Research on intelligent production scheduling optimization in the process manufacturing industry based on knowledge graphs and improved NSGA-II

With increasing competition in the fast-moving consumer goods industry, cigarette manufacturers face challenges such as multiple product types, small batch sizes, and rapid delivery. Traditional scheduling relies on human experience, struggling to handle complex constraints dynamically, leading to resource waste and low efficiency. This paper proposes an intelligent scheduling method that integrates knowledge graphs with an improved Non-dominated Sorting Genetic Algorithm II (NSGA-II). By constructing two-layer knowledge graphs for monthly planning and cigarette packaging machines, implicit knowledge is made explicit. A scheduling optimization algorithm based on heuristic rules and non-dominated sorting is presented. The model considers equipment capacity, changeover time, production continuity, and inventory levels, forming a multi-objective, multi-constraint optimization problem. To address NSGA-II’s slow convergence and limited population diversity, a two-level chromosome encoding and adaptive mutation strategy are introduced, enhancing Pareto front exploration. Experiments based on real production data from ZCT show the method reduces monthly finished goods inventory by 16.23%, increases equipment utilization by 11.4%, and decreases manual intervention. The study provides theoretical and practical support for intelligent manufacturing transformation in the cigarette industry and offers a generalizable methodology for process industry scheduling optimization.