This project addresses the problem of scheduling large job shops characterized by dynamic job arrivals, machine failures, due- date related performance measures and different types of workcenters. In spite of their practical importance, the difficulty of these problems has caused them to remain largely unstudied to date. However, with the exploitation of special problem structures present in industrial contexts provides a promising approach. An event classification approach to the dynamic aspects of the problem is used. It is combined with an approximation methodology that can obtain near-optimal schedules for a static problem. Events affecting system state are monitored and those needing immediate attention (exceptions) are identified. In the absence of exceptions, schedules are generated periodically and implemented on a rolling horizon basis. If an exception occurs, the system is rescheduled immediately. An approximation methodology is proposed to generate the schedules at the various rescheduling points. This requires the development of efficient scheduling procedures for different types of workcenters, leading to the formulation of a number of scheduling problems not extensively examined to date. The use of a specific industrial setting, that of semiconductor testing operations, as a research vehicle will enable the evaluation of the proposed research from a practical standpoint.
