The Research On Fault Diagnosis Method Of Rotor System Based On The Information Entropy Of Chaos Attractor Invariants

Rotating machinery is the most comprehensive industrial equipment, whose corecomponent is rotor-bearing system. It is an important method to ensure the stable andeffective thread of rotating machinery that applies the vibral signals to achievereal-time mornitoring, analysis and diagosis of the working process of rotor-bearingsystem. However, the elevation of revolution and emergence of new material andconstruction has brought obvious dissatisfaction of traditional linear theories, andconstantly been improving the role of nonlinear theories and methods in rotor faultdiagnosis. But, the existing achievements still has defects, such as no scientificnonlinear features for fault identification and no precise quatized evaluation of signals,which decreases the diagnosis accuracy at a certain extent. Aiming at the rotordynamic model with representative faults, I researches on the quatized extraction ofthe nonlinear feature of fault vibral signals with chaos theory and information entropymethod. The main contents and conclusions are as follows:(1) Through the creation and calculation of mathematic model of rotor-bearingsystem with different slowing-down fault parameters, the dynamic behavior analysis ofrepresentative faults based on chaos theory is accomplished, and the dynamic responsedivergences of four fauls in different revolution fields is generalized, which is able tobe the necessary features for rotor faults identification for their conformity in theprogress of system faults slowing-down change, making the proof for the rationally ofthe methods introduced in my research.(2) According to the demands for the processing of rotor fault nonlinear vibralsignals, the improved local projective algorithm is designed. The adaptive parameterselection link of optimum neighborhood diameter and noise subspace dimension isadded into normal algorithm, in order to completely preserve the chaotic signals underthe intensive noise background. The experimental application on rotor testing bed datahas achieved satisfied results.(3) In view of the actual complex system which is incapable of modeling, themethod which calculates correlation dimension and maximum Lyapunov indexaccording to revolution of different faults is applied to reflect the dynamic behaviorsof rotor systems. And the analysis of simulating sequences and experimental data of fault rotor is applied to verify the effect of that method. The result shows that thechaos attractor invariants is able to accruatly reflect the condition and level of rotornonlinear movements, which is the effective features of nonlinear characteristicidentification of actual complex system.(4) In the first-order crtitcal speed field and high frequency field, where thenonlinear features of rotor faults vary more greatly, the information entropy quatizedextraction of attractor invariants is accomplished. And the entropy band and entropycomponent spectrum are programmed as the reason of identifying the fourrepresentative faults. Meanwhile, the method that reflects the degrees of faultseverity with entropy component spectrum is researched, which provides a new imagebasis for rotor-bearing system fault diagnosis adopting nonlinear theories.The nonlinear dynamic characteristics of mechanical system, the quantization ofthe extraction as a new trend of the mechanical fault diagnosis, with a great researchspace and research value, but also for intelligent fault diagnosis, namely the datadriven technology in fault diagnosis in application to lay the foundation.