Development Of Fault Diagnosis System For Turbine Shaft Based On Neural Network

Steam turbine, as the typical rotating machinery, plays an important role in our modern industry like power, metallurgy, and shipbuilding. The normal and safe operation of Turbine can create not only a sound material wealth for the country, but also an enormous benefit for the society. Therefore, the research on turbine fault diagnosis system exerts great influence on its normal and safe operation.This paper developed the fault diagnosis system of turbine shaft in neural network on the basis of VC ++ and SQL Server. This diagnosis system consists of a series of functional modules, including the display of the vibration spectra, the extraction of the symptoms, the intelligent diagnosis and the re-checking of the filed diagnostic cases. In the module of vibration spectra displaying, the time-domain waveform, frequency spectrum and the axis orbit of filtering can all be manifested or reflected and the diagnosis sample can also be extracted from the frequency spectrum. The neural network training module employed the most advanced and sophisticated BP neural network as the diagnosis method based on the modification of the traditional BP algorithm, which includes adding momentum formula, adapting learning rate, and introducing the batch algorithm. Intelligent diagnosis module realized a joint multi-symptom diagnosis by employing the frequency spectrum symptoms, associated symptoms, symptoms of rising speed in cold conditions and vice versa. And in the final part, a favorable result has been obtained in testing for all the functional modules based on the experimental stage of Bentley rotor system.