材料非线性粘弹性本构关系的神经网络模拟

SIMULATION OF NONLINEAR VISCOELASTIC CONSTITUTIVE RELATION OF MATERIALS BY NEURAL NETWORK METHOD

对Green Rivlin非线性粘弹本构方程给出了用一个三层时延前馈神经网络表示的等价关系 ,从而利用该网络可以求解G R方程中的各阶核函数 ,并给出了根据有机玻璃的一组实验数据计算核函数的应用实例 .由此指出神经网络方法可广泛应用于材料本构关系的研究 .

Firstly, it is proved that the Green-Rivlin nonlinear viscoelastic constitutive equation can be expressed by a Three-Layer Feedforward Network with Time Delays. Thereby, the network can solve the kernel functions in all orders of the Green-Rivlin equation. Then, an application of computing the kernel functions depending on a group of experimental data of PMMA is conducted,and it is pointed out that the Artificial Neural Network method can be widely used in the study of the constitutive relation of materials.