Nonlinear Fault Detection For Rotating Machinery

Rotating machinery's operating status is good or bad,it directly affects the performance of the system.The fault detection technology of rotating machinery are studied.The nonlinear degree is evaluated by the nonlinear characteristics of the vibration signal.The influence of signal decomposition on reducing the degree of nonlinearity and method by decomposed signal are discussed.The signal decomposition,chaos characteristics and hidden Markov theory are combined to construct the fault detection model for accurately judging rotating machinery.It is important to ensure the normal operation of the rotating machinery.The main contents are as following.The nonlinear characteristic of vibration signal is studied,and the two important parameters of embedding dimension and delay time are determined.The nonlinear evaluation index is studied by using chaos and fractal theory.The computational methods of maximum Lyapunov exponent,Kolmogorov entropy,correlation dimension and box dimension are given.The nonlinarity degree of is reduced by signal decomposition.The nonlinearity degree after wavelet decomposition and ensemble empirical mode decomposition is compared.A combination denoising method is proposed for vibration signal.The fault detection model of vibration signal is constructed,and the time series modeling method is adopted to accurately extract the features that can express the characteristics of the fault.After the vibration signal is decomposed,the numerical features of chaos and fractal are calculated.According to the evaluation index,the nonlinear degree of decomposition signal is judged.The AR models are established for linear components to extract the linear model parameters.The Volterra models are established for low nonlinear components to extract Volterra model parameters.Based on the deep study of nonlinear feature extraction,the realization method of HMM technology is discussed.The HMM model is proposed for fault identification.The bearing signal of the rotating machinery is analyzed,the linear and nonlinear feature quantities are extracted after the signal decomposition and are input into the HMM model.The normal signal,inner ring fault,outer ring fault,rolling failure are adopted for pattern recognition,the experimental results show that the model can accurately identify the rotating machinery failure,and the recognition rate is high.