黄铜/铝双金属轧后厚比预测的神经网络模型

采用正交实验方法，对不同原始厚度、不同原始厚比和压下率的复合板进行轧制实验，发现不同的轧制工艺条件对成品厚比影响较大。对试验数据进行了神经网络建模。训练发现，在目标函数为0．05、隐层节点数为5、学习率为0．1时，系统误差较小。利用所建立的网络模型对其它8组不同实验条件下的成品厚比进行了预测，发现预测数据与实验数据吻合良好(总拟合度为8．7％)。说明神经网络模型很适合拟合轧制工艺参数和双金属板厚度比之间的非线性关系。

An Artificial Neural Network Model for Prediction of Thickness Ratio of the Rolled Al/Cu Sheet

The thickness ratio of aluminum and copper has a very important effect on the property of Al/Cu clad sheet. There are very complex nonlinear relations between the processing parameters (initial thickness ratio, initial total thickness, rolling reduction) and the thickness ratio of the clad sheet. An artificial neural network model was established based on the orthogonal experimental data. The results show that the systematic error is small with the objective function of 0.05, the number of nodes of 5 in the hidden layer and a learning-rate of 0.1. The total fitness is 8.7%. The results also show that neural network is a good tool to fit the complex data got from nonlinear relation between the parameters of the rolling process and the thickness ratio of the clad sheet.