基于BP神经网络的镁合金晶粒尺寸及流变应力模型

Modelling for grain size and flow stress of magnesium alloy based on BP neural network

通过等温压缩试验研究了热变形条件对AZ61B镁合金再结晶晶粒尺寸及其流变应力的影响,并采用人工神经网络方法分别建立了动态再结晶晶粒尺寸及流变应力的模型。利用反向传播算法对网络进行试探性训练研究,得到了最佳参数。结果表明,所建立的网络模型具有优良的性能,能精确预测AZ61B合金热变形条件下的再结晶晶粒尺寸及流变应力。

The effect of processing parameters on grain size and flow stress of AZ61B magnesium alloy have been investigated by isothermal compression, both models for dynamic recrystallization grain size and flow stress were constructed by applying artificial neural network. The experimental research on the networks have been made by backpropagation arithmetic and some optimal parameters are obtained. The result indicates that models are good enough to exactly predict recrystallization grain size and flow stress of AZ61B magnesium alloy under the condition of thermal deformation.