Escolar Documentos
Profissional Documentos
Cultura Documentos
L. Dai
Industrial Systems Engineering, University of Regina, Regina, Saskatchewan, S4S 0A2, Canada.
L. Fan
Computer Science, University of Regina, Regina, Saskatchewan, S4S 0A2, Canada.
L. Sun
Industrial Systems Engineering, University of Regina, Regina, Saskatchewan, S4S 0A2, Canada.
Uncertainties and imprecise information regarding customer demands, production, inventory, and
MRP are very common in performing aggregate production-planning (APP) for the manufacturing
in real world. This study presents a fuzzy linear programming approach for managing the
uncertainties and imprecise information involved in industrial APP applications. Detailed
discussions are given to the establishment of the Fuzzy Linear Programming approach with
converting the fuzzy constraints of uncertain and imprecise items into deterministic equivalents. A
mathematical model is developed for APP practice with the Fuzzy Linear Programming approach.
For numerically performing an aggregate production planning with the Fuzzy Linear
Programming developed, a computer simulation for an actual aggregate production-planning is
presented. It is demonstrated in the study, the employment of the Fuzzy Linear Programming
provides a great advantage in APP of manufacturing, if the parameters of the stochastic factors
involved in the production planning are neither definitely reliable nor precise. The present study
shows that the interrelated effects of the customer service level and facility capacity on the
effectiveness and efficiency of aggregate production-planning is significant and should be taken
into account in performing an aggregate production-planning
1. Introduction
One of the major decision making problems facing operations managers, or decision-makers in the
manufacturing industry is the aggregate production-planning (APP) problem. The research in APP area
mainly dealt with workforce size, inventory levels, and production rates. The literatures in this field do
offer a number of quantitative methods that can be used to perform aggregate production plans for a
firm (Tang et al., 2000, Hsieh and Wu, 2000, Silva Filho, 1999, Wang and Fang, 2001, Kaufmann and
Gupta, 1998). However, despite increasing of analytical tools and the substantial economic benefits to
be derived from using them, reports of successful application remain scarce. In reality, researchers
often ignore two factors: the customer service level and the facility capacity. In fact, these two factors
do affect managers’ decisions when they develop aggregate production planning. An extensive search
of the literature reveals that no previous investigation has systematically taken these two factors into
account with the purpose of improving total system costs, using a linear programming approach.
where r ∈ [0,1] is a parameter of variation. Constraint (5) is equivalent to Chanas’ (Chanas, 1983) and
Carlsson’s (Carlsson and Pekka, 1986) parametric programming formulation. Thus, fuzzy linear
constraints in equation (2) are transformed into conventional linear ones.
(b) Convert the equality constraints to Chanas’ formulation (Chanas, 1983)
The fuzzy equality constraint (1) can be transformed into the following two constraints using the
method of parametric programming with upper bounded variables.
Journal of Integrated Design and Process Science SEPTEMBER 2002, Vol. 7, No. 4, pp. 82
2
( Ax) i = bi + bi r − s i (6)
2 1
si ≤ bi − bi + (bi + bi ) r (7)
where fuzzy resource bi is a fuzzy number having the trapezoidal form as shown in Figure 2. This
transformation makes it possible to generate a set of solutions incorporating the whole range of
~
possible values ( bi 1 − b i r , bi 2 + bi r ) for fuzzy parameter bi .
