Fault Diagnosis Analysis Of Fan Spindle Bearing Based On CPSO-BBO Opimization SVM

The rapid development of the wind power industry has also promoted the development of related auxiliary equipment.Wind turbine is an important equipment to convert wind energy into electricity,and spindle bearing is the main component of wind turbine drive system and power generation system,and also a vulnerable component.Because of its bad working environment,complex and changeable load,it is easy to break down,which not only brings economic losses to enterprises,but also causes certain security risks.Therefore,on-line monitoring and fault diagnosis of spindle bearings is particularly important.In order to improve the efficiency and accuracy of fault diagnosis and recognition of spindle bearing,this paper will take 2.0MW direct drive wind turbine spindle bearings as the object of further research in data acquisition,denoising,feature extraction and fault diagnosis.The contents are as follows:Firstly,the operating conditions and load characteristics of spindle bearings are analyzed,and the fault characteristics and mechanism are analyzed,which provides theoretical basis for bearing data acquisition and diagnosis research.On this basis,reasonable selection of vibration detection equipment and monitoring points,acquisition of Jinfeng 2.0MW direct drive wind turbine spindle bearing fault vibration data.Four kinds of fault vibration signals(normal state,inner circle fault state,outer circle fault state and rolling element fault state)are denoised by soft threshold wavelet denoising method.The denoising effect of the method is verified by the improvement of signal-to-noise ratio.Secondly,the Modified Ensemble Empirical Mode Decomposition,MEEMD is used to decompose the vibration signal,and the components with obvious fault characteristics are selected to calculate their kurtosis,mean square root,margin and impulse values as the eigenvalues of the vibration signal.By histogram,the distinction degree of vibration signal characteristic parameters of four kinds of spindle bearings is compared.It is proved that root mean square value,pulse index,peak value index and waveform index can better characterize bearing fault information.Finally,aiming at the slow convergence and low precision of Biogeography-Based Optimization(BBO)algorithm,this paper proposes(Cloud Particle Swarm Optimization,CPSO)to optimize the BBO algorithm and proposes a(Cloud Particle Swarm-Biogeography Based Optimization,CPSO-BBO)algorithm combined with support vector machine identification method,fault diagnosis of 2.0MW direct drive wind turbine main shaft bearings.In order toimprove the training and test time and recognition accuracy of support vector machine,the CPSO-BBO algorithm is used to optimize the kernel function parameters and penalty coefficients of support vector machine,and the optimization of support vector machine and BBO algorithm for CPSO-BBO algorithm is supported by Matlab.The vector machine compiles related programs,and brings four kinds of fault vibration data into the classifier,and compares the training results.The results show that the CPSO-BBO algorithm is superior to the BBO algorithm in training time,training accuracy and training speed.This verifies the feasibility and superiority of the method.