Operational modal parameter identification based on PCA-CWT

Abstract

A new continuous wavelet transform (CWT) based modal identification method is proposed to apply to ambient vibration tests. Firstly, to improve computational efficiency, perform singular value decomposition on positive power spectrum matrix, determine system's order and corresponding frequency range according to the extreme values of the singular values; reduce data involved in analysis by principal component analysis (PCA); decompose the reduced data into the time-scale domain using CWT within the corresponding frequency range. Secondly, to locate the wavelet ridges of the covariance wavelet coefficient, PCA of the covariance matrix of multi-measurements is used to extract the ridges of covariance wavelet coefficient. Estimate eigenfrequencies and damping ratios with the obtained ridges while the ratios of cross-covariance wavelet coefficients to auto-covariance wavelet coefficients are used to determine mode shapes. Both real measurement data and simulation are adopted to illustrate the proposed method's practicability.