A Feature-Based Learning Differential Evolution Algorithm for the Flexible Job-Shop Scheduling With Occupational...

Area of Science:

• Operations Research
• Artificial Intelligence
• Industrial Engineering

Background:

• Learning differential evolution (DE) algorithms are common for flexible job-shop scheduling problems (FJSPs).
• Traditional DE methods often lack the ability to fully utilize problem-specific feature information.
• Integrating worker health considerations, like fatigue, into scheduling is crucial for realistic optimization.

Purpose of the Study:

• To propose a novel feature-based learning DE algorithm (FLDE) for addressing FJSPs.
• To incorporate worker health, specifically fatigue measured by the Occupational Repetitive Actions Index (OCRA), into the scheduling optimization process.
• To enhance the local optimization capabilities of DE algorithms for FJSPs.

Main Methods:

• Development of a feature-based learning DE algorithm (FLDE).
• Integration of the Occupational Repetitive Actions Index (OCRA) to ensure scheduling solution feasibility and worker well-being.
• Design of a feature-based decision model (FDM) for adaptive optimization operator selection.
• Introduction of a critical operation search method for extracting scheduling solution features.

Main Results:

• The feature-based decision model (FDM) significantly improves the local optimization performance of the FLDE algorithm.
• FLDE demonstrates superior performance compared to traditional algorithms across 40 diverse FJSP instances.
• The inclusion of OCRA ensures that generated schedules are feasible with respect to worker health constraints.

Conclusions:

• FLDE offers an effective approach to solving FJSPs, particularly when worker health is a critical constraint.
• The feature-based decision model is key to enhancing the adaptability and efficiency of learning DE algorithms.
• This research provides a valuable framework for developing more human-centric and optimized scheduling solutions.