Fault Diagnosis Of Wind Turbine Gearbox Based On Multi-source Information Fusion

The gearbox is an important mechanical transmission system in the wind turbine.If it fails,it will not only seriously threaten the safety of the whole machine,but also greatly increase the operation and maintenance costs and reduce the economic benefits of the wind farm.Therefore,carrying out research on fault diagnosis of wind power gearbox systems can discover gearbox system failures and diagnose their failure types in advance,which helps to contain the devastating failures of wind turbines in advance and is of great significance to the realization of intelligent health management of wind turbines.In this paper,based on multi-source information fusion technology,by combining BP neural network and DS evidence theory,a wind power gearbox fault diagnosis model based on system / component feature layer & decision layer is established,and the wind power gearbox test is run on a domestic wind farm The data is the object of analysis.The comparison of the analysis results of multiple sets of measured data verifies the effectiveness of the fault diagnosis model proposed in this paper.It has certain theoretical significance and practical value for the wind power gearbox system fault diagnosis.The main content of the paper is as follows:(1)It summarizes the working principle of wind turbines,categorizes the common failure modes of wind power gearbox systems,summarizes the current commonly used methods in the field of fault diagnosis and their advantages and limitations,and provides support for the subsequent fault diagnosis of wind power gearbox systems.(2)Considering the advantages of BP neural network self-learning,parallel learning of learning,and significant effects in pattern recognition,by combining the solid theoretical foundation of DS evidence theory and the advantages of processing uncertain information with BP neural network,a BP-based neural network is proposed Neural network-improved fault diagnosis algorithm of wind power gearbox system based on DS evidence theory,effectively solved the shortcomings that the basic probability distribution value of DS evidence theory is difficult to obtain and the conflicts of various evidence bodies,and realized the test signal of multiple wind power gearbox Information fusion.(3)Collect the vibration,oil temperature,current signal and other data of the wind power gear box through the wind farm SCADA system,and use BP for the five states of gear failure,bearing failure,shaft failure,normal state and uncertain state in the gear box The neural network carries out the characteristic layer diagnosis of the gearbox system;on this basis,combined with the data of the vibration,oil temperature,current,torque and speed signals of the wind power gearbox,the characteristic layer diagnosis of the gearbox bearing is performed by the BP neural network,To achieve the diagnosis of five states of bearing outer ring failure,inner ring failure,rolling element failure,normal state and uncertainty.(4)The DS evidence theory method based on weighted improvement is proposed,and the diagnosis result of the BP neural network feature layer is used as the basic probability distribution value(BPA value)of the DS evidence theory.The body is fused to obtain a diagnosis.Finally,the effectiveness of the proposed method is verified by analyzing the results of the sample data of the two groups in different time periods.