基于快速独立分量分析的模态振型识别方法研究

摘要：快速、准确地识别出结构的模态参数，特别是结构的振型是结构损伤精确识别与健康监测的重要前提。大多的模态参数识别时域方法都是从曲线拟合的角度或解算特征值的过程来实现。振型向量通过求解各阶模态的留数获得，这些方法依赖于模态频率与模态阻尼的识别。本文提出一种模态振型的直接提取方法，该方法基于快速独立分量分析技术，以模态响应之间的独立性构造目标函数，通过优化目标函数寻求振型向量的最优解，直接从结构自由响应或脉冲响应的数据矩阵中提取结构的振型向量。三自由度数值算例表明该方法有效，具有很高的识别精度且对测量噪声具有很好的鲁棒性。

A modal identification method based on fast ICA

It's important to identify structural modal parameters, especially, accurate modal shapes in time domain for accurate damage identification and health monitoring of structures. Normally, lots of time domain modal identification methods are generally realized through curve-fitting or calculating eigenvalues. Modal shape matrix is obtained by calculat- ing various modal residues. These methods depend on identifying modal frequencies and modal damping ratios. Here, a novel method based on Fast ICA for modal shapes identification was proposed. This method used the independence be- tween modal responses to create an objective function, and obtained the optimal solutions of the modal shape matrix by op- timizing the objective function. It extracted the modal shapes directly from free responses or impulse responses of struc- tures. A numerical example of a three DOF system was performed and showed the efficiency of this method, which gave a high identification accuracy and a good rebustness to measuring noises.