Minimizing total tardiness of jobs on batch processing machines with incompatible job families

Abstract

In this thesis, we present production scheduling algorithms for batch processing machines (at the diffusion workstation) in a semiconductor wafer fabrication facility. The diffusion workstation consists of identical parallel machines, in which lots from the same family are processed as a batch. A setup operation is required in these machines if the family type of a batch just processed is different from that of the batch to be processed next. For the scheduling problem with the objective of minimizing total tardiness of jobs, we develop two types of heuristics, a dispatching rule-based algorithm and an improvement algorithm. In the former heuristics, priorities of the batches are determined based on three factors, i.e., need for a batch setup, batch due date that is computed from due dates of the jobs included in the batch, and the composition ratio of the batch. In the latter heuristics, it is determined to process the batch (selected to be processed in the first algorithm) without delay considering the next arrival job of the selected family. A series of computational experiments was done to evaluate performance of the heuristics, and results show that the suggested heuristics outperformed a method that have been used in a real manufacturing system and a simple algorithms based on the minimum batch size rule and the earliest due date rules.