A decomposition methodology for scheduling complex job shops
This work addresses the problem of scheduling large complex job shops characterized by different types of workcenters, sequence-dependent-setup times, reentrant product flows, and due date related performance measures. These problems have been largely ignored by researchers to date due to their extreme complexity. However, the fact that these problems are encountered in a large variety of industries makes the task of developing effective scheduling methodologies vital for the well being of these industries. The availability of powerful workstations and Computer Aided Manufacturing (CAM) systems capable of tracking work-in-process inventory in real-time gives us the proper medium in which to develop these methodologies. The methodology developed decomposes the job shop into workcenters, schedules the workcenters in order of their criticality and constructs a schedule for the whole facility using the schedules for the workcenters. We use the disjunctive graph representation of the job shop to capture the interactions between workcenters and develop rolling horizon procedures for single and parallel identical machines to schedule the individual workcenters. We compare the performance of the decomposition methodology against those of several dispatching rules using randomly generated problems motivated by a semiconductor testing facility. Computational experiments show that the decomposition methodology performs significantly better than the dispatching rules both on average and in the worst case.
Uzsoy, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our