摘要
采用基于退火策略的混沌神经网络求解作业车间调度问题(JSP)。通过在Hopfield神经网络(HNN)中引入混沌机制,并结合退火策略控制混沌动态,有效避免了传统HNN极易陷入局部极小的缺陷;同时改进了表示JSP的换位矩阵,给出了包含目标函数的能量函数,保证了网络的稳态输出为全局可行解。与基于模拟退火的HNN和暂态混沌神经网络算法比较表明,该算法在收敛速度和求解准确性上有了很大改进。
Chaotic neural network based on simulated annealing （ACNN） method for job-shop scheduling problem was proposed. ACNN would not be stuck into local minima by introducing the chaos mechanism in Hopfield neural network, which utilized the annealing strategy to control chaos dynamics. Permutation matrix of job-shop scheduling problem was improved. An improved energy function with objective function was given, which guaranteed the network steady-state output for the overall situation feasible solution. The results obtained by the proposed algorithm were compared with those obtained by Hopfield neural network based on simulated annealing and transient chaotic neural network, it is shown that the proposed algorithm greatly improves optimization convergence speed and the accuracy of solution.
出处
《机床与液压》
北大核心
2009年第7期11-14,20,共5页
Machine Tool & Hydraulics
基金
河北自然科学基金资助项目(F2008000861)
关键词
组合优化
混沌神经网络
作业车间调度
能量函数
Combinatorial optimization
Chaotic neural network
Job-shop scheduling
Energy function
