DOE-based enhanced genetic algorithm for unrelated parallel machine scheduling to minimize earliness/tardiness costs

Subject Areas : Design of Experiment

Parsa Kianpour ^{1 *} , Deepak Gupta ² , Krishna Krishnan ³ , Bhaskaran Gopalakrishnan ⁴

1 - Senior Data Scientist, Cummins, USA
2 - Department of Industrial, Systems, and Manufacturing Engineering, Wichita State University, USA
3 - Department of Industrial, Systems, and Manufacturing Engineering, Wichita State University, USA
4 - Industrial and Management Systems Engineering, West Virginia University, USA

Accepted : 2023-08-01 Published : 2024-09-01

Keywords:

Abstract :

This study presents an enhanced genetic algorithm (E-GA) to minimize earliness/tardiness costs in the job shop environment. It considers an unrelated parallel machine scheduling problem with a limit on maximum tardiness levels. This problem is motivated by the experience of one of the authors in a job shop supporting the local aircraft industry that requires strict control on delivery times. Current literature does not consider this critical restriction and unsuccessfully tries to deal with them using higher penalty costs. The proposed method uses the design of experiment (DOE) concept while optimizing the GA operators. Furthermore, it improves the initial solution using a hybrid dispatch rule through a strategic combination of construction and improvement heuristics. The model was applied to a local job shop. The results indicate that E-GA provides a schedule with lower cost and reduced computational time compared to existing dispatch rules in the literature and existing algorithms (OptQuest).