Forecast-corrected production-inventory control policy in unreliable manufacturing systems

In traditional research on production-inventory control problems with failure-prone manufacturing systems, a stationary demand process is an essential assumption. However, such a situation may not be true. This study extends the hedging-point-based production-inventory control problem into the case with non-stationary demand. The demand forecasting process is simulated and categorised into two different cases. First of all, a two-level control policy is proposed to solve the problem with a Markov modulated Poisson demand process which is often used in qualitative forecasting. Then the quantitative forecasting process using time series methods is modelled and a forecast-corrected control policy is proposed accordingly. The impact of forecasting on the system performance is then investigated. An integrated simulation and experimental design method was adopted to solve the modified optimal control problem. The results show that the proposed control policy can outperform the traditional stationary policy when the forecasting error is limited to a certain level.