The Research Of The Early Vibration Fault Identification Technology Of Steam Turbine Based On Wavelet Analysis And ANN

This paper focuses on wavelet analysis, aims at early vibration faults which have strong background noises and weak characteristic signals, improves the traditional Donoho Hrad-threshold denoise algorithm, brings forward base on Shannon entropy of the optimization Wavelet Packet Analysis denoise algorithm, and obtains favourable effect. It combines wavelet analysis and BP neural network which has high precision of diagnosis and rapid study speed, analyzes the vibration signals include five conditions . The paper pick up energy of frequency range as the input parameter, and uses Levenberg-Marquardt algorithm to BP neural network the training and models under teachers' controls. It is proved that wavelet neural network can diagnoses early vibration faults and achieve constringency with a quick speed.In the process of analysis, author makes use of Matlab that is of great ability of visual data processing of digital signals of big amount and realizes direct and valid data processing.