Robust scheduling for feed production: a stability-driven approach to minimize makespan and energy under uncertainty

Abstract The combination of multi-stage processing, sequence dependent setup time along with intensive energy demands makes it a challenging task to deal with scheduling of feed production. A modified multi-objective evolutionary algorithm, termed non-dominated sorting genetic algorithm II-adaptive-dynamic differential evolution (NSGA-II-ADDE), is proposed seeking to balance between makespan and total energy consumption. Coupled with a warm-start heuristic initialization strategy that utilizes reciprocal weighting of time and energy, a hierarchical hybrid chromosome encoding is employed to enforce campaign production constraints, such that resource-efficient machinery is prioritized in the framework. To accelerate convergence, a diversity-driven adaptive mutation strategy is introduced to dynamically switch between global exploration and local exploitation. Experimental results show that NSGA-II-ADDE consistently outperforms standard baseline algorithm NSGA-II in locating superior global optima on Pareto front as well as much achieving faster convergence depicted by hypervolume. Besides, the proposed scheduling approach is proved to improve substantially in operational fidelity and resilience against real-world manufacturing disruptions with a much higher stability index under stochastic Monte Carlo simulations.