| This dissertation considers several dynamic scheduling problems associated with continuous-flow flexible manufacturing systems (FMS). FMS consist of one or more machines that each are able to perform operations on multiple part-types. Appropriate use of this flexibility results in reduced inventories and faster responses to demand fluctuations.; Using a general framework, conditions are provided under which the optimal production policy is chosen from the class of myopic scheduling policies. Myopic policies are particularly appealing since only information about the current system state is required to implement them. Also, several counterexamples, explicitly illustrating performance limitations of myopic policies, are presented.; For a failure-prone system producing multiple part-types, a linear switching curve (LSC) policy is proposed as a practical, near-optimal solution. The objective is to minimize a weighted sum of expected quadratic buffer costs. In general, the optimal solution to this problem is unknown. However, by further restricting the allowable control set to the class of prioritized hedging point (PHP) policies, simple, closed-form expressions for the optimal hedging points are determined, Also obtained are some structural results for determining the optimal priority ordering.; These analytical results are then extended to systems operating under general LSC policies. An interesting byproduct of this analysis is that, in the shortage region, the slope of the optimal LSC policy is equal to that of the myopic policy, which is the overall optimal control in that region.; For weighted, absolute-value instantaneous buffer costs, the optimal slope and hedging points for the LSC policy cannot be obtained analytically. Thus, a numerical stochastic optimization technique is employed that uses infinitesimal perturbation analysis to obtain derivative estimates from a single simulation run.; Also considered is the effect of yield uncertainty in the supply chain. It is assumed that the probability distribution of the random yield rate is known, but the inventory level is observable only intermittently. The optimal production control, that minimizes a linear combination of expected surplus and shortage costs over the planning horizon, is shown to be piecewise constant, and the appropriate production levels and control break-points are determined as functions of the yield rate distribution. |
