基于人工神经网络的微合金钢热轧奥氏体晶粒尺寸模型的研究

A Model for Austenite Gain Size Based on Artificial Neural Networks During Controlled Hot Rolling of Micro-alloyed Steel

基于人工神经网络原理，对微合金钢热轧控制参数的选取进行了研究．利用Gleeble－1500热力模拟机提取了轧制温度、应变量、应变速率和相应的应力应变曲线，并通过显微组织观察获取了实验后样品断面的奥氏体晶粒尺寸．通过归一化把实验所得数据进行处理．采用BP算法训练网络，对热轧控制参数（轧制温度、应变量、应变速率）和奥氏体晶粒尺寸之间的映射关系进行了函数逼近，建立了奥氏体晶粒尺寸神经网络模型．根据网络估测的结果可定量地进行热轧控制质量预报．

In actual production, it is important to predicts quality by generating a Model between hot rolling parameters and microstructures. In this paper, based on the method of Artificial Neural Networks, a new approach is provided to estimate hot rolling quality of the mico-alloyed steel. Firstly, a new means was developed, which obtained data of hot rolling. In the research, rolling temperature, strain, strain rate and the curve of stress and strain were gained by using Gleeble-1500 thermal mechanical machine and then grain sizes were measured.In addition, the data were deal with by ANN. In the research, the artificial neural networks originated BP arithmetic is built through training. Based on the result of networks＇estimating,the prediction of austenite grain size of microalloyed steel in hot rolling can realized and the mathematics model generated by ANN has higher accuracy than by regression method.