Order planning for manufacturing systems with load sequence-dependent degradation under human errors and imperfect inspection
针对制造系统中负载顺序相关退化、人为错误和不完美检测的交互影响，提出一个综合订单规划框架，通过扩展随机流制造网络模型和聚类增强NSGA-II算法，同时优化任务性能和成本，实验验证其优于多种先进算法。
Daily order planning is the core of dynamic decision-making that ensures the reliability and cost-effectiveness of manufacturing system delivery. However, order planning faces significant challenges due to the interplay of load sequence-dependent degradation, human errors, and imperfect inspection. The interplay of these factors introduces significant uncertainty, often leading to overestimated system performance and suboptimal planning outcomes. Thus, a comprehensive order planning framework is developed in this paper to optimise order scheduling. First, an Extended Stochastic Flow Manufacturing Network (ESFMN) model is introduced to characterise the dynamic interactions among processing machines, inspection machines, buffers, operators, and heterogeneous feedstocks (qualified and unqualified feedstocks) in manufacturing systems with load sequence-dependent degradation. Second, a semi-Markov model is employed to evaluate system mission performance that combines human errors and imperfect inspection. Third, a Clustering-Enhanced NSGA-II (CNSGA-II) algorithm is developed to tackle the order planning problem through the simultaneous optimisation of mission performance and cost. Comparative experimental results demonstrate that proposed order planning framework outperforms several advanced algorithms, including Memetic-NSGA-II, Reinforcement Learning-NSGA-II, Greedy algorithm-NSGA-II, and NSGA-II. Through a case study of a servo valve spool manufacturing system, the framework's effectiveness in practical applications is validated, confirming its ability to achieve reliable and cost-effective order planning.