Hopfield neural networks approach for job shop scheduling problems

A new method based on Hopfield neural networks for solving job-shop scheduling problems (JSP) is proposed. All constraints of job-shop scheduling problems and its permutation matrix express are developed. A new calculation energy function included all constraints of job-shop scheduling problems is given. A corresponding new Hopfield neural network construction and its weights of job-shop scheduling problems are given. To avoid Hopfield neural network to converge to local minimum volume, and to produce some non-feasible scheduling solutions for JSP, simulated annealing algorithm is applied to Hopfield neural network. Hopfield neural network converging to minimum volume 0, can keep the steady outputs of neural networks as feasible solution for job-shop scheduling problem. This paper improved existing method based on Hopfield neural network for solving job-shop scheduling problems. Compared with the method, modified method can keep the steady outputs of neural networks as feasible solutions for job-shop scheduling problems.