A Learning-Driven Dynamic Dual-Population Evolutionary Algorithm for Green Flexible Job Shop Scheduling with Preventive Maintenance

Production scheduling is widely acknowledged as a pivotal element in the manufacturing process. Within the context of modern sustainable manufacturing, energy consumption reduction has emerged as a significant challenge faced by manufacturing enterprises. Among various strategies for energy reduction, green production scheduling provides a practical and cost-effective solution that can be seamlessly integrated into existing production systems without significant financial investment. This paper investigates a green flexible job shop scheduling problem with the consideration of preventive maintenance activities (GFJSP-PM). A mathematical model is established with the objective of minimizing total energy consumption within the workshop. Then, a learning-driven dynamic dual-population evolutionary algorithm (LDDEA) is proposed for solving the problem. During the evolutionary process, the entire population is split into two subpopulations to enable differentiated search strategies. A Q-learning-driven population management mechanism is presented to adaptively adjust the individual distribution between the subpopulations. Furthermore, problem-specific genetic operators and a local search algorithm are designed to enhance the solution quality. To test the LDDEA’s performance, extensive computational experiments are carried out based on both benchmark and randomly generated instances, along with a real-world engineering case. The experimental results indicate that LDDEA exhibits competitive performance in addressing the problem under consideration.