Study On Signal Analysis And Feature Extraction For Turbo-Generator Unit

As human have been exploring further in-depth theoretical study on the various fields of science, which largely contributing to a rapidly increase of the level of the modern industrial technology as well as the rapid development of the science technology. As the main equipment in electric power industry turbo-generator unit continues toward the development of highly automated, large-scale, high-efficiency, intelligent direction. But this leads to the increasing factors that can impact its security, so it is important that improving the level of condition monitoring and the efficiency of fault diagnosis to ensure the equipment is running with safety and economy.Signal analysis and feature extraction is considered a key step in the unit equipment condition monitoring and fault diagnosis process, playing a role in processing and transforming the sensitive parameters to extract the fault symptoms. The unit is producing a large number of signals every second, including pressure, temperature and vibration signal. They have a close relationship with the health status of unit, so the preprocessing and feature extraction is a prerequisite for revealing the status information of unit. Aiming at this goal, the paper regards the monitoring signals of unit as the object, which is processing based on the research and simulation of the wavelet transform and Hilbert-Huang transform analysis method. The main research works are as follows:In turbo-generator fault information research, first, through the mechanism analysis of failure mode we can understand the nature reason of the failure and its impact. Then combining the previous expertise and live case data for the set of fault symptoms to establish FMEA analysis table, while fault tree analysis method is top-down for the failure mode adopted to tease out the underlying causes. Finally, the fault warning model is composed of data signs, curve signs, graphic signs and the descriptive signs.In signal analysis and processing, wavelet transform is adopted to the de-noising processing of the unit signals and it gets a better noise cancellation, which has an important role to reveal unit failure, using the time-frequency analysis characteristic of wavelet transform and Hilbert Huang transform to achieve the function of time-domain mutation detection and trend forecasting and frequency domain analysis, verifying that the two transforms can be applied to the unit failure warning and fault diagnosis. In feature extraction, through simulation of the traditional Fourier transform spectral analysis results show that the method has limitations in dealing with non-stationary signals, while using the wavelet transform band energy feature extraction method to processing unit signals, which can effectively extract the energy distribution characteristics of the signals in the respective frequency bands. Simulation results show that feature method of band energy extraction based on wavelet transform is an effective method for fault diagnosis.