Working modal identification method based on time-frequency domain single-source-point sparse component analysis

Abstract

The invention provides a working modal identification method based on time-frequency domain single-source-point sparse component analysis. The working modal identification method specifically includes the following steps that vibration signals of an equipment target position under the working state are obtained through measurement; time-frequency domain conversion is conducted on the mixed vibration signals; a single-source-point method is used for extracting the mixed vibration signals used for estimating a hybrid matrix in a time-frequency domain; a hybrid matrix estimation method based on K hyperline clustering sparse component analysis is used for estimating the hybrid matrix; after the hybrid matrix is solved, the time-frequency domain is returned, the l1 minimization method is used for reconstructing each order of source signals, and modal vectors of a structure are extracted; then the working modal frequency and the damping ratio are obtained through signal index expression. According to the method, the calculation amount in the hybrid matrix estimation process is reduced, under the poor condition that the number of measuring points is less than that of the source signals, modal parameters are identified effectively, and the method has good anti-interference capacity for incomplete sparsity of noise, abnormal values and the source signals.

Description

technical field [0001] The invention relates to a working mode identification method based on time-frequency domain single source point sparse component analysis, which belongs to the field of working mode identification. Background technique [0002] At present, modal analysis technology has become an important means to identify the characteristics of dynamical systems. Through modal analysis, important dynamic properties such as natural frequencies, damping ratios, and mode shapes of the system can be extracted. Commonly used modal parameter extraction methods include experimental modal analysis and working modal analysis. Experimental modal analysis requires excitation of the structure, which is very difficult for large complex mechanical systems. The working modal analysis method does not require external excitation, and only relies on the vibration response signal of the mechanical structure to extract the structural modal parameters. This analysis method has been po...

AI Technical Summary

Problems solved by technology

However, this application only uses the frequency-domain sparsity of the signal in the mode separation process, ignoring the time-domain characteristics of the signal. At the same time, there are a large number of invalid cluster points in the cluster analysis process, and the actual engineering application requires a large amount of calculation and is easy to disturbed by noise

Estimation methods of the mixing matrix in the process of sparse component analysis include K-means, fuzzy C clustering, linear geometric independent component analysis, etc., such as references (Wang Xiang, Huang Zhitao, Ren Xiaotian, Zhou Yu, based on time-frequency single Underdetermined mixed blind identification algorithm for verification technology, Journal of National University of Defense Technology, 35(2), 2013) gives the process of using K-means clustering to obtain a mixed matrix after single-source point detection, but this method requires sufficient sparsity of the source signal , noise and incompletely sparse source signals will seriously affect the practical application of this method. In addition, the estimation process does not clearly show the source signal scatter diagram after single source point detection

[0005] In view of the above problems, the purpose of this method is to solve the underdetermined modal parameter identification problem in which the number of sensors measuring vibration signals is less than the number of source signals in actual working modal analysis, and make full use of the sparseness of source signals in both time domain and frequency domain It can reduce the calculation amount of the mixing matrix estimation process, separate the effective components in the source signal, and overcome the existing clustering method for estimating the mixing matrix that has too strong requirements on the sparsity of the source signal, and the incomplete sparseness of noise, outliers and source signals Good anti-interference ability

Images