Track: Facilities Planning and Management

Abstract

The Facility Layout Problem (FLP) is all about obtaining the optimal arrangement of machines or departments in a given area or plant. Most of the research in the area of the FLP assumes that the material flow between the departments is constant. However, in today’s competitive and volatile manufacturing environments, this assumption does not usually hold. Layouts constantly need to be changed from one period to the other to adapt to the changes in the product mix and demand. The Dynamic Facility Layout Problem (DFLP) searches for a layout for each period of the planning horizon minimizing its respective total material handling costs, for all periods,  as well as the total rearrangement costs between successive periods. Because of the prohibitive combinatorial nature of DFLP, absolute optimality is not usually sought, and larger instances of the problem are solved using near-optimal algorithms. In this paper, Dynamic Programming (DP) is used to solve the DFLP problem. The layouts considered in each period forms the state space for DP; the problem is decomposed to stages— one for each DFLP period. This state space is arrived at using a hybridized heuristic approach that includes the use of a Simulated Annealing (SA) metaheuristic. Consistent with the rest of the literature, the proposed method is comprised of two phases; in the first phase, the set space to be considered and searched in each phase of DP are arrived at using SA and two different heuristic procedures. To obtain these layouts and for smaller manageable instances of the problem, exact optimum layout for each period is found by solving the QAP formulation of the static facility layout problem. As for larger instances and due to the inherent combinatorial nature of the problem, sub-optimum layouts for each period are arrived at using SA. Two other heuristics are utilized to complete the set space of layouts for each stage by applying different neighborhood functions to the previously obtained exact- and sub-optimum solutions. In the second phase, the DP recursive formulation is solved. Computational experience exhibits promise.