Study On Construction And Application Of Gear Fault Sensitive Feature Set Of Wind Turbine Gearbox

Wind power,as one of the fastest growing renewable energy sources at the present stage,is growing year by year in the global power production structure and has broad prospects for development.However,with the high development of wind power industry,the failure rate of wind turbines in actual operation is also increasing at the same time.Because of the harsh working environment of wind turbines,gear transmission system is prone to failure,which will lead to serious potential safety hazards in the operation of wind turbines,resulting in a large amount of waste of wind energy.How to accurately identify the running state of gear drive and improve its maintenance efficiency is a major technical problem to ensure the normal operation of wind turbine.The quality of the fault feature set of gears constructed will greatly affect the accuracy of the recognition results of the gear transmission system.Therefore,the research on how to construct a sensitive fault feature set for gear fault type identification will improve the reliability of the fault condition monitoring results of wind turbine gear transmission system,which has important scientific research and practical engineering application value.First,the main gear failure causes are studied.And then the vibration mechanism of the gear signal is analyzed based on the gear vibration signal model.Finally,the fault characteristics of gear vibration signal on the experimental platform are analyzed,which lays a theoretical foundation for the analysis of gear vibration signal in the following chapters.This paper proposes a central differential energy operator feature enhancement algorithm based on empirical wavelet transform.The principle of empirical wavelet transform and variational mode decomposition algorithm is introduced,and the adaptive improvement of the decomposition layer of variational mode decomposition method is presented.The effect of two decomposition algorithms on normal gear signal is compared and analyzed.The decomposition effect of the signal is finally carried out by using the central differential energy operator to test the decomposed gear signal.It is verified that the central differential energy operator based on the empirical wavelet transform enhances the fault characteristic signal and promotes the recognition effect of the gear fault.Aiming at the problem of large fault feature data sample and large redundancy,which is not conducive to gear fault recognition,the fault feature set processing method is studied from the perspective of dimensionality reduction,and a fault sensitive feature set construction algorithm based on EM-PCA is proposed.Firstly,the experimental data samples are processed by time-domain and frequency-domain indices respectively,and the more sensitive characteristic indices for gear state are expressed.Then,the proposed method is used to optimize the feature set and obtain the sensitive feature set.Experiments and comparative analysis show that this method can construct sensitive feature sets with good results,and is conducive to distinguishing faulty gears.This paper analyzes the algorithm principle of BP and support vector machine,and obtains BP-SVM recognition algorithm by combining the advantages and disadvantages of the two algorithms.BP-SVM is used to identify the fault of the sensitive feature set,which verifies that the sensitive feature set constructed by EM-PCA effectively improves the accuracy of gear fault recognition.