论文部分内容阅读
根据182组实测焊缝金属奥氏体分解温度值,分别采用线性回归方法、非线性回归方法和人工神经网络技术建立了奥氏体分解温度的预测公式或模型。结果表明:线性回归公式难以准确体现各因素与奥氏体分解温度之间的关系,引入Mo指数和ln(t8/3)函数,预测精度有所提高;考虑了各因素之间交互作用的神经网络模型预测精度高于线性和非线性回归公式的,更适合于奥氏体分解温度预测研究。
According to the 182 sets of measured austenite decomposition temperature of the weld metal, the prediction formula or model of austenite decomposition temperature was established by linear regression method, nonlinear regression method and artificial neural network respectively. The results show that the linear regression formula can not exactly reflect the relationship between each factor and the austenite decomposition temperature. The Mo index and ln (t8 / 3) function are introduced to improve the prediction accuracy. Considering the interaction between the various factors The prediction accuracy of network model is higher than that of linear and non-linear regression formulas, which is more suitable for the prediction of austenite decomposition temperature.