Professional Documents
Culture Documents
*
This research has been carried out in the framework of a project funded by the Science and Technology Ministry of
the Spanish Government, titled ‘Business process integration, knowledge management and decision support tools in
supply chain of industrial SMEs’. Ref. DPI2002-01755. www.gnosis.cigip.org.
547
EUSFLAT - LFA 2005
2 Formulation of the problem Billington, Mcclain and Thomas [2] although the
setup times have not been included. The decision
A Linear Programming (LP) model for the capacity variables Tocrt and Texrt are not limited by any
constrained MRP problem originally proposed in [9]
established parameter but are penalized with the
and called MRPDet is adopted as the basis for our
corresponding costs in the objective function. This is
work. MRPDet is a model for the optimization of the
to provide the most possible generality to the model.
medium term production planning problem in a
The limitation of these variables for specific
capacity constrained MRP, multi product, multi applications could be easily considered taking into
level and multi period manufacturing environment. account that if those limits are exceeded the solution
Equation (1) shows the total costs to be minimized: of the model could be no feasible.
costs of the inventories cii, costs of the extra time
A constraint has also been added (4) to finish with
used by resources, ctexrt, and costs of the lazy time
the delays in the last period (T) of the planning
of resources, ctocrt. The MRPDet includes a plan to horizon.
satisfy the delayed demands penalized with a cost,
crdi. It is assumed that this cost is linear to the The model also contemplates the non negativity
number of backlogs in every period. constraints (5) for the decision variables.
The balance equations for the inventory are given by Finally, the decision variables Pit, INVTit and Rdit
the group of constraints (2). These equations take will be defined as continuous or integer variables
into account the backlogs of the demand which depending on the manufacturing environment where
behave as a negative inventory. It is important to the model is applied.
highlight the consideration of the parameter RPit that Let us consider the following fuzzy formulation of
guarantees the continuity of the MRP along the the MRPDet model. Decision variables and
successive explosions carried out during a given parameters for the model are defined in Table 1.
planning horizon.
The production in every period is limited by the
availability of a group of shared resources. The
equation (3) considers the limits of capacity of these
resources. This equation has been thought in a
similar way that in the model proposed by
I T
~
R T
Minimize z = ∑∑ cp P
i it + cii INVTit + crd i Rd it + ∑∑ (ctocrt Toc rt + ctexrt Texrt ) (1)
i =1 t = 1 r =1 t =1
Subject to
I
~
INVTi ,t −1 + Pi ,t −TSi + RPit − INVTi ,t − Rdi ,t −1 − ∑ aij ( Pjt + RPjt ) + Rdit = d it i = 1…I, t = 1…T (2)
j =1
I ~ ~
∑ AR
i =1
ir Pit + Toc rt − Tex rt = CAPrt r = 1…R, t = 1…T (3)
548
EUSFLAT - LFA 2005
1
r − r ( x − r2 ) + 1
2 1 if ( r1 ≤ x ≤ r2 )
1
µ r ( x) = ( x − r2 ) + 1 if ( r2 ≤ x ≤ r3 ) (6)
r2 − r3 if ( x ≤ r1, r3 ≤ x )
0
I T
R T
∑∑ + + − α + α it + ∑∑ (ctocrt Tocrt + ctexrt Texrt )
Minimize z = cp P
i it cii INVTit ((1 ) crd i1 crd i2 ) Rd (7)
i =1 t =1 r =1 t =1
Subject to
I
INVT i ,t −1 + Pi ,t −TS i + RPit − INVT i ,t − Rd i ,t −1 − ∑ a ij ( Pjt + RP jt ) + Rd it ≤ (1 − α ) d it 3 + αd it 2 i = 1…I, t = 1…T (8)
j =1
I
INVT i ,t −1 + Pi ,t −TS i + RPit − INVT i ,t − Rd i ,t −1 − ∑ a ij ( Pjt + RP jt ) + Rd it ≥ (1 − α ) d it1 + αd it 2 i = 1…I, t = 1…T (9)
j =1
∑ ((1 − α ) AR
i =1
ir1 + αARir 2 ) Pit + Tocrt − Texrt ≤ (1 − α )CAPrt 3 + αCAPrt 2 r = 1…R, t = 1…T (10)
∑ ((1 − α ) AR
i =1
ir 3 + αARir 2 ) Pit + Tocrt − Texrt ≥ (1 − α )CAPrt1 + αCAPrt 2 r = 1…R, t = 1…T (11)
549
EUSFLAT - LFA 2005
550
EUSFLAT - LFA 2005
The evaluation method consists on the comparative In the case of the fuzzy model that provides a fuzzy
analysis of the performance of the two models, i.e. solution, the fuzzy set of the decision has been
the fuzzy model and the deterministic model, obtained. Next, it has been chosen as crisp solution
according to a group of parameters defined in [9]: (i) which obtains the best results in the highest number
the total costs; (ii) the service level; (iii) the levels of of the evaluated indicators. Along this section it will
inventory; (iv) the planning nervousness respect to only be considered for this model the selected crisp
the planned period and the planned quantity (see solution, where the parameter α was established at 0.
