The Research On Multi-fault Diagnosis Of Wind Turbine Gearbox

With the development of wind power,the safety operation and fault diagnosis of wind turbine have been attracting more and more attention of scientific researchers.Gearbox is an important transmission part in wind turbine whose performance affected by various factors and its failure may break the whole transmission line down.Thus,the research of the fault diagnosis of gearbox to ensure safe operation of wind turbine has the vital significance.The research of thesis is mainly about the multi-fault diagnosis of gearbox and this thesis solve the multi-fault diagnosis problem by two different method:1.In this thesis,a novel UBSS based algorithmic solution is proposed to solve the multi-fault diagnosis problem.The algorithm divides the UBSS problem into two sub-problems:source number estimation and source signal recovery.The number of source signals are estimated by applying the empirical mode decomposition(EMD),singular value decomposition(SVD)and K-means joint approach.Then,the obtained source signals are transformed into time-frequency(TF)domain by the short-time Fourier transform(STFT)to obtain the sparse signal representations.At last,the mixing matrix is estimated by using fuzzy C-means(FCM)clustering and the source signal recovery is realized through minimizing the l1 norm of the source signals.The experiment result clearly confirms the effectiveness of the proposed algorithm for nonlinear multi-fault diagnosis of gearbox.2.The other method proposed by this thesis is based on support vector machine(SVM)probability estimation.The method constructs a SVM model for each accelerometer which mounted at different place of wind turbine gearbox.Every SVM model will output the class probabilities respectively.Together these class probabilities determine the final class probabilities of the sample.In order to improve the accuracy of SVM model prediction,this paper utilizes the ensemble empirical mode decomposition(EEMD)for fault feature extraction.The solution is validated through simulation data and real data.The result confirms its effectiveness.