Managing stochastic multiproduct systems: Model, measures, and analysis

We consider a model for managing a single stage that produces multiple items. The production rates are finite, and there are switchover times. The interarrival times and quantities of demands for the items are random, and demand may occur for a set of items. We consider order focused measures: cost based on response times; service levels based on quoted lead times and Type-1 service. We operate the stage in the following manner: (1) there is a cyclic schedule that determines the sequence of items and the number of times a particular item is produced in a cycle; (2) given a cyclic schedule, production of each item follows a modified base-stock policy or a (s, S) policy. We present a simulation based procedure to obtain good values for the base stock levels or S (for any fixed S–s) for each of the above performance measures. Numerical results indicate that good solutions can be obtained with modest computational effort. We also report on a real world implementation of this model.