.png){kind=icon}
.png){kind=icon}
Return
A Q-learning-driven multi-objective algorithm for dynamic hybrid flow shop rescheduling with setup and new job arrivals
.png){kind=icon}
PRE
AI
Y
J
H
Y
L
H
DOI：10.1016/j.swevo.2026.102495.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In real production environments, sequence-dependent setup times and the random arrival of new jobs often coexist. To this end, this paper investigates the dynamic rescheduling problem in a hybrid flow shop with sequence-dependent setup times and new job arrivals,which is solved by a dynamic Q-learning enhanced non-dominated sorting genetic algorithm (DQ-NSGA-II). First, a multi-objective optimization model is constructed, aiming to simultaneously optimize total energy cost, carbon trading cost, and weighted earliness/tardiness penalties. Second, the decoding process is restructured for dynamic job insertions, and a hybrid initialization strategy combining multi-objective heuristics with random solutions is designed to balance exploration and exploitation. Third, an adaptive crowding distance mechanism is introduced to improve the distribution uniformity of the Pareto optimal set in the objective space. Furthermore, to effectively address dynamic disruptions, a partial rescheduling strategy is designed, and a stability metric is incorporated to balance the trade-off between scheduling performance improvement and system stability. Finally, the proposed DQ-NSGA-II is compared in detail with three high-performance algorithms. Experimental results demonstrate that the proposed method can generate high-quality and well-distributed Pareto optimal sets, exhibiting superior responsiveness and stability.
Journal
IF:
8.5
Papers:
2.1K
Citations:
1.0W
Organization
Cited Papers