Research On Feature Extraction Method Of Non-stationary Vibration Signal Of Hydropower Unit

Hydroelectric unit is the largest clean energy in the world,which has been widely used in various countries.However,the structure of hydropower unit is complex and the operation condition is bad.Many factors including hydraulic,mechanical electrical and soon will affect the stable operation of the hydroelectric unit.Therefore,the unit is also prone to various faults,and the vibration signal of the unit has strong nonlinearity,which makes it difficult to extract the fault feature signal of the hydropower unit.Therefore,it is of great significance to research the feature extraction method of non-stationary vibration signal for the operation and maintenance of hydropower units.In this paper,the vibration mechanism and typical faults of hydroelectric units are analyzed firstly,and the classification of vibration faults and the vibration characteristics of faults of hydroelectric units are studied emphatically.Finally,five typical faults of hydropower units are analyzed in detail,such as rotor unbalance fault,rotor misalignment fault,rubbing fault between stator and rotor,eccentric vortex band of draft tube and extreme frequency vibration.The analysis will lay a theoretical foundation for the subsequent fault feature extraction of hydropower units.Then,aiming at the difficulty of extracting weak feature signals and gradual feature signals in the early stage of hydropower unit fault,this paper proposes a sensitive component extractmethod based on variational mode decomposition(VMD).Firstly,VMD is used to decompose the vibration signal of the unit,and then the sensitivity coefficient is constructed based on mutual information,and the component with larger sensitivity coefficient is selected as the sensitive component of the signal.Finally,the spectrum analysis of the obtained sensitive component is carried out to analyze the effective information contained in the sensitive component.The simulation analysis and example verification show that the proposed method can accurately extract the weak characteristic signal,low frequency signal and abrupt signal in the early stage of fault,which has practical application value.Finally,aiming at the problem of misjudgment or missing judgment in the single channel fault analysis method of hydropower units,a feature extraction method based on two-dimensional empirical mode decomposition(BEMD)and full vector spectrum(FVS)is proposed.In this method,the orthogonal vibration signal of the unit is obtained by utilizing the orthogonal sampling technique,and the BEMD decomposition is carried out in the vibration signal to obtain the inherent modal function with phase information.Then the sensitivity coefficient of each component is calculated,and the component with larger sensitivity coefficient is selected as the sensitive component.Finally,the full vector envelope technology is used to fuse the selected sensitive components for full vector envelope spectrum analysis.Simulation analysis and example verification show that this method can extract fault feature components of the unit comprehensively and accurately,which has good practical application value.