Table 2); and (v) the computational efficiency
(Table 3).
Table 2: Evaluation of the results.
Number of
minimum Planning Planning
Service Inventory nervousness nervousness
Model level (%) levels (period ) (quantity) Total costs (€)
MRPDet 99.44 8 0.35 14.05 3.988.276,07
Fuzzy MRP 99.49 19 0.35 13.50 2.938.857,69
Table 3: Efficiency of the computational experiments for a MRP execution (first week).
Array
Elements Density CPU Time
Model Iterations Variables Integer Constraints non zero (%) (seconds)
MRPDet 0 4237 5612 2797 8239 0.07 0.86
Fuzzy MRP 0 4237 5612 2857 8477 0.07 3.74
Both models present an average service level above requirements of information storage and a moderate
99.5% The fuzzy model provides a lightly better increment of the required CPU time.
value in this performance indicator. Contrarily, the
6. Conclusions.
fuzzy model generates a higher number of minimum
inventory levels, i.e. 19 from the 46 evaluated items In many manufacturing environments, such as the
present lower inventories with the production plans automobile industry, the production planning
obtained by the fuzzy model in contrast with the 8 decisions have to be made under conditions of
items with lower inventories provided by the uncertainty in parameters as important as costs,
MRPDet model. market demand or capacity data.
Both models have presented a similar nervousness A model based on fuzzy mathematical programming
with respect to the planned time period. On the other for production planning under conditions of
hand, the fuzzy model presents the best value of uncertainty has been proposed. In a general way, the
nervousness with respect to the planned quantity. structure of the fuzzy model has been able to
increase the group satisfaction (level of service,
Also, the fuzzy model generates lower total costs
inventory levels, planning nervousness and total
than MRPDet. These differences in the total costs
costs) without causing an explosive growth of the
are due, mainly, to two aspects: (i) The
computational effort.
consideration of possible future variations of the
demand that originates larger production and Presented the research conclusions to the company,
inventories with the objective of avoiding the the staff in charge of planning have shown their
strongly penalized demand backlogs, and (ii) the interest in the new model that it would allow them to
strict constraints of the deterministic model, where consider in a better way the demand variability and
the required capacity and the available capacity of to introduce their perceptions.
the assembly line are fixed rigidly.
References
Finally, the fuzzy model has provided freedom of
action with regard to problems where ambiguous [1] Bellman, R. and Zadeh, L.A., “Decision-making
values appear with a minimum increment of the in a fuzzy environment”, Management Science,
vol. 17, pp. 141-164, 1970.
551
EUSFLAT - LFA 2005
[2] Billington, P.J., Mcclain, J.O. and Thomas, L.J. Journal of Production Economics, vol. 33, pp.
“Mathematical programming approaches to 45-52, 1994.
capacity constrained MRP systems: Review, [8] Maximal Software Incorporation, MPL
formulation and problem reduction”, modelling system. Release 4.11, USA, 2000.
Management Science, vol. 29´, pp. 1126-1141, [9] Mula, J., Models for production planning under
1983. uncertainty. Application in a company of the
[3] CPLEX Optimization Inc., Using the CPLEX automobile sector, PhD (in Spanish), Polytechnic
callable library, 1994. University of Valencia, Spain. Ed. SP-UPV,
[4] Gen, M. Tsujimura, Y. and Ida, K., “Method for ISBN: 84-688-6987-2, 2004.
solving multiobjective aggregate production [10] Mula, J., Poler, R., García, J.P. and Ortiz,
planning problem with fuzzy parameters”, A.,”Supply planning and demand behaviour in an
Computers and Industrial Engineering, vol. 23 , automobile industry supply chain”, International
(1-4), pp. 117-120, 1992. Conference on Industrial Engineering and
[5] Hsu, H. and Wang, W., “Possibilistic Production Management (IEPM’03), Porto
programming in production planning of (Portugal), May, 26-28, 2003.
assemble-to-order environments”, Fuzzy Sets and [11] Orlicky, J., Material Requirements Planning,
Systems, vol. 119, pp. 59-70, 2001. McGraw Hill, London, 1975.
[6] Inuiguchi, M. and Ichihashi, H., “Relative [12] Vollmann, T.E., Berry, W.L. and Whybark,
modalities and their use in possibilistic linear D.C., Manufacturing planning and control
programming”, Fuzzy Sets and Systems, vol. 35, systems. Third Edition, Irwin, Homewood,
pp. 303-323, 1990. Illinois, 1992.
[7] Inuiguchi, M., Sakawa, M. and Kume, Y., “The [13] Zadeh, L.A., “Fuzzy sets”, Information Control,
usefulness of possibilistic programming in vol. 8, pp. 338-353, 1965.
production planning problems”, International
552