基于神经网络的贮仓结构参数识别  

PARAMETERS IDENTIFICATION OF SILO USING NEURAL NETWORK

作　　者：董清华[1] 黄义[1] 

机构地区：[1]西安建筑科技大学土木工程学院

出　　处：《振动与冲击》2006年第2期83-87,共5页Journal of Vibration and Shock

摘　　要：利用神经网络技术,提出了识别结构物理参数的一种方法。用单元刚度矩阵基本值和模态应变能来选择基本模态,用修正的Latin超立方采样技术和模态准入准则来产生网络的输入数据。贮仓在动载作用下的自振频率和模态作为网络的输入,子矩阵参与系数作为网络的输出,用Levenberg-Marquardt算法训练网络。仿真计算表明,方法是可行的。A method to identify the physical parameters of structure system has been developed by using neural network . Element-stiffness matrix baseline parameters and modal strain energy are employed for the selection of the base modes. An updated-Latin hypercube sampling and modal assurance criteria are adopted for efficient generation of the patterns for training the neural network. The neural network is composed in which the input signals are the silo natural frequencies and its mode shapes ,and output signals are the submatrix scaling factor . The Levenberg-Marquardt algorithm is applied to modify the weight matrices of neural network. Results from computer simulation studies show that the method is valid and feasible.

关 键 词：参数识别 神经网络 修正的Latin超立方采样技术 模态应变能 

分 类 号：O175.3[理学—数学] TU352.11[理学—基础数学]