Study On Intelligent Diagnosis Technology Of Bearing Based On Demmodulated Resonance And Artificial Neural Networks

The condition monitoring and error diagnosis of the rolling bearing, which is the most widely used device in mechanical area, is a key research subject in the engineering & technology fields of the world. The rolling bearing is an easy-mar device. It is reported that nearly 30 percent of the revolving machinery fault is caused by the rolling bearing's fault.This paper introduced a new process of intelligent diagnosis for rolling bearing. After designing and building the experiment platform suitably and programming the data capture program, vibration signals of rolling bearing are collected and the bearing's condition corresponding to the signal are identified. Then artificial neural network techniques are applied to the fault diagnosis process. The network is built by choosing the appropriate network structure and input layer characteristic parameter. The third step is to complete the intelligent diagnosis of rolling bearing by using the test sample to test the network's recognition capability. Lastly but not least, the Graphical User Interface is developed to further improve the intelligent diagnosis process.In this paper, the main method of the signal processing is the demodulated resonance technique which uses the system connatural vibration caused by the pulse force. The pulse force is caused during the working progress of the rolling bearing. Some connatural vibration is separated through the band-pass filter and Hilbert transform is applied to envelope and demodulate. This step can remove the frequency components of the high frequency damped vibration and the low frequency enveloping signal which contains the fault information is then produced. The fault information of the rolling bearing can be provided by carrying out the spectrum analysis to this enveloping signal.Based on the demodulated resonance diagnosis, the neural network technique was applied to carry out intelligent diagnosis. The author chose the BP network, the RBF network and the Elman network respectively to classify the rolling bearing fault pattern. Three different groups of features are put into the networks for comparison. The three groups of features are the time range parameters, the frequency range parameters and the demodulated resonance parameters of the vibration signal. The large number of experiments showed that demodulated resonance parameters can diagnose rolling bearing faults more correctly and effectively. The network can identify accurately the normal state, fault of the outer race and fault of the inner race. The capability of the three different nets was compared according to different kinds of inputting features.Based on the intelligent diagnosis process, demodulated resonance technique are combined with artificial neural network technique. A graphical user interface is developed, this friendly interface improves the efficiency of the fault diagnosis process even further.