基于神经网络的Job-Shop类调度问题

A Job-Shop scheduling solution based on neural network

提出了将Hopfield神经网络与模拟退火相结合以求解Job-Shop类调度问题的算法。该算法给出了Job-Shop类调度问题的约束条件,并且直接把问题的各种约束条件表示为Hopfield神经网络的能量函数项。为避免Hopfield神经网络容易收敛到局部极小解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神经网络收敛到能量函数的全局最优解,从而保证神经网络输出是一个可行的调度方案。通过仿真实例验证了该算法的可行性。

Based on Hopfield neural network and simulated annealing, an algorithm for the Job-Shop scheduling problem was proposed. This algorithm put forward the constraint of the Job-Shop scheduling, then formulates the energy function of Hopfield neural network according to constraints directly. Lest the network converged into the local minimum of the function the simulated annealing was applied. Thus the global minimum of the Job-Shop scheduling problem as a feasible scheme appears. The results of simulation calculation show that this algorithm is efficient.