Real-time dynamic shop floor scheduling using evolutionary algorithms

Evolutionary Algorithms (EA) possess significant potential for solving manufacturing control problems. Despite the fact that the value of EA has become increasingly apparent over the past years, there are only few real-world applications of these methods. Proponents of the EA approach argue that one reason for this is the potential of these algorithms to adapt to sudden changes like machine breakdowns immediately by defining improved schedules in a continuous process. The new schedules often differ significantly from the previous schedules used by the supervisor before the changes occurred. This contradicts organizational demands for predictable schedules that fix the work order of a machine at the beginning of a planning period allowing only minor changes afterwards. This is primarily caused by the inability of today's organizational systems to distribute new schedules quickly without causing major disturbance in the manufacturing process. While plant data acquisition (PDA) has been automated for years, distribution of work order information is in most cases still done manually. The authors introduce an evolutionary search algorithm for shop floor scheduling and show how to integrate the scheduler with a bidirectional PDA-system used for data collection as well as distribution of sequencing information. In addition, quality improvements of evolutionary tools by separating the time-consuming scheduling procedure from the EA and distributing it throughout intelligent data terminals is discussed.