| Previous research on the multiple orders per job (moj) scheduling problem, as motivated by 300-mm semiconductor manufacturing, has focused on the development of optimization-based solution procedures for a variety of machine environments. Due to the NP hard complexity of this problem, heuristic solution procedures are necessary for analyzing problems in a computationally efficient manner. This thesis investigates the identical, parallel machine version of moj scheduling by developing a random keys genetic algorithm which can solve reference problem instances for total weighted completion time, as well as instances for a corresponding total weighted tardiness objective, to within 1% of optimal, on average, for 10-order problems. A non-dominated, sorting-based genetic algorithm is then developed to find the efficient frontier of solutions considering both objective function values. |
