Research On Feature Extraction And Fault Diagnosis Of Wind Turbine Bearings

Since the entry of the world into the industrial age,more and more energy is needed in the production and life of mankind,which has led to the sharp depletion of traditional fossil energy.As the result,problems such as energy crisis,greenhouse effect,and deterioration of the ecological environment have arisen.In order to eliminate the impact which caused by the overuse of fossil energy,all countries of the world have accelerated the development of renewable clean energy,and wind energy is one of the important clean energy sources.As the main equipment for wind power generation,the installation capacity of wind turbine is increasing year by year.But the harsh environment might lead to frequent accidents.As the key component of the wind turbine,the rolling bearing is also the part of frequent failure.Therefore,it is necessary to study the fault diagnosis of bearing.Two key links of the process of bearing fault diagnosis,the feature extraction and fault diagnosis,have been studied.The main work is as follows:(1)Based on the analysis of the history of bearing fault diagnosis technology,the research status of feature extraction algorithms and fault diagnosis methods is studied.(2)The structure of the bearing in the wind turbine and the main failure forms of the bearing are studied,then the failure mechanism and fault feature of the bearing are analyzed.The source of the experimental data and the parameters of the data acquisition equipment are briefly described,and the fault feature frequencies of the bearing are obtained through the formula calculation.(3)EMD algorithm is combined with SVD for feature extraction.Aiming at the end effect problem of EMD,an improved extreme value extension algorithm is proposed to suppress the end effect.Then the effectiveness of the algorithm is verified through contrast experiments.The envelope spectrum analysis technology is used to analyze the fault signals of the bearing,it is observed that most of the fault information of original signal is contained in the first four intrinsic mode components(IMF),so only features of the first four IMF components need to be extracted.Finally,the feature information of the bearing signal is successfully extracted through the improved algorithm.(4)The related theory of fuzzy cognitive maps is studied.Aiming at the learning algorithm,a modified ant colony algorithm based on non-uniform mutation operator is proposed,and the flow of improved algorithm is described in detail.The advantages of the improved algorithm in learning fuzzy cognitive maps are verified through contrast experiments.(5)The common classification algorithms are studied,and the advantages and disadvantages are briefly described.The improved learning algorithm is applied to construct fuzzy cognitive map classification model.The effectiveness and accuracy of the classification model which constructed by improved algorithm are verified through contrast experiments on Iris data set.Finally,the fuzzy cognitive maps classifier constructed by the improved algorithm is applied for the fault diagnosis of bearings,which can accurately classify the types of bearing faults,with a accuracy rate of 99.17%.The validity and accuracy of the classification model in fault diagnosis are proved.