基于KPCA的结构振动信号特征提取研究

Feature Extraction of Structural Vibration Signal Using KPCA

本文提出基于核主元分析(Kernel Principal Component Analysis,KPCA)的结构振动信号特征提取方法。通过对原始信号进行KPCA分析得到非线性主元,根据非线性主元的累计贡献率确定非线性主元个数,然后根据结构信号在不同损伤状态下的KPCA特征构造结构特征指标,由此判断结构相对于基准状态是否发生了损伤。试验结果表明,该方法在基准状态为无损状态和小损伤状态下都能很好的判断结构是否发生损伤,用于结构损伤识别特征提取是有效的。

A feature extraction method of structural vibration signals based on the Kernel Principal Component Analysis （KPCA） is proposed in this paper. The number of nonlinear principal components is confirmed according to the cumulative contribution rate of the nonlinear principal components which were obtained by using of KPCA on original data, and then the structural characteristic indicators are constructed by KPCA character- istics of the signals which are collected at different structure damage status. Finally, according to the structural characteristic indicators, it is judged whether there has been a serious structural damage compared with the reference state. The experimental results show that structure damage happened can be judged by the proposed method, regardless of the reference state which is non-destructive or smaller injury. And the method is effective on extract the feature of structural damage identification.