Study On Fault Diagnosis Of Steam Turbine Based On Wavelet Packet And Improved DE-ELM

Steam turbine is an important component in the production process of power plant, is mainly used in the production process of heat engine, the coal combustion for heat into mechanical rotating parts, and the generator is converted into electricity. Its efficient and stable work is the most important of economic production in power plant. Along with the power plant equipment to the large unit, high parameter direction to develop, the complexity of the equipment and the level of automation is also increasing. Therefore, in order to ensure the safe operation of equipment, reduce the safety cost and improve the utilization rate of the equipment, it is necessary to take effective methods to detect and diagnose the condition of the steam turbine.In this paper, using the method of wavelet packet analysis, for steam turbine vibration signal contains characteristics of a large number of mutations and short-time impact component of typical denoising methods to improve the implementation of, the way based on Shannon entropy, the optimal wavelet packet commissure of different frequency band threshold selected for vibration signal de-noising and achieved good results. And use "wavelet packet energy spectrum method to decompose the denoised signal, and the nodes of each band separately reconstructed to obtain the energy data reconstruction, and normalization of the vibration signal of vibration signs, construct the characteristics of steam turbine vibration fault. Extreme learning machine has the advantages of fast speed and good generalization ability compared to other classification network, but its input weights and implied the biasing layer is chosen at random and can not ensure the optimal, so the improved differential evolution algorithm (IMDE) of global search and rapid evolutionary advantages of poor, on the ultimate learning machine was optimized and designed IMDE-ELM classifier. According to the vibration fault characteristic table of steam turbine, the IMDE-ELM classifier is used to train and recognize the vibration state of the steam turbine. Matlab software through the preparation of the program, and the diagnosis of the results of simulation and comparison.The simulation results show that the IMDE-ELM model is faster than other models, and the diagnostic accuracy can reach 100%. The model can be used to identify and classify the vibration state of the turbine, which provides a solid theoretical basis for the research of vibration fault diagnosis of steam turbine.