| Summary: | 博士 === 國立清華大學 === 工業工程與工程管理學系 === 88 === The goal of this thesis is to develop an interactive computer aided scheduler (ICAS) for solving scheduling problems of wafer sorting in semiconductor manufacturing. The ICAS composed of three major components, via. simulation mechanism, parameterized dispatching mechanism, and alternative schedules generator. In the interactive scheduling environment, ICAS provides a set of alternative schedules for a human, by using the three components, and then the human is responsible for evaluating scheduling performance and selecting one of the alternatives to implement. Since both production information and scheduling methods are isolated from human, the difficulties of scheduling task can be minimized.
In constructing the simulation mechanism, the input/output data of simulation models has to be defined first. In order to facilitate the construction of simulation mechanism, transformation between colored timed Petri net and simulation pseudo code is presented in this thesis by using a three-step development procedure. In the parameterized dispatching mechanism, since the multi-criteria and sequence-dependent setup time are present in the wafer sorting scheduling problem, an interactive scheduling tool which is controlled by a set of dispatching parameters is proposed. This tool can generate schedules with different achievement of multiple scheduling objectives by properly adjusting the values of dispatching parameters. In the alternative schedules generator, a genetic algorithm is designed to determine the values of dispatching parameters. This algorithm observes the gap between current schedules and the goal the human wants to achieve. Three genetic operators are utilized to reduce the gap by exploring new settings of dispatching parameters.
To evaluate the performance of ICAS, two experiments performed on a parallel machines scheduling problem and an industrial case study were conducted. In the former, ICAS was shown to outperform other ten well-known priority rules in all the 7 objectives and 8 experimental environments. In the latter, ICAS performed significantly better than the manual scheduling approach and 6 priority rules when hit rate of demand, machine setup, and machine idle time were simultaneously considered.
|
|---|
