Learning dispatching rules via novel genetic programming with feature selection in energy-aware dynamic job-shop...

Area of Science:

• Operations Research
• Industrial Engineering
• Artificial Intelligence

Background:

• Integrating energy conservation into production efficiency is vital for modern industry.
• Dynamic job shop scheduling presents challenges in optimizing both production and energy consumption.
• Existing methods often lack interpretability and efficiency in energy-aware scheduling.

Purpose of the Study:

• To develop interpretable and high-quality dispatching rules for energy-aware dynamic job shop scheduling (EDJSS).
• To propose a novel genetic programming (GP) method with online feature selection for automatic rule learning.
• To enhance the balance between exploration and exploitation in GP for superior rule generation.

Main Methods:

• A novel genetic programming approach with an online feature selection mechanism was developed.
• The GP method dynamically adjusts population diversity based on stopping criteria and elapsed duration.
• The proposed approach was benchmarked against three other GP algorithms and 20 traditional rules.

Main Results:

• The proposed approach significantly outperformed existing GP methods and benchmark rules.
• It generated more interpretable and effective dispatching rules for energy-aware scheduling.
• Average improvements of 12.67% (EMS), 15.38% (EMWT), and 11.59% (EMFT) were observed.

Conclusions:

• The novel GP method is highly effective for learning dispatching rules in energy-aware dynamic job shop scheduling.
• The approach offers a superior balance of interpretability and performance compared to existing techniques.
• This research contributes to more sustainable and efficient industrial production practices.