Research On Fault Diagnosis Method Of Wind Turbine Based On Sparse Decomposition Of Vibration Signals

The operation condition of the wind turbine is poor,and the load of the wind turbine is complex and changeable,which makes the unit fault rate higher,resulting in low utilization rate and utilization hours of wind turbines and high operating and maintenance costs.In order to ensure the safe operation of the unit and improve the economic and market competitiveness of the wind power,it is necessary to monitor the state and diagnose the fault of the wind turbine.The condition monitoring system(CMS)of wind turbine can judge the operational status of units by monitoring the vibration of transmission chain components.The main methods of vibration signal analysis are based on Fourier transform,wavelet transform and other orthogonal linear transformation.This kind of analysis method often breaks the fault information into too many base functions,which is not conducive to the extraction of fault features.Sparse decomposition is based on the redundant base function library(redundant dictionary),and decomposes the signal into a series of linear combinations of basis functions.The algorithm can adaptively select the base function which matches the signal adaptively according to the characteristics of the signal.Because the dictionary is highly redundant,the fault features can not be scattered into too many base functions when sparse decomposition is used to represent complex vibration signals.Aiming at the early fault feature extraction of the wind turbine transmission chain,the sparse decomposition method is introduced into the vibration signal processing of wind turbines.The sparse decomposition and minimum entropy deconvolution algorithm are combined to enhance and extract the fault feature in the signal,and the validity of the selected method is verified by the simulation signals and the actual vibration signals of the unit.The main contents of this paper are as follows:(1)Two key problems in the implementation of sparse decomposition are studied and analyzed,namely sparse decomposition algorithm and the construction of redundant dictionary.By analyzing the simulation signals and the actual vibration signals,the characteristics of different sparse decomposition algorithms and redundant dictionaries are demonstrated.(2)The theory of minimum entropy deconvolution(MED)and its implementation are introduced.The minimum entropy deconvolution and the sparse decomposition method are combined to extract the fault features,which are verified by the simulation signals and the actual vibration signals.(3)Collect Bearing vibration data for a period of time before a fault occurs,then screen the collected data at regular intervals.The fault feature components in the selected data are extracted by using the proposed method.The trend of the fault characteristic value is generated,and compared with the common time-domain eigenvalue trend.It is proved that the selected method can effectively characterize the fault development trend.The above research shows that the sparse decomposition method has a good application prospect in the feature extraction of the early fault signals of wind turbines.