The impact of non-stationary demand and forecasting on a failure-prone manufacturing system

Stationary demand process is mostly an assumption in the problem of production/inventory control. The objective of this paper is to investigate the performance of non-stationary control policy and stationary control policy under the condition of non-stationary demand and to study the impact of forecasting on the system's performance in a failure-prone manufacturing system. The hedging-point-based (HPP) production/inventory control policy is adopted and modified to solve this specific problem. The problem is formed as a dynamic programming model. Non-stationary demand is forecasted using a few time-series forecasting methods. Discrete event simulation, experimental design and response surface method are combined together to simultaneously obtain the optimal lot size and hedging point considering the production cost, inventory cost and setup cost.. The results show that different forecasting methods produce varies accuracy and excessive forecasting inaccuracy deteriorates the performance of the non-stationary control policy. Non-stationary control policy generally can provide better performance when compared with the traditional stationary one.