Font Size: a A A

Fault Feature Extraction And State Prediction Of Complex Mechanical Systems Based On Chaos Time Series

Posted on:2005-12-11Degree:MasterType:Thesis
Country:ChinaCandidate:W A LiFull Text:PDF
GTID:2168360152966865Subject:Systems Engineering
Abstract/Summary:PDF Full Text Request
It is one of the primary methods to extract systemic eigenvalue from the time series examined on the spot and identify systemic running state or fault types from the eigenvalue. As known to all, for the existence of all kinds of nonlinear factors, the mechanical system is nonlinear inherently, and the vibration signals got from the complex system by sensor also shows nonlinearity usually. The present methods of eigenvalue extraction mainly concentrate on the frequence and chart linear analysis. Diagnosis practice shows that identification of the diagnosis with obvious nonlinear property (such as gear-rotor system) has strong limitations usually. Chaotic time series analysis provides a new approach to analyzing, examining and diagnosing dynamics system, and resolves problem according to the essence of chaos dynamics signals. If we can classify system conditions, identify system abnormal behavior and predict fault evolution by Chaotic time series analysis combined with the frequence and chart analysis method, this makes up for the limitations of traditional methods to low frequency and broad band fault signals without doubt. Furthermore we can acquire a much higher diagnosis ratio.Based on the fact, this work can be summarized as follows emphasis:1. The theory and methods of chaotic time series signals analysis is systematic- ally summarized with creativity according to chaotic dynamics property of complex mechanical system and needs of fault diagnosis.2. This paper puts forward a modified version according to the shortcomings of the traditional G-P algorithm such as consistency, real time, data stream, ability to resist noise etc. And signals simulation shows a better performance than before.3. This paper systematically summarizes the basic prediction methods of chaos time series and puts forward a modified phase trajectory prediction method according to the present methods. Digital simulation shows that it has a better prediction precise, higher real time and a longer prediction steps than those of Maximum Lyapunov exponent and Neural network method prediction method.4. This paper takes three kind of low frequency and broad band fault signals of a gear-rotor system(such as Pedestal looseness, Larger bearing clearance and Rub-impact failure) for instance, combined with the methods feature extraction and fault prediction given in the context, and extra identifies the three kind of faults which can't be classified with the traditional frequence and chart analysis method. Diagnosis experiment shows validity and practicability.
Keywords/Search Tags:Complex mechanical system, fault diagnosis, nonlinear, chaotic time series analysis, feature extraction, state prediction
PDF Full Text Request
Related items