Research Methods Of Rotating Machinery Typical Fault Feature Extraction

Rotating machinery is currently the most widely used industrial production machinery and equipment. The research and application of Rotating machinery fault diagnosis technology can greatly reduce, avoid accidents and economic losses. This paper studies the content and substance of rotating machinery fault diagnosis technology, summarizes the common types of rotating machinery fault the failure mechanism and traditional characteristics of rotating machinery fault diagnosis and fault diagnosis reasoning extraction methods.In the first, we built rotating machinery fault analysis experiment system platform. The experimental system includes:rotating machinery test bed, eddy current sensors, data acquisition cards, signal conditioning modules, experimental data analysis computer. The experiment main uses the modeling crack fault in the laboratory bench. It monitors and collected vibration signal of rotating machinery fault for acquisition and analysis. In the time domain, using time-domain characteristics, in the frequency domain using conventional FFT-power general method and wavelet packet analysis and energy analysis of frequency domain method to extract the crack fault signal feature. Using the principle of feature evaluation, choose the most suitable for fault diagnosis fault features. Using BP neural network, SOC self-organizing competitive neural network and SVM support vector machine building the fault diagnosis model and to compare with their performance. SOC network and the SVM structure easy to determine, and the performance is better than BP network. The rotating machinery fault cracks, fault characteristics of a relatively poor degree of differentiation, presents a wavelet packet decomposition and time domain feature extraction combined with the new method to extract a new feature to analysis. The results show that the new feature extraction programs can be improved to some extent, distinguish between degrees of fault characteristics.