论文部分内容阅读
用神经网络模型代替传统的数学模型,达到提高轧制参数预报精度的目的。在分析了轧制原理的基础上设计了神经网络冷连轧参数预报模型,并针对前向网络反向传播算法(BP)收敛速度缓慢和易陷入局部极小点的缺点,将有全局寻优特性的模拟退火算法(SA)与之结合得到具有较快收敛速度和较高逼近精度的神经网络轧制参数预报模型,提高了网络的快速性和精确性。最后以轧制力预报为例,证明了该方法收敛速度快,稳定性好,可信度高,具有较好的应用前景。
The neural network model instead of the traditional mathematical model, to achieve the purpose of improving the accuracy of rolling parameters forecast. Based on the analysis of rolling principle, the parameter prediction model of cold tandem rolling based on neural network was designed. In view of the shortcoming of BP convergence and slow convergence, it will have global optimization The simulated annealing algorithm (SA) is combined with it to get the prediction model of the rolling parameter of neural network with faster convergence rate and higher approximation accuracy, which improves the speed and accuracy of the network. Finally, taking the prediction of rolling force as an example, it is proved that this method has the advantages of fast convergence, good stability, high reliability and good application prospect.