基于RBF神经网络的马氏体开始转变温度预测研究

Study on Martensite Start Temperature Prediction Based on RBF Neural Networks

分析马氏体转变温度的影响因素,基于RBF神经网络建立马氏体开始转变温度预测模型,对其训练至稳定,预测钢的马氏体开始转变温度。与经验公式计算结果对比,基于RBF神经网络的马氏体开始转变温度预测模型具有较高预测精度。对4种钢的合金元素进行定量分析,结果表明:增加C含量能降低马氏体开始转变温度;马氏体开始转变温度与C、Si、Mn、Cr、Ni和Mo含量一般呈非线性关系。

Ms prediction model was established based on RBF neural network. The factors which can affect the martensitic transformation temperature were analyzed. The martensite starting the transition temperature was forecasted after the model was trained to stable steel. The results were contrasted with the empirical formula based on RBF neural network. Ms prediction model has higher prediction accuracy. The quantitative analysis of the alloying element of the four kinds of steels shows that increasing C content can significantly reduce Ms starting temperature, which are in non-linear relationship between Ms temperature and C, Si, Mn, Cr, Ni and Mo.