Research On Fast Fault Diagnosis Methods Oriented To Motor Bearing

Motor bearings employed in a manufacturing environment must be able to operate as long as possible having as little downtime as possible. Therefore, maintenance is crucial in order to allow for the equipment to perform its designated tasks without failure, especially on critical systems. In recent years, several researches and developments have been made in engineering field, which aim to improve the healthcare diagnosis and treatment. Signal analysis methods such as Fourier transform and Wavelet transform are widely used in this field. However, some methods are limited to the linear and stationary signal analysis. As a result, they are not all well adaptive methods for bearing vibration signals, which in fact are mostly nonlinear and non-stationary signals.This thesis introduces the research object and methods at first, then introduces the domestic and foreign development and research status in the field of bearing fault diagnosis. An important objective of this thesis is to detect and distinguish various fault types. This work aims at developing an ensemble diagnosis method for bearing fault. The system also synchronize update data as well as provides a right diagnosis results for the machine. Empirical mode decomposition has very strong ability in non-stationary signal decomposition, signal can be decomposed into a series of components from high frequency to low frequency; The feature of the vibration signals of the motor bearing, usually non-Gaussian, was suitable for extracting by the fixed-point iteration fast independent component analysis algorithm. Then the online incremental method was adopted to optimize the probabilistic neural network structure and train probabilistic neural network parameters to improve the classification adaptability of probabilistic neural network. Incremental probabilistic neural network has strong ability in classification. Comprehensive consider the characteristics of each method, this thesis proposes an ensemble motor bearing fault diagnosis method EMD-FICA-IPNN. Analysis approach was applied in the data from bearing fault diagnosis experiment of Case Western Reserve University, lastly. The results show that the new approach has better precision, adaptability and quickness.The construction of experimental platform contains two parts, software and hardware. The hardware was designed by the Multifunction DAQ Card PCI-1710 HG and the accelerometer LC0159. The corresponding data collection and fault detection soft-ware were designed based on the virtual instruments Lab VIEW. After the evaluation, the experimental results show that the adaptability, accuracy and timeliness by ensemble approach are better than that of traditional probabilistic neural network.