Modular scheduling of tightly coupled production lines

Area of Science:

• Industrial Engineering
• Operations Research
• Computer Science

Background:

• Optimizing tightly coupled modular production lines is complex due to diverse products, varied constraints, and precise timing needs.
• Modular setups offer flexibility and cost savings but complicate scheduling due to distributed decision-making and limited system views.
• Local scheduling decisions in modular systems can significantly impact global schedule feasibility.

Purpose of the Study:

• To propose a distributed scheduling method for tightly coupled modular sequential production lines.
• To address the challenge of coordinating modules with limited local information for global schedule optimization.
• To develop a multi-agent framework that facilitates consensus on job handover times.

Main Methods:

• Developed a multi-agent framework with a central system agent and local agents.
• The system agent propagates timing constraints and seeks consensus on job handover times between modules.
• Integrated existing local schedulers without modification through local agents.

Main Results:

• The distributed scheduling method was tested on complex re-entrant flow-shop instances with setup times and due dates.
• Schedules generated by the distributed method showed near-optimal performance, with an average makespan only 1.15% larger than a hypothetical optimum.
• The approach effectively managed timing constraints and achieved global schedule feasibility in a modular environment.

Conclusions:

• The proposed distributed scheduling method is effective for tightly coupled modular sequential production lines.
• The multi-agent framework successfully enables consensus-building for globally feasible schedules.
• This approach offers a practical solution for optimizing productivity in industries like semiconductor manufacturing.