Both the stochastic
breakdown time of the machine and the stochastic defective rate are considered. As far as our knowledge goes no such
stochastic production model has been considered in inventory literature. A numerical example is provided to demonstrate
its practical usage. For further research, one interesting direction among others to investigate economic lot-sizing model
on an imperfect quality with rework process. The model can also be formulated and solved considering other probability
distributions in stochastic or hybrid environments.

β
Appendix. Calculation of E (TP ) considering machine failure with a Weibull distribution i.e. f (t ) = αβt β−1 e−αt ,
where α > 0, β > 0

Expected shortage cost is given by,

∫ b ∫ ∫  
∞ ∞
q(t )
 
ESC = Cs D τ− φ(τ )dτ f (t )dt ψ(x)dx
q(t ) D
a 0 D
∫ b ∫ ∫    
∞ ∞
{(1 − x)P − D}t β−1 −α t β
= Cs D τ− φ(τ )dτ αβ t e dt ψ(x)dx
{(1−x)P −D}t D
a 0 D

∫ b ∫ ∞
  
(1−x)P −D
β−1 −{ t +α t β }
= Cs D µαβ t e µD dt ψ(x)dx. (20)
a 0

Expected cost per break down during idle time is given by,
∫ b ∫ ∫  
∞ ∞
q(t )
 
EIC = Id τ− φ(τ )dτ f (t )dt ψ(x)dx
q(t ) D
a 0 D
 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400 2399

∫ b ∫     
∞ ∫ ∞
{(1 − x)P − D}t β
= Id τ− φ(τ )dτ αβ t β−1 e−αt dt ψ(x)dx
{(1−x)P −D}t D
a 0 D

∫ b ∫ ∞
  
(1−x)P −D
β−1 −{ t +α t β }
= Id µαβ t e µD dt ψ(x)dx. (21)
a 0

Expected selling price is given by,

∫ b ∫
∞

ESP = s (1 − x)P tf (t )dt ψ(x)dx
a 0
 
sP Γ 1 
β a+b
= 1
1− . (22)
βα β 2

Expected holding cost is given by,

∫ b  ∫  ∫ b ∫ 
C1 ∞
C1 {(1 − x)P − D}2 ∞
EHC = (1 − x)P − D tf (t )dt ψ(x)dx + tf (t )dt ψ(x)dx
2 a 0 2 a D 0
   
C1 Γ 1
C1 Γ 1
 
β P (a + b) β P 2 (a2 + ab + b2 )
= 1
(P − D) − + 1
(P − D) − P (P − D)(a + b) +
2
. (23)
2βα β 2 2Dβα β 3

Expected maintenance cost is given by,

∫ b ∫ ∞ ∫ ∞  
EMC = cm τ φ(τ )dτ f (t )dt ψ(x)dx
a 0 0

= cm µ. (24)

Total expected production cost is given by,

∫ ∞
EPC = η(P )P tf (t )dt
0
 
η(P )P Γ 1
β
= 1
. (25)
βα β
So, expected total profit per break down is,

E (TP ) = ESP − EPC − EHC − ESC − EIC − EMC

     
sP Γ 1
η(P )P Γ 1
C1 Γ 1

β a+b β β
= 1
1− − 1
− 1
(P − D)
βα β 2 βα β 2βα β
 
C1 Γ 1

P (a + b) β
− − 1
(P − D)2 − P (P − D)(a + b)
2 2Dβα β
 ∫ b   
P (a + ab + b2 )
2 2 ∞
∫
(1−x)P −D
β−1 −{ t +α t β }
+ − Cs D µαβ t e µD dt ψ(x)dx
3 a 0

∫ b ∫ ∞
  
(1−x)P −D
β−1 −{ t +α t β }
− Id µαβ t e µD dt ψ(x)dx − cm µ. (26)
a 0

β
Since the derivation of ESC and EIC are very complicated by taking Weibull distribution of f (t ) i.e. f (t ) = αβ t β−1 e−α t ,
where α > 0, β > 0, so the numerical results are not given. In our model we consider the particular case of Weibull
t
distribution by taking α = λ1 and β = 1 i.e. f (t ) = λ1 e− λ .
2400 D. Das et al. / Computers and Mathematics with Applications 61 (2011) 2388–2400