µ b~ j
1.0
0
bj bj + bj ( Ax) j
µ ~
A
1.0
0
a1 − a
1
a1 a2 a2 + a
⎧1, T *
if C x ( r ) ≤ Z
1
⎪
⎪ C x(r ) − Z (8)
T
µ ~ ( x ( r )) = ⎨1 −
* 1 T * 2
2 1
, if Z ≤ C x ( r ) ≤ Z
G
⎪ Z −Z
⎪0, T *
if C x ( r ) ≥ Z
1
⎩
Equation (8) must be greater than the satisfaction level λ = 1 − r , therefore, it can be expressed as:
T 1 2
C x ≤ Z + (Z − Z )r
1
(9)
(b) Reformulate the FLP
Equation (9) is added to the transformed constraints set, the problem becomes to minimize r or
maximize λ and subject to the new constraints set of equations (6) and (7). The model can then be
defined as:
Min r
Subject to: C T x ≤ Z 1 + ( Z 2 − Z 1 )r
2
( Ax )i + si = bi + bi r
si ≤ bi − bi + (b i + bi ) r ,
2 1
for i = 1, 2, …, p
( Ax ) j ≤ b j + b j r , for j = p + 1 , …, q
( Ax ) m ≤ b m , for m = q + 1, …, s
x, s i , r ≥ 0
or,
Max λ
Subject to: C x ≤ Z + ( Z − Z )(1 − λ )
T 1 2 1
( Ax )i + si = bi + bi (1 − λ )
2
s i ≤ b i − b i + (b i + b i )(1 − λ ) ,
2 1
for i = 1, 2, …, p
( Ax) j ≤ b j + b j (1 − λ ), for j = p + 1 , …, q
( Ax ) m ≤ b m , for m = q + 1, …, s
x, s i , r ≥ 0
(c) Find the optimal solution (Bellman and Zadeh, 1970)
Journal of Integrated Design and Process Science SEPTEMBER 2002, Vol. 7, No. 4, pp. 84
Given that the fuzzy objective and fuzzy constraints are fuzzy sets µ G~ and µ C~ respectively, a set of
decision space (possible alternatives) can be formed by the intersection of µ G~ and µ C~ . If one wants to
~
find the optimal solution, the best x ∈ X = { x Ax ≤ b , x ≥ 0} must be found to simultaneously satisfy both
the objective and the corresponding constraints. A formula of maximizing decision can be defined as:
Max µ D~ ( x ) = Max [ µ G~ ( x ) ∧ µ C~ ( x ) ].
The decision maker may either terminate the decision process by accepting the production fuzzy
constraints or continue the process to modify (a) the goal range, (b) the membership function of the
goal and/or the parameters in the constraints, or (c) the model formulation.
Subject to:
Wt = Wt-1+Ht – Lt t=1, 2, …, T (11)
H t Lt = 0 , t = 1, 2, …, T (15)
bPot ≤ γ t M t t = 1, 2, …, T (17)
t = 1, 2, …, T (19)
Prt + Pot + I t −1 + S t ≥ Dt min
t = 1, 2, …, T (20)
Prt , Pot , Wt , I t , S t , H t , Lt ≥ 0
t = 1, 2, …, T (21)
The proposed fuzzy linear programming model for the single objective aggregate production-
planning problem differs from the traditional linear programming model in having a fuzzy goal, a fuzzy
workforce level, (E.q. 12), allowed in each period, and fuzzy demands, (E.q. 19). Where ~ denotes the
fuzzified version of the variable. Equation (20) implies that the total available product quantity should
be equal to or greater than the minimum demand, in period t. This will avoid the penalty of backlogs.
Journal of Integrated Design and Process Science SEPTEMBER 2002, Vol. 7, No. 4, pp. 86
(4) Overtime production is limited to no more than 30% of regular time production. The overtime
charge is based on $15 per worker per hour.
(5) The initial inventory ( I 0 ) is 0, the inventory carrying cost is $2 per unit per period. The
subcontracting cost is $67 per unit.
(6) Regular time machining capacity Mt= 800, 700, 820, 650, 750, and 720 machine- hours; fraction
of regular machine (facility) capacity available for use in overtime γ t = 0.5, 0.6, 0.5, 0.6, 0.4, and
0.4, in periods 1 to 6, respectively. For every period, the lower bound on the utilization of
machine capacity Mtmin is 400 machine-hours. And machine time for the product is 2.5 machine-
hours per unit.
