采用RBF网络预测含氢TC4合金的高温流变应力

Prediction of flow stress in TC4 titanium alloy with hydrogen at high temperature by a radial basis function neural network

研究了含氢TC4合金的热变形行为,基于径向基函数(RBF)人工神经网络建立了含氢TC4合金热变形流变应力的预测模型,该模型的样本数据取自热压缩试验数据,模型的输入量为变形温度、应变速率、应变量和氢含量,输出量为流变应力.研究表明:随着变形温度的升高和应变速率的降低,合金的流变应力降低;随着氢含量的增多,流变应力先降低后升高;RBF网络有较好的非线性逼近能力,训练相关性系数为0.999,训练速度较快,网络测试结果的最大相对误差为11.8%.

The hot deformation behaviour of TCA titanium alloy hydrogenated was studied and the radial basis function （RBF） neural network （NN） model on flow stress of TCA alloy with hydrogen under hot deformation was established. The results showed that the flow stresses of alloy decreased with increasing the deformation temperature and decreasing the strain rate, and they dropped first and then rose with enhancing the content of hydrogen. The RBF network had good nonlinear approach abihty. The training correlation coefficient of the model was 0. 999, the convergence rate was fast, and the maximum relative error of network was 11.8%.