Multi-objective optimization of production control mechanisms for multi-stage serial manufacturing-inventory systems

Single-model multi-stage serial production/inventory systems with stochastic order arrival and service times are examined. The manufacturing systems are controlled by Kanban, Base Stock, CONWIP, and CONWIP/Kanban Hybrid mechanisms. Discrete-event simulation models of the manufacturing systems are developed. Four simulation cases are examined where optimal or near-optimal parameters for the control policies are obtained by integrating the simulation models with multi-objective evolutionary algorithm in order to minimize mean WIP and mean number of backordered demands simultaneously. The non-dominated sets are compared in terms of several metrics for comparing Pareto fronts.