基于小波包和支持向量机的结构有限元模型修正

The finite element model updating of structure based on wavelet packet analysis and support vector machines

提出了基于小波包变换和支持向量机(SVM)原理对大型空间网架结构有限元模型进行修正的方法.首先建立半刚性节点固结系数表示的空间网架单元刚度矩阵,通过对结构加速度信号进行小波包分析,提取小波包信号成分节点能量作为SVM模型的训练样本,建立SVM模型.其次由训练好的SVM模型,识别出节点的固结系数.最后将得到的节点固结系数代入空间网架结构的单元刚度矩阵,再由坐标变换得到修正后的结构总的刚度矩阵,从而实现网架结构的有限元模型修正.数值仿真结果表明该方法能对空间网架结构的有限元模型进行有效修正.

The finite element （FE） model updating of spatial truss structure based on wavelet packet （WP）analysis and support vector machines （SVM） is developed in this paper. Firstly, the element stiffness matrix expressed by the semi-rigid node with fixity coefficient is established. The calculated structure acceleration response signal is decomposed by WP, and then a proper SVM model is established by training samples which are obtained by extracting WP signal component node energy. Secondly, the WP component node energy achieved by the measured signal is input into the trained SVM model to identify the fixity coefficient, which is substituted to update the element stiffness matrix. Finally, the total stiffness matrix is obtained by the coordinate transformation of the element stiffness matrix, and the FE model of the structure is updated. The simulation result has confirmed that the proposed method is validity.