基于Hopfield神经网络的作业车间生产调度方法

该文提出了基于Hopfield神经网络的作业车间生产调度的新方法.文中给出了作业车 间生产调度问题(JSP)的约束条件及其换位矩阵表示,提出了新的包括所有约束条件的计算能 量函数表达式,得到相应的作业车间调度问题的Hopfield神经网络结构与权值解析表达式,并 提出相应的Hopfield神经网络作业车间调度方法.为了避免Hopfield神经网络容易收敛到局部 极小,从而产生非法调度解的缺点,将模拟退火算法应用于Hopfield神经网络求解,使Hopfield 神经网络收敛到计算能量函数的最小值0,从而保证神经网络输出是一个可行调度方案.该文 改进了已有文献中提出的作业调度问题的Hopfield神经网络方法,与已有算法相比,能够保证 神经网络稳态输出为可行的作业车间调度方案.

HOPFIELD NEURAL NETWORK APPROACH FOR JOB-SHOP SCHEDULING PROBLEMS

A new Hopfield neural network approach for job shop scheduling problems(JSP) is proposed. All constraints of job shop scheduling problems and its permutation matrix expression are proposed. A new computational energy function including all constraints of job shop scheduling problem is given. A corresponding new Hopfield neural network construction and its weights of job shop scheduling problem are given. To avoid the Hopfield neural network convergence to a local minimum to produce non feasible scheduling for JSP, the simulated annealing algorithm is applied to the Hopfield neural network and the network converges to a minimum volume 0, making the steady outputs of the neural network as feasible solution for job shop scheduling problem. Compared with the existing methods, our modified method can keep the steady outputs of neural networks as feasible solution for job shop scheduling problem.