Abstract: Facing the characteristics of aerospace component inspection with many varieties,small batches,centralized inspection tasks,large differences in test procedures and difficulty in data reuse,a scheduling framework for multi-category dual-feature detection resources was established.A multi-objective optimization model with the minimum average time consumption and the minimum total cost as the optimization goals was constructed,and an improved Multi-Objective Optimization Evolutionary algorithm based on Decomposition (MOEA/D) algorithm based on local optimization was proposed.The algorithm considered multi-stage real number encoding and decoding rules,and combined with local optimization operators and adaptive penalty functions based on the feasibility of the solution to ensure the quality and diversity of individuals in the population.The improved MOEA/D algorithm was used in actual cases and compared with other algorithms,and the effectiveness and superiority in solving such problems were verified.

Key words: component inspection, scheduling optimization, multi-objective optimization, multi-objective optimization evolutionary algorithm based on decomposition

摘要： 针对航天元器件检测品种多、批量小,检测任务集中、试验流程差异大、数据复用难度大的特点,建立面向航天元器件检测的订单调度框架,对多品类双特性资源进行差异化调度处理。构建以订单平均耗时与检测总成本最小化为目标的多目标优化模型,并提出一种基于局部优化的改进MOEA/D算法。同时考虑了多段式实数编码解码方案,结合以解方案可行程度为基础的局部优化算子与自适应惩罚函数,保证了种群中个体的质量与多样性。最后,以某航天元器件检测单位实际业务为案例进行方法验证,对比了所提算法与改进NSGA-Ⅱ算法、经典MOEA/D算法、NSGA-Ⅲ算法的优化效果,验证了所提方案在解决此类问题上的优越性。

关键词: 元器件检测, 调度优化, 多目标优化, 基于分解的多目标进化算法