Research On Fault Identification Method For Key Components Of Wind Turbine Gearbox

The gearbox is an important component connecting the wind turbine to the main shaft and the generator.Its failure accounts for only 4% of the total failures.However,the downtime caused by it is the longest of all failures,resulting in the operation and maintenance of the wind turbine Increased costs have seriously affected the efficiency and economic benefits of power generation.Therefore,it is urgent to carry out research on fault identification methods for key components of wind turbine gearboxes.This research topic mainly has the following three research contents:(1)Noise reduction of signals of key components of wind turbine gearbox.Aiming at the problems of large background noise interference and weak fault information of wind turbine gearbox,signal denoising based on adaptive maximum correlation kurtosis deconvolution algorithm is proposed.Due to the influence of time-varying wind load service environment and intermittent working mechanism,the vibration signal collected by the wind turbine usually has strong background noise.If the gearbox or one of its components fails at this time,the fault signal is very large It will be overwhelmed by noise,so that the fault cannot be found in time,and the loss caused is immeasurable.Therefore,in this paper,under the noise reduction effect of the maximum correlation kurtosis deconvolution,the key parameters are optimized in combination with the kurtosis value,in order to further improve the noise reduction performance of the maximum correlation kurtosis deconvolution.(2)Extraction of fault features of key components of wind turbine gearbox.Aiming at the problem that "non-stationary" and "non-linear" characteristics make it difficult to extract the fault features of key components of wind turbine gearboxes,a combination of variational mode decomposition optimized by immune genetic algorithm and improved multi-scale arrangement entropy is proposed to extract fault features Constructed with feature vector set.The components of the fan have a strong coupling with each other,combined with the particularity of the service environment and working mechanism.Therefore,the vibration signal has the characteristics of coupling modulation,showing a non-stationary vibration signal,and the traditional signal processing method is processing The non-stationary signal is weak.Because of the superior performance of non-stationary signal processing by variational mode decomposition,this paper combines variational mode decomposition and immune genetic algorithm to extract the improved multi-scale permutation entropy of the decomposed signal in order to construct a fault feature vector.(3)The learning method of fault migration of key components of wind turbine gearbox.The typical fault diagnosis is based on "sufficient fault data" and "clear fault type" as the prerequisites for model training and fault recognition in terms of pattern recognition,that is,a large amount of fault data can be obtained,and the fault type of the acquired data is known.The data of the key components of the wind turbine gearbox is contrary to this,with low value density(the data in the normal state is much more than the data in other states),and the low availability(the data type of most data is not clear).However,since the wind turbine cannot be shut down frequently for self-check failure and manual labeling data is time-consuming and laborious,the labeling data is costly,resulting in a lack of health labeling information in the monitoring data.Therefore,transfer learning based on improved multi-core semi-supervised transfer component analysis is proposed to make full use of the rich typical fault information and sufficient available samples in the test equipment data to make up for the problem of lack of typical fault information and lack of health marker information in actual data.This research topic takes the key components of the wind turbine gearbox as the research object,based on the maximum correlation kurtosis deconvolution,immune genetic algorithm,variational mode decomposition,improved multi-scale permutation entropy and improved multi-core semi-supervised migration component analysis related theory And methods,focusing on noise reduction and fault identification of vibration signals,multi-sample feature extraction in the field of intelligent diagnosis,and fault identification based on semi-supervised migration component analysis.The effectiveness of the proposed method is verified through experiments.