References

[1] M.K. Salameh, M.Y. Jaber, Economic production quantity model for items with imperfect quality, International Journal of Production Economics 64
(2000) 59–64.
[2] L.E. Cardenas-Barron, Observation on: economic production quantity model with imperfect quality-a practical approach, International Journal of
Production Economics 67 (2000) 201–212.
[3] S.K. Goyal, L.E. Cardenas-Barron, Note on: economic production quantity model for items with imperfect quality a practical approach, International
Journal of Production Economics 77 (2002) 77–85.
[4] P.A. Hayek, M.K. Salameh, Production lotsizing with the reworking of imperfect quality items produced, Production Planning and Control 12 (2001)
584–590.
[5] Y.P. Chiu, Determining the optimal lot-size for the finite production model with random defective rate, the rework process and backlogging,
Engineering Optimization 35 (2003) 427–437.
[6] H.M. Wee, J. Yu, M.C. Chen, Optimum inventory model for items with imperfect quality and shortage backordering, Omega 35 (2007) 7–11.
[7] K.J. Chung, C.C. Her, S.D. Lin, A two-warehouse inventory model with imperfect quality production processes, Computers & Industrial Engineering 56
(2009) 193–197.
[8] M.Y. Jaber, M. Bonney, I. Moualek, An economic order quantity model for an imperfect production process with entropy cost, International Journal of
Production Economics 118 (2009) 19–25.
[9] M.Y. Jaber, S.K. Goyal, M. Imran, Economic production quantity model for items with imperfect quality subject to learning effects, International Journal
of Production Economics 115 (2008) 143–150.
[10] M. Khouja, A. Mehrez, Economic production lot size model with variable production rate and imperfect quality, Journal of the Operational Research
Society 45 (1994) 1405–1417.
[11] M. Khouja, Economic production lot size model under volume flexibility, Computers and Operations Research 22 (1995) 515–525.
[12] I. Moon, G. Gallego, D. Simchilevi, Controllable production rates in a family production context, International Journal of Production Research 29 (1991)
2459–2470.
[13] G. Gallego, Reduced production rates in the economic lot scheduling problem, International Journal of Production Research 31 (1993) 1035–1046.
[14] S. Sana, S.K. Goyal, K.S. Chaudhuri, On a volume flexible inventory model for items with an imperfect production system, International Journal of
Operational Research 2 (1) (2007) 64–80.
[15] C. Lee, Fuzzy logic in control systems: fuzzy logic controller, Parts I and Parts II, IEEE Transaction Systems, Man & Cybernetics 20 (1990) 404–435.
[16] H. Heiiendoorn, C. Thomas, Defuzzification in fuzzy controller, Intelligent and Fuzzy Systems 1 (1993) 109–123.
[17] T.S. Liou, M.J.J. Wang, Ranking fuzzy numbers with integral values, Fuzzy Sets and Systems 50 (1992) 247–255.
[18] P. Grzegorzewski, Nearest interval approximation of a fuzzy number, Fuzzy Sets and Systems 130 (2002) 321–330.
[19] S.H. Chen, C.H. Hsieh, Graded mean integration representation of generalized fuzzy numbers, Journal of Chinese Fuzzy Systems 5 (2) (1999) 1–7.
[20] J.W. Chinnek, K. Ramadan, Linear programming with interval coefficients, Journal of the Operational Research Society 51 (2000) 209–220.
[21] H. Ishibuchi, H. Tanaka, Multi-objective programming in optimization of the interval objective function, European Journal of Operational Research 48
(1990) 219–225.
[22] P.L. Yu, M. Zeleny, The set of all non-dominated solutions in linear cases and a multicriteria simplex method, Journal of Mathematical Analysis and
Applications 49 (1975) 430–448.
[23] L.M. Boychuk, V.O. Ovchinnikov, Principle methods for solution of multicriterion optimization problem (survey), Soviet Automatic Control 6 (1973)
1–4.