Rotor State Detection And Identification Based On Wavelet Analysis And Neural Network Research

With the development of the Rotating Machinery, the vibration signal processing and the vibration condition identification on the Roter Machinery become more and more important. In this paper, the application of Discrete Wavelet transform and Neural Network techniques in the field of rotor status monitoring and identification is researched deeply.In this paper, the DWT is introduced to the vibration signal analysis because it is good at processing the non-linear fault signal. On the base of studying the methods of drawing cha- racteristics of the non-linear fault signal by DWT, an improved method is presented in this paper. The characteristics drawn by this improved method are more integrated, which can show the degrees and locations of the non-linear vibration fault. And the validity of the im- proved method is proved by the experiment data from the rotor rig.In this paper, the method of improving Resilient back propagation is presented on the base of comparing the various improved BP algorithms. The improved Resilient Back propagate- ion has better performances in training time, memory consumption and predication exactitu- de. And the classifier of rotor conditions based on the improved RPROP algorithm and the improved method of drawing characteristics of vibration signal by DWT is designed.In this paper, DWT, NN and NI techniques are used to design an experiment system of rotor vibration status monitoring and identification with sensors, DAQ-card, signal adjusted card and PC. This system can complete the real time testing, vibration signal analysis and status identification on the rotor vibration rig. It is carried out well in experiments and it provides some refrences to status monitoring and identification of the Rotating Machinery.