基于LS-SVM的模态参数识别方法

LS-SVM-based Method for Modal Parameter Identification

将最小二乘支持向量机回归用于系统的模态参数识别研究。针对经典的最小二乘支持向量回归缺少鲁棒性和稀疏性的缺陷,提出了一种兼具鲁棒性和稀疏性的最小二乘支持向量回归的算法,并保持了它原有的计算速度快的优点。最后,结合结构动力学方程的自回归滑动平均时间序列形式,给出了结构的模态参数提取方法和流程,给出了相应的数值算例以及进行了实验的检验证明。结果表明,本文的方法能够快速、准确地提取出系统的模态参数。

A least-squares support vector regression （LS-SVR） technique is applied to modal parameter identification in this article. While the present least squares support vector machines （LS-SVM） exhibit two natural drawbacks of insufficient robustness and sparseness, a novel algorithm that can overcome these drawbacks is proposed. An LS-SVM-based method employing the auto regression moving average （ARMA） time series is presented for linear structural parameter identification using the observed vibration data. Both numerical evaluation and experimental validation demonstrate that the LS-SVM-based method identifies structural modal parameters accurately and quickly.