Modeling uncertainties in plant layout problems

Abstract The plant layout problem deals with the problem of allocation of departments to sites to minimize the total material handling cost. The problem is generally solved based interdepartmental flows for a single period. Due to the dynamic nature of businesses — growth, fluctuating demands, changes in product mix — the optimal layout should in general, be different from period to period. A relocation cost is incurred whenever a layout is changed from one period to the next. Moreover, due to uncertainties in future predictions, system variables such as growth, demand and product mix cannot be deterministically known for the future periods. Often the interdepartmental flows for future periods can only be probabilistically predicted. In the face of these uncertainties, what should the plant layout be in future periods so that the sum of the expected material handling cost and relocation cost is minimized? This paper focuses on this issue of modeling uncertainties in plant layout problems. An exact method and heuristics are suggested to solve the resulting stochastic dynamic plant layout problem. The heuristics proposed were able to generate good solutions in a reasonable amount of time for problems with up to 40 departments. Our simulation studies indicate that a rolling horizon approach yields better results instead of using a fixed horizon approach.