参数灵敏度分析的有限元-神经网络混合法

THE FEM-NEURAL NETWORK METHOD OF PARAMETER SENSITIVITY ANALYSIS

为了在力学参数位移反分析中进行力学参数灵敏度的分析 ,将神经网络方法与有限元正分析有机结合 ,利用有限元正分析获取神经网络的学习样本 ,利用神经网络输出变量对输入变量的偏导数求解力学参数在每一个测点的灵敏度 ,提出了一种力学参数灵敏度计算的新方法———有限元 -神经网络混合法。

To analyze the parameter Sensitivity in the mechanics parameters back analysis, a new method--the FEM-neural network method--in which the FEM and neural network are combined is putted forward. In this method, the studied samples which are utilized by neural network are acquired with FEM, and the parameters sensitivity of very measured point is solved by the neural network output variables derivative about input variables.