人工神经网络在金属塑性本构建模中的应用  

Application of artificial neural networks in metal plasticity constitutive modeling

作　　者：张韩旭 方刚[1] Zhang Hanxu;Fang Gang(Department of Mechanical Engineering,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学机械工程系,北京100084

出　　处：《锻压技术》2024年第7期1-18,共18页Forging & Stamping Technology

基　　金：国家自然科学基金资助项目(52075288)。

摘　　要：从宏观和细观两个层面回顾了金属塑性本构模型,并指出模型应用的限制主要集中在本构参数的获取和模型的数值实现上。介绍了3种广泛应用的人工神经网络:反向传播神经网络、卷积神经网络和循环神经网络。从模型的标定和计算两个方面总结了人工神经网络在塑性本构建模中的应用。此外,介绍了基于物理信息的神经网络,该类模型基于传统物理理论在人工神经网络中引入约束,能够有效提高训练效率和泛化能力。最后,指出了人工神经网络应用在金属塑性本构建模中面临的挑战和未来发展方向。The metal plasticity constitutive models were reviewed from macroscopic and microscopic perspectives,and the limitations of the model application were mainly focused on the acquisition of constitutive parameters and the numerical realization of mdel.Then,three widely used artificial neural networks,namely,back propagation neural networks,convolutional neural networks and recurrent neural networks were introduced,and the applications of artificial neural networks in plastic constitutive modeling were summarized from two aspects of calibration and calculation of the model.Furthermore,physics-informed neural networks were introduced,and the training efficiency and generalization ability were effectively improved by taking constraints into the artificial neural network based on the traditional physical theories.Finally,the challenges and future development directions of the application of artificial neural networks in metal plasticity constitutive modeling were indicated.

关 键 词：塑性本构模型 唯象模型 晶体塑性 人工神经网络 训练效率 泛化能力 

分 类 号：TG301[金属学及工艺—金属压力加工]