摘要
本构关系体现了材料在热态塑性加工过程中对热力参数的动态响应,关系到有限元模拟的准确性与精度。文章以TA15钛合金等温压缩实验数据为基础,构造一个3×10×10×1四层BP神经网络结构形式的本构关系模型,采用Bayesian规则化调整法训练网络以提高网络的泛化能力。预测结果、外推结果和实验结果对比表明,利用Bayesian规则化调整法训练的BP神经网络结构形式的TA15钛合金本构关系能够描述其高温变形力学行为,适用于热变形过程的数值模拟。
Constitutive relationship incarnates the material dynamic response to thermal parameters in hot plastic working process, which is related to FEM simulation accuracy and precision. Based on TA15 alloy isothermal compressive experiment data, Constitutive relationship is constructed in the shape of 3× 10× 10× 1 BP artificial neural network. The networks is trained using Bayes- ian regularization adjustment method in order to improve the generalization capacity. The comparison of the predicted, extrapolation and experimental results show that the TA15 alloy constitutive relationship, constructed in the shape of BP artificial neural network which is trained by Bayesian regularization adjustment method, can describe the TA15 alloy hot deformation mechanical behavior and be the same with the numerical simulation in hot deforming process.
出处
《塑性工程学报》
EI
CAS
CSCD
北大核心
2007年第4期101-104,132,共5页
Journal of Plasticity Engineering