Abstract:  In order to solve the flexible flow shop production scheduling optimization problem (flexible flow shop schedule problem FFSP), this paper proposes a Hopfield neural network algorithm based on the principle of simulated annealing as a global optimization algorithm. This algorithm puts forward the permutation matrix of FFSP problem, and gives the energy function expression of FFSP problem, and to overcome the standard hopfield neural network algorithm (hopfield neural networks HNN) in solving FFSP easy to fall into local minimum solution of defects, the simulated annealing algorithm is applied to the hopfield neural network to solve, ensure that the output can be dispatched when the energy function tends to be stable. Finally, the advanced HNN algorithm is tested by using examples of different scales, the advanced HNN algorithm is compared with the traditional genetic algorithm and compact genetic algorithm the experiment results show that the advanced HNN algorithm is an effective method of solving FFSP problem.

Key words: Flexible flow shop, neural network, simulate annealing algorithm, permutation matrix, energy function

摘要： 为了解决柔性流水车间中的排产优化问题（Flexible Flow Shop Schedule Problem，FFSP），提出了一种基于模拟退火原理的Hopfield神经网络算法作为全局优化算法。该算法提出了FFSP问题的换位矩阵，给出了FFSP问题的能量函数达式，并且为克服标准Hopfield神经网络算法（Hopfield Neural Networks，HNN）在解决FFSP问题时容易陷入局部最小解的缺陷，将模拟退火算法应用于Hopfield神经网络求解，确保当能量函数趋于稳定时输出可行调度解。最后，选用不同规模的实例对改进的HNN算法进行测试，并与遗传算法、紧致遗传算法、HNN算法进行对比研究，实验结果表明改进的HNN算法是求解FFSP问题的一种有效方法。

关键词: 柔性流水车间, 神经网络, 模拟退火法, 换位矩阵, 能量函数