基于反向传播神经网络的陶瓷损伤参数反演分析

Reversion Analysis of Ceramic Damage Based on Back Propagation Neural Network

陶瓷材料冲击加载条件下的损伤累积过程伴随着裂纹扩展、体积膨胀等因素,为准确获取陶瓷损伤参数,以钨合金球弹丸高速撞击陶瓷复合装甲的侵彻深度实验为基础,获得了陶瓷面板破碎情况。根据现有文献[14]数据中陶瓷材料JH-II本构模型的损伤参数范围,确定了反向传播(BP)神经网络的样本点;采用有限元分析软件AUTODYN对所有样本点的侵彻过程进行数值模拟,结合仿真和实验数据完成了BP神经网络模型的建立和Ti B2-B4C复合材料损伤参数的反演。仿真结果和实验侵彻深度、回收陶瓷面板的损伤比对,充分验证了所建立的BP神经网络模型对陶瓷损伤参数反演的有效性。

Damage accumulation of ceramic material is accompanied by crack propagation and bulking under shock loading.In order to obtain the accurate damage parameters of ceramic,the tungsten alloy ball projectiles were used to penetrate into a ceramic composite armor at high speed,and the depth of penetration and the fracture degree of ceramic plate were gained.The sample points of back propagation neural network are determined according to the damage parameters of JH-II constitutive model in Ref.[14].The process of penetration into all sample points is numerically simulated by using the finite element analysis software AUTODYN.The establishment of BP neural network and the damage inversion of TiB2-B4C composites are accomplished by simulation and experimental data.The validity of BP neural network model for damage inversion is verified by comparing the simulated and experimental penetrating depths and fractures of recovered ceramic plates.