Control-Relevant System Identification By Using Wavelet Analysis

点击翻译按钮获取中文摘要

1.We studied a condition for getting the estimate of transfer function from the ratio of wavelet spectra of input/output data.2.In connection with 1, we proposed an "indirect method" based on wavelet spectra and a "direct method" based on wavelet transform of input/output data, and compared two methods. Consequently, we showed that desired filter characteristics (e.g.LPF) are realized with some degrees of freedom.3.In connection with 2, we showed that when adopting a complex wavelet such as Gabor wavelet, the bias of estimate of spectrum is suppressed.4.Through the identification of control systems of a thermal power plant and the identification of control systems for Nox decomposition, we verified the feasibility and effectiveness of our approach.5.For the case that there is a stable transfer function with dead time in the forward loop and also there is a feedback loop, we considered a dead-time estimation problem. Consequently, we were able to prove that the previous result is extensible to this practical case.6.We showed the relation between tire sounds and tread patterns by using the wavelet-based correlation analysis.7.Applying the wavelet transform to the blind separation of sound source for a convolutive mixture case, we made the previous methods applicable to the convolutive mixture case by transforming time-frequency dependent signals into temporal signals by fixing the frequency