A STUDY OF DENOISING METHOD FOR EIGENSYSTEM REALIZATION ALGORITHM BASED ON WAVELET ANALYSIS

ERA/DC (eingensystem realization algorithm using data correlation) adopts the data correlation approach when constructing the Hankel matrix with impulse response data. Therefore, ERA/DC can restrain the white noise in the signal, but the practical data also contain other kinds of noise, such as random noise, impulsive noise, and excursion. In order to get better identification results, the impulse response data is denoised using the wavelet analysis method. Then, the ERA/DC is adopted to identify the denoised data. Simulations of a two-degree-of-freedom mass-spring-dashpot system and a satellite model are demonstrated. The results show that the denosing method based on wavelet analysis can suppress different kinds of noise, and improve the identification precision of ERA/DC. Therefore, it can be used as the denoising method for ERA/DC.