Comparison of Deterministic and Stochastic Production Planning Approaches in Sawmills by Integrating Design of Experiments and Monte- Carlo simulation

摘要

AbstractudForest industry is one of the key economic initiatives in Quebec (Canada). This industry has recently faced some obstacles, such as the scarcity of raw material, higher competitiveness in the market and new obligations applied in North America regarding sustainable development. These problems force lumber industries to improve their efficiency and become more service sensitive, in order to ensure the on time demand fulfillment. To achieve the goals aforesaid, one solution is to integrate the uncertainties more appropriately into production planning models. Traditional production planning approaches are based on deterministic models which in fact, ignore the uncertainties. A stochastic production planning approach is an alternative which models the uncertainties as different scenarios. Our goal is to compare the effectiveness of deterministic and stochastic approaches in sawing unit of sawmills on a rolling planning horizon. The comparison is performed under different circumstances in terms of length of planning horizon, re-planning frequency, and demand characteristics defined by its average and standard deviation. The design of experiments method is used as a basis for performing the comparison and the experiments are ran virtually through Monte-Carlo simulation. Several experiments are performed based on factorial design, and three types of robust parameter design (Taguchi, combined array, and a new protocol) which are integrated with stochastic simulation. Backorder and inventory costs are considered as key performance indicators. Finally a decision framework is presented, which guides managers to choose between deterministic and stochastic approaches under different combinations of length of planning horizon, re-planning frequency, and demand average and variation.ududKey words: sawmills, production planning, design of experiments, robust parameter design, uncertainty, Monte- Carlo simulationud