Identification Of Cavitation State Of Kaplan Turbine Based On Vibration Signals

As the main type of medium and low head hydropower resources,axial flow turbines have prominent problems in operation,such as low efficiency,serious abrasion damage,and large coupling deviation.Cavitation,as one of the major influencing factors during operation,can cause blade damage and reduce efficiency.Therefore,this article takes the axial flow turbine as the research object to study the time-frequency variation rules of the runner vibration signal and the draft tube vibration signal during the process of the axial flow turbine from normal state to severe cavitation state.By extracting key characteristic values,the cavitation state of the axial flow turbine is identified and analyzed.The main research results of this article are as follows:（1）A comprehensive testing system including high-speed photography,LDV,and vibration testing was constructed.The former recorded cavitation images inside the runner,and the latter collected vibration signals from the turbine runner chamber and draft tube.Through time and frequency domain analysis of the vibration signals of the runner and draft tube,it is found that as the degree of cavitation increases,the time and frequency domain eigenvalues reflect nonlinear changes,which cannot effectively identify and analyze the cavitation state;Through short time Fourier transform,it is found that the amplitude of the spectrum component in the low frequency region increases significantly when cavitation is severe;Through MUSCI analysis,it is found that cavitation can cause energy migration from high frequency regions to low frequency regions,increasing the risk of overall unit resonance.（2）The MF-DFA method is introduced to decompose the vibration signals of the runner and draft tube,and the results show that the Hurst exponent exhibits a significant zoning phenomenon for the vibration signals of the runner.The slope value of the scaling exponent varies slightly at different cavitation stages,and the characteristic value in the multifractal spectrum α₀ and α₍Max can effectively identify the degree of cavitation in the runner;For draft tube vibration signals,both the slope of Hurst exponent and the linearity of scale exponent can be used as effective eigenvalues for identifying cavitation states,and the Δα And Δf can also be determined as characteristic values.（3）Reliability analysis of characteristic vectors in time and frequency domain is carried out by correlation coefficient method.For runner vibration signal,four eigenvalues,peak value of eigenvalue,root mean square value,rectification average value and frequency variance,are selected to combine.For vibration signal of draft tube,4 characteristic values of center of gravity frequency,mean square frequency,mean square root frequency and rectification mean value are selected to combine.The recognition rate of BP neural network is basically similar to that of SVM,but the recognition rate of BP neural network is higher for draft tube vibration signal.Therefore,considering the feasibility of real machine test,it is better to select BP neural network for recognition.