The Analysis Of The Vibration Signal In Hydraulic Turbine Based On Wavelet Analysis And Artificial Neural Networks

With the energy restructuring and sustainable development strategies, present, China is speeding up hydropower development, hydropower equipment levels continue to move towards a new height, Hydroelectric units is moving in high head, high-capacity, high parameter development. Meanwhile the stability and security of the hydropower plant operation is more difficult than ever before .With the improvement of single capacity ,the increases of urbine unit size, the abate of relative stiffness ,the reduced of inherent frequency, the possibility of local resonance increases. .Stability problem increasingly prominent. Turbines stablity is an important problem. According to the actual investigation ,it shows that about 10 percents of the turbines once had a vibration. From Excitation source , most vibration hapeends on wheel and spindle system. 60 percent of the intensifying vibration is due to "water pressure pulsation" and "the unbalanced" revolver. In this paper, the state of draft tube pressure fluctuation is researched by monitoring its vibration signals. As the motor vibration signals are non-stable and random , the signal analysis based on traditional Fourier Transform can't completely meet the requirements for feature extraction .The status mathematical model of draft tube pressure fluctuation is very complex. This paper presents a state identification method combining with wavelet analysis and self-organizing artificial neural network . Wavelet transform is used to extract the feature from vibration signal and the neural network is used to identify the feature.In view of the non-stationary and time-varying characteristics of the pressure fluctuation signal of the turbine draft tube, this paper presents a state identification method combining with wavelet analysis and self-organizing artificial neural network. In order to verify the validity of the method ,this test simulates a certain type of pressure fluctuation signal of the turbine draft tube as an analytical signal, at first reduces noise interference of the signal based on wavelet thresholding, then lists the signal component of different frequency bands under reconstruction of wavelet decomposition coefficients , and the algorithm is realized with MATLAB software .at last extracts the band energy and enters them into the self-organizing artificial neural network as the feature vectors for pattern recognition. and the algorithm is realized with MATLAB software . The results show that the method can effectively identify the turbine running.