Generating daily production plans for complex manufacturing facilities using multi-objective genetic algorithms

Modern production tracking and planning systems produce good inventory plans. Unfortunately, daily production problems often produce an inventory profit that is significantly different from that of the production plan. Several manufacturing objectives are available to remedy inventory profile problems, but a weighted combination of them can be more effective. We propose using genetic algorithms to search for best weight assignments in multi-objective daily production planning problems. This approach takes snapshots of a factory's inventory, equipment capacity and demands to generate a near optimal shop floor daily production plan. The implemented system runs on a daily basis at midnight for 40 minutes and generates the following day's production plan. Tests of this system in a large-scale semiconductor manufacturing facility show the proposed approach generates production plans of high quality in reasonable run times under many factory conditions.