The numerical solution procedure is as follows:
or
~
M i n Z= ∑6 {20 Prt + 65 Pot + 64Wt + +2 I t + 67 St + 30 H t + 40 Lt } (22)
t =1
Subject to:
W1 = W0 + H 1 − L1 (23)
W2 = W1 + H 2 − L2 (24)
W3 = W2 + H 3 − L3 (25)
W4 = W3 + H 4 − L4 (26)
W 5 = W 4 + H 5 − L5 (27)
W6 = W5 + H 6 − L6 (28)
~
W1 ≤ 9 0 (29)
~
W2 ≤ 10 0 (30)
~
W3 ≤ 12 0 (31)
~
W4 ≤ 12 0 (32)
~
W5 ≤ 12 0 (33)
~
W6 ≤ 12 0 (34)
3 Pr 1 ≤ 8W1 (35)
3 Pr 2 ≤ 8W2 (36)
3 Pr 3 ≤ 8W3 (37)
3 Pr 4 ≤ 8W4 (38)
3 Po1 ≤ 2.4W1 (39)
3 Po 2 ≤ 2.4W2 (40)
3 Po 3 ≤ 2.4W3 (41)
3 Po 4 ≤ 2.4W4 (42)
3 Po 5 ≤ 2.4W5 (43)
Journal of Integrated Design and Process Science SEPTEMBER 2002, Vol. 7, No. 4, pp. 88
⎧1, if W1 ≤ 90
⎪
⎪ W1 − 90
µ W~ = ⎨1 − , if 90 ≤ W1 ≤ 100
1 mzs
⎪ 10
⎪⎩0 , if W1 ≥ 100
W1 ≤ 90 + 10 r (71)
This equation demonstrates that a decision maker has the highest degree of satisfaction λ = 1 − r = 1
if the workforce level in period 1 is equal to or below 90. Equations (30) – (34) are transformed (from
FLP forms to conventional LP forms) as follows:
W2 ≤ 100 + 15 r (72)
W3 ≤ 120 + 30 r (73)
W4 ≤ 120 + 30 r (74)
W5 ≤ 120 + 30 r (75)
W6 ≤ 120 + 30 r (76)
(b) Change equality constraints to Chanas’ formulation
According to the discussion about equality constraints in (Chanas, 1983), the equations (45) – (50)
can be converted to the following forms:
µW~
1 max
1.0
0 100 W1max
90
µ D~
1
1.0
0
250 270 D1
µ D~
3
1.0
0
430 450 500 D3
Journal of Integrated Design and Process Science SEPTEMBER 2002, Vol. 7, No. 4, pp. 90
It should be noted that the decision maker does not have a precise idea about the total system costs
before an accurate planning is performed. The decision maker may also not be certain for a specific
production plan with a hundred percent satisfaction. Under this situation, the additional information
obtained from the eleven sets of feasible production plan will assist the decision maker to choose a
fuzzy goal range. The extreme function values are the costs of $83168 and $93186, from which the
decision-maker can choose any possible goal range. The decision maker looks at the numbers
associated with a set of production plans, then chooses the goal range in [$83000, $90000], which
means that Z 1 = $83000 is the goal with the tolerance of Z 2 − Z 1 = $7000.
(e) Reformulate and solve the FLP
The membership function of the fuzzy goal is:
⎧1, C T x* ( r ) ≤ 83000
⎪
⎪ C x ( r ) − 83000
T *
µ G~ ( x* ( r )) = ⎨1 − , 83000 ≤ C T x* ( r ) ≤ 90000
⎪ 90000 − 83000
⎪0 , C T x* ( r ) ≥ 9000
⎩
Since the grade of satisfaction must be greater than λ = 1 − r , the equation above can be expressed
as:
C T x* ( r ) ≤ 83000 + 7000 r (89)
Equation (89) is added into the transformed constraint set. Solving the LP formulation for the APP
problem then becomes to minimize r or equivalently maximize λ subject to the new constraint set.
