Evolutionary optimisation of hedging points for unreliable manufacturing systems

The production rates of manufacturing systems are notoriously difficult to control, since such systems are dynamic, uncertain and non-linear. However, the introduction of hedging-point policies for such systems has led to much progress in optimal production control. But the theoretical results so far obtained for such hedging-point policies are still far from complete, since the optimal hedging points (i.e., the optimal inventory levels) are analytically available only for simple systems and under restrictive assumptions. In this paper, an evolutionary stochastic optimisation procedure is proposed to estimate the short-run optimal hedging points for failure-prone manufacturing systems under crisp-logic control. This methodology is illustrated by examples and is validated by comparing the evolutionary results with the available analytical long-run solutions. The proposed evolutionary methodology is also shown to be capable of generating optimal hedging points for unreliable systems producing multiple products with different priorities. In addition, the relative merits of genetic algorithms, evolution strategies and adaptive evolution strategies in hedging-point optimisation are compared.