Dissertation > Excellent graduate degree dissertation topics show

Based on the time-frequency analysis of the mechanical failure of blind source separation method

Author: LvYaPing
Tutor: LiZhiNong
School: Zhengzhou University
Course: Mechanical and Electronic Engineering
Keywords: Time-frequency analysis Blind Source Separation Independent Component Analysis Fault Diagnosis Fractional Fourier transform Empirical Mode Decomposition Wigner-Ville distribution Ambiguity function Due to blind source separation
CLC: TH165.3
Type: Master's thesis
Year: 2009
Downloads: 138
Quote: 1
Read: Download Dissertation

Abstract

Most mechanical failures blind separation method is limited to non-Gaussian, stable and mutually independent source signals, and requires the observation signal than the number of source signals, tend to produce many of the problems in the machinery and equipment fault diagnosis, because the mechanical source signals often do not meet these assumptions. For this deficiency, the National Natural Science Foundation of China (No.: 50775208), the Education Department of Henan Natural Science Foundation of China (No.: 2006460005,2008 C460003) funded under the Cohen class frequency distribution of fractional Fourier transform and empirical mode decomposition as an example, combination of frequency distribution and blind source separation, in-depth study machinery and equipment based on time-frequency analysis of non-stationary signals blind source separation method, were compared with the traditional mechanical source separation method, the innovative achievements, The main contents are as follows: Chapter 1: This paper discusses the issues raised and its research significance, research status, blind source separation and blind source separation in the Fault Diagnosis Research. On the basis of insufficient analysis of the existing mechanical failure source separation, the main content of the paper and innovation. Chapter II: in layman's language and discusses the basic concepts of blind source separation, the two blind source separation inherent uncertainty, the amplitude and phase of uncertainty and sort of uncertainty, these two uncertainties objective existence does not affect the correctness of the results of the blind source separation. Blind source separation of some important concepts, such as principal component analysis, singular value decomposition, independent component analysis were introduced, and compared, pointed out their differences and connections. Finally, discusses three typical algorithms used in this thesis blind source separation: JADE algorithm, Infomax algorithm and FastICA algorithm, three algorithm calculation and steps. In addition, blind source separation performance indicators used in this article were introduced. Chapter 3: respective advantages for blind source separation methods based on mechanical failure ignore the lack of non-stationary signals, combined with the Cohen class time-frequency analysis and blind source separation, real-time frequency analysis is a powerful tool for processing non-stationary signals, describe its spectral characteristics change over time, blind source separation of highly relevant to a variety of signal aliasing separation. Time-frequency analysis based on Cohen's machinery and equipment blind separation of non-stationary signals, the existing blind source separation of mechanical fault diagnosis method extended to the Cohen class time-frequency distribution, with the signal time-frequency distribution, to achieve the purpose of separation of multiple faults in machinery and equipment. The same time, the proposed method with the traditional mechanical equipment non-stationary signals blind separation method were compared. Simulation results show that the proposed method is superior to the traditional mechanical failure blind source separation methods, machinery and equipment of non-stationary signals blind separation must take full advantage of non-stationary signals, in order to achieve good separation. The experimental results further validate the effectiveness of the method. The method is characterized as long as the source with a different time-frequency distribution, you can achieve effective separation. Finally, based on the quadratic time-frequency distribution of mechanical failure blind source separation, reflected by the root mean square error of the source signal separation. Chapter 4: mechanical source separation method based on time-frequency analysis is limited to the Cohen class time-frequency distribution, and not extended to other time-frequency distribution. Fractional Fourier transform as a new time-frequency analysis method, is the promotion of the classical Fourier transform. Both classical Fourier transform is a natural link, the Fourier transform do not have some of the characteristics. The fractional Fourier transform is a powerful tool to deal with non-stationary signals, combined with fractional Fourier transform and blind source separation, machinery and equipment based on fractional Fourier transform blind separation of non-stationary signals, the method first on whitening treatment observation signals obtained new observation signal, re-calculate a new observation signal FRFT a generalized correlation matrix, whereby an estimated generalized correlation matrix of the estimated approximate joint diagonalization, to thereby obtain the estimate of the source signal. The significant features of the method are: it does not have to assume that the energy of the signal varies with time, and it does not require a pretreatment stage in the time-frequency domain selection point. Finally, this approach is applied to the bearing inner ring fault blind separation Experimental results show that the method is effective. Chapter 5: based time-frequency analysis of mechanical failure source separation method requires observation signal than the lack of signal sources, combined with empirical mode decomposition (Empirical mode decomposition, EMD) and BSS, two mechanical failure source due to blind separation method, EMD-BSS method. In the proposed method, the use of the EMD method to decompose the mixed observed signals, all IMF component decomposed mixed observed signals and the original re-form a new observation signals, underdetermined BSS problem into overdetermined BSS problem. Then, a new observation signal whitening process and joint diagonalization to obtain estimates of the source signal. A notable feature of the method is not only able to handle the separation of stationary signals, but also to deal with the separation of non-stationary signals, another notable feature is applicable both in the number of sources than the separation of the number of observed signals, also applies to the number of sources is less than the observation the number of signal separation. The simulation results demonstrate the effectiveness of the method, and superior to the traditional mechanical failure blind source separation method. Finally, the proposed method is applied to the motor - gearbox coupling experiments further validate the effectiveness of the method. The same time, the combination of EMD and principal component analysis (PCA) their respective advantages, based on EMD-PCA of mechanical failure underdetermined blind source separation methods, the basic idea is similar to the EMD-BSS method. The difference is that the commonality analysis using PCA on the observed signals in EMD-PCA method to obtain the main component of the source signal. EMD-BSS method, it also has excellent characteristics. The bearing failure source separation experiments verify the effectiveness of the method. Chapter 6: full text content summary, and pointed out that warrant further research.

Related Dissertations

CLC: > Industrial Technology > Machinery and Instrument Industry > Machinery Manufacturing Technology > Flexible manufacturing systems and flexible manufacturing cell > Fault diagnosis and maintenance
© 2012 www.DissertationTopic.Net Mobile