Bearing Soft Fault Monitoring Using Compressed Sensing And Long Range Dependence Random Model

Because of the rotating machinery system load structure has become increasingly complex and continuous operation, the mechanical operation condition online monitoring and fault prediction are researched to ensure the safe and reliable operation of the rotating machinery system. Based on the theory of compressive sensing (CS) and long range dependence random model, the rotating machinery system non-stationary condition monitoring and fault trend prediction were proposed to realize rotating machinery system predictive maintenance. The main achievements are presented as follows:1、Based on the basic theory and methods of condition monitoring and fault prediction, this paper discusses the vibration mechanism of the rotary unit and conventional signal processing methods, then introduces the conventional prediction methods in the field of fault prediction and vibration characteristic extraction algorithm, and provides mechanical fault monitoring criterion.2、In order to overcome the problems such as mechanical vibration signals high speed transmission and long-term storage online are limited by hardware, and can not satisfy requirement of the high resolution and real-time monitoring. Based on the thepry of local mean decomposition(LMD), vibration signal sparse representation, measurement matrix and reconstruction algorithms, A bearing fault vibration signal reconstruction approach based on local mean decomposition and nonconvex penalized Lq minimization compressed sensing is proposed. This method combines sampling and compression with a small amount of sample to reconstruct signal well. The result show that this methods not only improves compression accuracy and reduces the reconstruction computational complexity, but also proved the reconstruction effective of the bearing actual operation signal is better than the traditional algorithm.3、Owing to the influence of operation mode and equipment itself nonlinear, the bearing vibration signals are often show the features of non-stationary and poor reproducibility. Due to the static analysis of the traditional method, and ignores the dynamic changing bearing failure, so a bearing fault pattern recognition based on harmonic wavelet sample entropy and hidden markov model is proposed, by applying the harmonic wavelet to decompose each bearing fault signal, the frequency layer characteristics of harmonic wavelet three-dimensional time- frequency trellis is used to estimate the reasonable sample entropy parameter, and the feature vector sequence was constructed by extract sample entropy of rolling bearing each layer vibration signal. After that, the HMM are trained by given feature vector sequence and the bearing fault types are identified by comparing the logarithmic likelihood probability value. The results show that this method can achieve intelligent identification of the bearing fault types with high success rate and stability.4、A bearing fault feature extraction method based on LMD and spectral kurtosis diagram is proposed, by analysis spectral kurtosis diagram can accurately identify the band-pass filter center frequency and bandwidth, and overcome the bearing fault carrier frequency and size estimate defects by the subjective judgment, then analysis the square envelope spectrum of the band-pass filter fault signal, The result show that this method can identificate the proper rotation frequency and its harmonics, fault characteristic frequency and its harmonics.5、Extract the trend prediction of the vibration signal characteristic, the vibration intensity value, and then treat the early stage of the vibration intensity values as non-stationary time series, by using long range dependence to forecast the vibration intensity value. The result show the predict superior of the long range dependence, at the same time this method also proved the effectiveness of the R/S method to test the non-stationary time series. This random predict model have great significance in rotating machinery equipment state real-time assessment and prediction, its laid a solid foundation with the rotating machinery equipment status maintenance, and provides a new thought in equipment running status maintenance model.