The corresponding aggregate production plan with LP is performed and the results are tabulated in
Table 4. As can be seen from the table 4, the optimal total system cost obtained by solving the LP
problem is $87114 and the r value is 0.5933, or λ = 1- r = 0.4067.
The fuzzy decision thus made is based on the intersection of the fuzzy goal and fuzzy constraints.
This decision is depicted in Figure 4. The decision making process can be terminated if the decision
maker accepts the compromise production plan. Otherwise, the decision maker can modify the goal
range, the membership functions related to the goal and/or parameters in constraints, or the model
itself. For example, he/she can choose the goal range in [$83000, $88000], and with the same
procedure, solve the FLP problem. The corresponding optimal total cost is then $86301, r-value is
0.6737, or λ equals 0.3263. Accordingly, the aggregate production plan can also be obtained. The
decision maker may carry out these procedures repeatedly until a satisfying production plan is reached.
1.0
0.9
0.8
0.7 C G
0.6
0.5
0.4
0.3 D
0.2
0.1
00
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
5. Conclusions
To response to the imprecise and uncertainties in aggregate production planning, fuzzy set theory
can be employed as a mathematical tool and a fuzzy linear programming approach can be utilized for
modeling the APP of manufacturing industry. The numerical simulations presented in this paper show
a fuzzy linear programming approach provides powerful tools to the administration or the decision
makers of a manufacturing industry in developing the aggregate production plans with optimal results
and higher satisfaction
6. References
Bellman, R. and Zadeh, L. “Decision-making in a Fuzzy Environment”, Management Science, No.17, pp.141-
164, 1970.
Carlsson, C. and Pekka, K. “A Parametric Approach to Fuzzy Linear Programming”, Fuzzy Sets and Systems,
No.20, pp.17-30, 1986.
Chanas, S. “The Use of Parametric Programming in Fuzzy Linear Programming”, Fuzzy Sets and Systems.
No. 11, pp.243-251, 1983.
Hsieh, S. and Wu, M., “Demand and Cost Forecast Error Sensitivity Analyses in Aggregate Production
Planning by Possibillistic Linear Programming Models”, Journal of Intellligent Manufacturing 11, No.4, 355-364,
2000.
Kaufmann, A. and Gupta, M. Fuzzy Mathematical Models in Engineering and Management Science, North
Holland, New York, 1998.
Silva Filho, O.S., “An Aggregate Production Planning Model with Demand Under Uncertainty”, Production
Planning and Control, Vol.10, No.8, 745, 1999.
Tang, J., Wang, D. and Fung, R. “Fuzzy Formulation for Multi-product Aggregate Production Planning,”
Production Planning and Control 11, No.7, 670-676, 2000. Libraries: 46
Wang, R. and Fang, H., “Aggregate Production Planning with Multiple Objectives in a Fuzzy Environment”,
European Journal of Operational Research. Vol.133, No.3, 521-537, 2001.
Journal of Integrated Design and Process Science SEPTEMBER 2002, Vol. 7, No. 4, pp. 92
7. Appendix (Tables)
0.5 $87972
Journal of Integrated Design and Process Science SEPTEMBER 2002, Vol. 7, No. 4, pp. 94
L6 0 0 0 0 0 0 0 0 0 0 0
Pr 2 = 280 Po 2 = 53 I 2 = 29 S2 = 0 W2 = 105 H2 = 9 L2 = 0
Pr 3 = 280 Po 3 = 84 I3 = 0 S 3 = 45 W3 = 105 H3 = 0 L3 = 0
Pr 4 = 260 Po 4 = 16 I4 = 0 S4 = 0 W4 = 101 H4 = 0 L4 = 4
Pr 5 = 270 Po 5 = 30 I5 = 0 S5 = 0 W5 = 101 H5 = 0 L5 = 0