基于BP神经网络的管材材料参数逆向识别

A new method based on back propagation neural network for identifying the material parameters of tube

管材材料塑性本构参数是研究管材弯曲成形的关键因素之一。对于大直径管材的力学性能,可通过取样拉伸试验测得材料的应力应变曲线与弹性模量;对于小直径管材的材料参数,则较难直接通过实验测得。该文利用ABAQUS有限元软件,对小直径厚壁管材绕弯成形及回弹过程进行数值模拟,提出基于BP神经网络算法与数值模拟仿真实验相结合的管材材料参数逆向识别的方法,实验数据的对比表明,该方法能够有效的预测管材材料参数。

Plastic constitutive parameters of tube are one of key factors for studying the forming qualities during tube bending.The stress-strain curve and elastic modulus of large-diameter tube can be measured by the tensile test.However,it is difficult to measure the material parameters of small-diameter thick-walled tube directly by experiment.Using finite element software ABAQUS,the bending and springback processes of small-diameter thick-walled tube were simulated.A new method combining back propagation neural network and numerical simulation experiments was presented for parameters identification.Comparison between the analyzing results and the experimental data show that the new method can be effectively used for predicting the material parameters of tube.