Research On Fault Diagnosis Method Of Motor Bearing Based On Vibration Signal Processing

Rolling bearing is an important part of motors and this kind of rotating machinery, and rolling bearing is the most prone to failure in various parts of the rotating mechanical equipment, thus detecting the cause of rolling bearing failure timely is particularly important, Wherein the information contained in the vibration signal of bearing is rich, which has a clear physical meaning and reflect the operational status of equipment most directly and fully, therefore collecting vibration signals of motor bearing for fault diagnosis is also the most common and effective. This paper is based on analysis of giving expansion and research on rolling bearing fault vibration signal, focusing the principle of wavelet noise reduction, empirical mode decomposition principle and BP artificial neural network theory in detail. Two methods are mainly used for judging and identification on rolling bearing fault type. Method One combine methods of wavelet threshold de-noising method, correlation coefficient, kurtosis criterion and Complementary Ensemble Empirical Mode Decomposition for extraction on characteristic frequency of different fault types bearing fault, to achieve judgment on the type of bearing failure.Method two is based on the type of fault features of vibration signal of rolling bearings which is under varying degrees of damage, extracting their associated characteristic feature amount, while selecting sensitive feature vector to construct artificial BP neural network, making fault pattern recognition of fault type through the trained BP neural network.This paper mainly analyzes the mechanism of vibration signal generated of motor bearing, discloses the rationality of using vibration signal of rolling bearing to giving fault diagnosis on bearing, getting the fault type through fault extraction of three typical rolling bearing fault type, correspond with the fault characteristics frequency of different types of rolling bearing got by the calculation method of fault characteristics frequency. In the noise reduction processing for rolling bearing fault vibration signal, the main method used is the wavelet threshold noise reduction methods, and analyzes the flaws and shortcomings of the traditional wavelet threshold noise reduction methods meantime, proposing an improved wavelet threshold noise reduction algorithm, combined with vibration signals’ sensitive kurtosis parameters, using the new algorithm in the noise reduction processing on machinery vibration signal, experiments show that the improved wavelet threshold noise reduction algorithm can reduce high frequency noise signal components On the premise of keeping useful failure signal in mechanical vibration signal unwanted.Then the improved wavelet threshold noise reduction algorithm and Complementary Ensemble Empirical Mode Decomposition method combines, which is based on the correlation coefficient method and the failure vibration signals kurtosis criterion at the same time, proposing a new method for rolling bearing fault signal extraction, it illustrates this method giving a clearer fault signal feature information in the extraction process of rolling bearing fault signal.Besides, this article studies the role of BP neural network recognition in multi-mode rolling failure, based on collecting a large number of rolling bearing fault vibration signal’s sample data, making extraction of time domain parameters and frequency domain parameter of fault vibration signals, establish matrix-vector by these characteristic parameters, establish artificial BP neural network by training feature vectors to achieve multi mode pattern recognition of rolling bearing fault judgment, the BP neural network under four operation pattern recognition and ten operation pattern recognition for rolling bearing are established, and the experiment shows that the two neural network for rolling bearing fault identification play a good effect equally.