一种Hopfield神经网络能量函数改进下的作业车间调度

Hopfield neural networks for Job-Shop scheduling problems based on improved energy function

由于作业车间调度问题的目标函数目前还无法用换位矩阵的元素以数学公式的形式表示,无法保证求出全局最优解。首先对换位矩阵表示方法进行了改进,给出新的带有目标函数的能量函数表达式,然后提出改进的Hopfield神经网络作业车间调度方法,并将模拟退火应用于Hopfield神经网络求解,避免了陷入局部极值。仿真结果表明,该方法具有全局搜索能力,并能够保证神经网络的稳态输出为全局最优或近似全局最优。

Because the objective function of Job-Shop scheduling problem（JSSP） can not be transposed matrix elements in the form of a mathematical formula,there is no guarantee that a global optimal solution.First,this paper improved the permutation matrix,gave a new energy function with objective function.Then it proposed the modified Hopfield neural network for JSSP,and applied the simulated annealing algorithm to the Hopfield neural network to avoid a local maximum.The simulation results show that the method has the ability of searching for the global minimum.Moreover,the method guarantees the steady output of the neural network is the global optimal or similar global optimum.