Research On Condition Degradation Assessment And Remaining Useful Life Prediction For Wind Turbine High-speed Shaft Bearings

With the global energy shortage and environmental pollution aggravating,wind energy,as a kind of renewable energy with abundant reserves,clean and safe,has been highly concerned by people.Wind power generation is the most effective way to use wind energy,and it has been developing rapidly in recent years.The operational safety assessment and fault prognosis of wind turbine components are two important factors which directly affect the availability and economic benefits of wind farms,and have gradually become an important research topic in the field of wind power.The transmission system is the core part of wind turbine,and the bearing is the key component of the transmission system.Once the fault occurs,it will lead to the failure of the whole transmission system.Therefore,it is of great engineering significance to carry out condition monitoring and fault prognosis of wind turbine bearing.The main purpose of condition monitoring is to judge the health status of bearings,so as to deal with the faults that have occurred or will happen in time,ensure the reliability of operation and reduce the maintenance cost.The complex operating conditions and obvious individual differences of the wind turbine make it difficult to maintain.The existing safety assessment and fault prognosis theory can not meet the demand of wind power industry for wind turbine health status management.In this paper,a new model for intelligent operation and maintenance is presented,which integrates signal processing,feature extraction,cross-domain diagnosis,state identification,and life prediction.The methods for state degradation evaluation and performance trend prediction of the wind turbine bearing are studied.A deep transfer learning network for cross-machine fault diagnosis is constructed.A effective method for early degradation detection is proposed.The degradation process tracking and online remaining useful life prediction are realized.And then,a new prognosis model of multi-source information fusion is built.This study focuses on the key theories and methods of degradation assessment and remaining useful life prediction for wind turbine high-speed shaft bearings.The main contents are as follows:(1)To solve the problem that the bearing running states are difficult to be divided and the low accuracy of performance trend prediction,two methods are proposed,that is,the degradation state evaluation method based on VMD-AFCM-SVM and the improved LSTM degradation trend prediction method.The VMD-AFCM-SVM algorithm uses the relative features to establish sensitive feature data set,and clustering evaluation indicators to construct adaptive function,then the automatic updating of the clustering results of the model is realized,the optimal state number in the process of bearing operation is obtained,the time interval of bearing in different degradation states is determined,and the health level assessment of bearing are obtained.In addition,the improved LSTM method is a real-time updating method based on error minimization,which can update the model online with less sample data,and solve the problem that the traditional LSTM model can't use the online data reasonably.Taking the life-cycle bearing data set provided by Cincinnati University and the real-world wind turbine high-speed shaft bearing data set as examples,the effectiveness of the proposed method is verified.(2)In order to solve the problem that the generalization ability of laboratory bearing fault diagnosis technology is not strong and the real bearing fault data is difficult to be labeled,a novel intelligent method named deep transfer network(DTN)with multi-kernel dynamic distribution adaptation(MDDA)is presented to address the problem of cross-machine fault diagnosis.The DTN has wide first-layer convolutional kernel and several small convolutional layers,which is utilized to extract transferable features across different machines and suppress high frequency noise.Then,the MDDA method constructs a weighted mixed kernel function to map different transferable features to a unified feature space,and the relative importance of the marginal and conditional distributions are also evaluated dynamically.The proposed method is verified by three transfer learning tasks of bearings,in which the health states of wind turbine bearings in real scenario are identified by using diagnosis knowledge from two different bearings in laboratories.The results show that the proposed method can achieve higher diagnosis accuracy and better transfer performance even under different noisy environment conditions than many other state-of-the-art methods.The presented framework offers a promising approach for cross-machine fault diagnosis.(3)Aiming at the problem that the weak fault characteristics of rolling bearings are susceptible to noise and the starting point of degradation is difficult to detect,a method is proposed for detecting early degradation of bearings based on adaptive variational mode decomposition(VMD)and envelope harmonic-to-noise ratio(EHNR).Firstly,the minimum average envelope entropy is used as the objective function to search the optimal parameters of VMD adaptively by using the grey wolf optimization algorithm.Subsequently,the effective weighted sparseness kurtosis indicator is introduced to separate the effective modal components and the noise modal components,so that the reconstructed signal can filter out interference and retain fault information.Finally,the EHNR of the reconstructed signal is calculated,and its sensitivity to cyclical fault shock is used to detect the early degradation starting point of the bearing.The experimental results show that the proposed method not only solves the problem of false alarm in the early stage of bearing operation,but also can recognize the starting point of the bearing degradation process earlier.Its robustness and sensitivity provide the basis for early fault diagnosis and remaining useful life prediction of wind turbine bearings.(4)Due to the lack of sensitivity and robustness to periodic fault shocks,the traditional degradation indicators are unable to track the degradation process of wind turbine bearings timely and predict remaining useful life accurately.In this paper,a real-time remaining useful life(RUL)prediction method for wind turbine bearings based on the combination of EHNR and unscented particle filter(UPF)is proposed.Firstly,the early degradation starting point of the bearing is detected by calculating the EHNR of the vibration signal and the trend characteristic of the EHNR is extracted as the novel degradation indicator.Secondly,the degradation model of bearing is constructed on the basis of historical data,and then the UPF algorithm is used to update the model parameters in order to realize the tracking and prediction of the bearing degradation stage.Finally,the actual monitoring data of wind turbine bearings is taken as an example to validate the proposed method,the results show that our method can start the life prediction mechanism in time and effectively solve the problem of particle degradation in traditional particle filter(PF)algorithm.(5)A single signal source cannot fully reflect the degradation trend of bearings,influencing the RUL prediction precision.In this paper,a novel general log-linear Weibull(GLL-Weibull)model is presented.The model considers the influence of vibration and temperature monitoring signals on bearing deterioration by constructing covariates,and the prognosis method is divided into two stages.During the feature extraction stage,the relative root mean square(RRMS)is derived from the monitored vibration signal,and the relative temperature trend value is extracted from the monitored temperature signal to eliminate individual differences in bearings and random signal fluctuations.Then,a fuzzy operator is introduced to describe the degree of an "overheated bearing" and "excessive bearing vibrations".During the RUL prediction stage,both the extracted vibration and temperature characteristics are used to create the GLL-Weibull model.The best parameters are attained by employing the maximum likelihood estimation approach.The algorithm performance is checked with criteria like the root mean square error(RMSE)and the mean absolute percentage error(MAPE).