动态多目标优化问题具有多个相互冲突的目标,而且这些目标也受环境的影响不断变化,为了快速准确跟踪不断变化的Pareto前沿和Pareto解集,提出一种基于迁移学习的拐点预测策略(a knee points prediction strategy based on transfer learn...动态多目标优化问题具有多个相互冲突的目标,而且这些目标也受环境的影响不断变化,为了快速准确跟踪不断变化的Pareto前沿和Pareto解集,提出一种基于迁移学习的拐点预测策略(a knee points prediction strategy based on transfer learning, TKPS)。TKPS根据记忆过去时刻种群中优秀个体,使用迁移学习算法得到映射矩阵W,然后通过映射矩阵W,把当前时刻的拐点集映射到高维希尔伯特空间,从中找到下一时刻的拐点集,引导种群收敛;同时,在拐点的邻域内选出若干个伴随个体,增加种群多样性,避免种群陷入局部最优。TKPS采用8个测试函数,并与其它3个算法结果对比分析,实验结果表明TKPS算法具有更快的响应环境变化的能力。展开更多
为更好地应对动态多目标优化中的环境变化,提出了一种对差分向量进行角度修正以及分级多种群协同进化(Angle Correction and Hierarchical Multi-Population,ACHMP)的进化算法.根据历史信息,使用无迹卡尔曼滤波模型来预测种群的中心点,...为更好地应对动态多目标优化中的环境变化,提出了一种对差分向量进行角度修正以及分级多种群协同进化(Angle Correction and Hierarchical Multi-Population,ACHMP)的进化算法.根据历史信息,使用无迹卡尔曼滤波模型来预测种群的中心点,通过不同时刻的中心点产生不同的差分向量,再使用无迹卡尔曼滤波对差分向量进行角度修正;提出的多种群协同进化模式将种群分为三部分并使其沿不同的方向进化,子种群监督主种群进化,在提升了算法性能的同时,也保证了种群的多样性.与10种对比算法在不同测试问题上的实验结果显示,ACHMP算法的性能总体优于其他算法,证明了本文提出的角度修正和分级多种群方法在处理动态多目标优化问题时具有较强的竞争力.展开更多
Intelligent production is an important development direction in intelligent manufacturing,with intelligent factories playing a crucial role in promoting intelligent production.Flexible job shops,as the main form of in...Intelligent production is an important development direction in intelligent manufacturing,with intelligent factories playing a crucial role in promoting intelligent production.Flexible job shops,as the main form of intelligent factories,constantly face dynamic disturbances during the production process,including machine failures and urgent orders.This paper discusses the basic models and research methods of job shop scheduling,emphasizing the important role of dynamic job shop scheduling and its response schemes in future research.A multi-objective flexible job shop dynamic scheduling mathematical model is established,highlighting its complex and multi-constraint characteristics under different interferences.A classification discussion is conducted on the dynamic response methods and optimization objectives under machine failures,emergency orders,fuzzy completion times,and mixed dynamic events.The development process of traditional scheduling rules and intelligent methods in dynamic scheduling are also analyzed.Finally,based on the current development status of job shop scheduling and the requirements of intelligent manufacturing,the future development trends of dynamic scheduling in flexible job shops are proposed.展开更多
文摘动态多目标优化问题具有多个相互冲突的目标,而且这些目标也受环境的影响不断变化,为了快速准确跟踪不断变化的Pareto前沿和Pareto解集,提出一种基于迁移学习的拐点预测策略(a knee points prediction strategy based on transfer learning, TKPS)。TKPS根据记忆过去时刻种群中优秀个体,使用迁移学习算法得到映射矩阵W,然后通过映射矩阵W,把当前时刻的拐点集映射到高维希尔伯特空间,从中找到下一时刻的拐点集,引导种群收敛;同时,在拐点的邻域内选出若干个伴随个体,增加种群多样性,避免种群陷入局部最优。TKPS采用8个测试函数,并与其它3个算法结果对比分析,实验结果表明TKPS算法具有更快的响应环境变化的能力。
文摘为更好地应对动态多目标优化中的环境变化,提出了一种对差分向量进行角度修正以及分级多种群协同进化(Angle Correction and Hierarchical Multi-Population,ACHMP)的进化算法.根据历史信息,使用无迹卡尔曼滤波模型来预测种群的中心点,通过不同时刻的中心点产生不同的差分向量,再使用无迹卡尔曼滤波对差分向量进行角度修正;提出的多种群协同进化模式将种群分为三部分并使其沿不同的方向进化,子种群监督主种群进化,在提升了算法性能的同时,也保证了种群的多样性.与10种对比算法在不同测试问题上的实验结果显示,ACHMP算法的性能总体优于其他算法,证明了本文提出的角度修正和分级多种群方法在处理动态多目标优化问题时具有较强的竞争力.
基金supported by the National Key Research and Development Program Project(No.2021YFB3301300).
文摘Intelligent production is an important development direction in intelligent manufacturing,with intelligent factories playing a crucial role in promoting intelligent production.Flexible job shops,as the main form of intelligent factories,constantly face dynamic disturbances during the production process,including machine failures and urgent orders.This paper discusses the basic models and research methods of job shop scheduling,emphasizing the important role of dynamic job shop scheduling and its response schemes in future research.A multi-objective flexible job shop dynamic scheduling mathematical model is established,highlighting its complex and multi-constraint characteristics under different interferences.A classification discussion is conducted on the dynamic response methods and optimization objectives under machine failures,emergency orders,fuzzy completion times,and mixed dynamic events.The development process of traditional scheduling rules and intelligent methods in dynamic scheduling are also analyzed.Finally,based on the current development status of job shop scheduling and the requirements of intelligent manufacturing,the future development trends of dynamic scheduling in flexible job shops are proposed.
