Condition Monitoring And Fault Diagnosis Of Wind Turbine Gearboxes

Wind turbine gearboxes are key and expensive parts of wind generating sets.Breakdowns of wind turbine gearboxes may make wind generating sets shut down for a long time,which will cause serious economic losses.Therefore,promoting research into condition monitoring and fault diagnosis(CMFD)of wind turbine gearboxes is important for ensuring normal running of wind generating sets and improving economic and social efficiency of wind power plants.After summarizing existing theories and methods for CMFD of machinery,this dissertation proposes the following schemes for feature extraction of gearboxes.Consequently,this dissertation overcomes some problems of existing methods and improves results of CMFD of gearboxes.(1)Enhancement of weak fault features of rolling bearings based on singular spectrum decomposition(SSD).The proposed method can determine embedding dimensions and choose principal components for reconstruction automatically.Furthermore,an additional wrapping-around can enhance oscillatory contents in original data and reduce mode mixing.Also,the effectiveness of the proposed method is confirmed numerically.Then,the proposed method is applied to fault diagnosis of bearings and also compared with envelope analysis and the method based on empirical mode decomposition(EMD)or ensemble EMD(EEMD).The results indicate that the proposed method has an advantage over the others mentioned above in enhancing weak fault features of rolling bearings.(2)Extraction of fault features of machinery based on nonlinearity and determination of data.This dissertation develops a method for quantifying nonlinearity and determination of data and employs the nonlinearity and determination for describing running conditions of machinery.Firstly,a simulation shows that the proposed method has good robustness to noise.Next,the proposed method is used in fault diagnosis of gearboxes and also compared with each of Approximate Entropy,Sample Entropy,Permutation Entropy and Delay Vector Variance.The results show that the proposed method outperforms the others mentioned above in feature extraction of complex data.(3)Condition monitoring of machinery based on adaptive multiscale symbol-dynamics entropy(AMSDE).This dissertation introduces multiscale analysis to statistical linguistic analysis and then defines AMSDE.Next,an entropy vector transformed from original data by AMSDE is used to describe running conditions of machinery.Afterwards,the proposed method is applied to identify conditions of gearboxes and rolling bearings.The results suggest that the proposed method can disclose temporal and spacial structures of data and is superior to traditional temporal statistics and nonlinear measures in condition monitoring of machinery.(4)Adaptive variable-bandwidth cost function(AVBCF)for estimations of instantaneous frequency of machines.The AVBCF can adaptively determine search regions of a ridge and relieve difficulties of the one-step cost function(OSCF)in adaptively determining search regions of a ridge.Firstly,the feasibility of the AVBCF is confirmed numerically.Moreover,the AVBCF is employed to estimate instantaneous frequency of a wind turbine gearbox and also compared with the OSCF and some traditional methods.The results indicate that the AVBCF performs well and outperforms the others mentioned above in estimations of instantaneous frequency of machines.(5)Extraction of fault features of wind turbine gearboxes based on Fourier decomposition method(FDM).Firstly,a comparison between filtering properties of EMD and FDM suggests that FDM shows good frequency resolution both in low frequency and in high frequency.Also,this dissertation numerally proves that FDM can transcend the limit of performance of EMD in separating adjacent-frequency components of a signal.Furthermore,the performance of the proposed method is benchmarked against the method based on EMD or EEMD by diagnosing faults of a wind turbine gearbox.The results show that the proposed method delivers a good performance and has an advantage over the others mentioned above in extraction of fault features.