Condition Diagnosis Methods For Wind Turbine Bearings Based On Hidden Markov Models

Much power loss and high maintenance cost will be brought by the fault of the wind turbine.And the rolling bearing is one of the most easily damaged parts of rotating machinery in the wind turbine.Therefore,studying how to identify the bearing condition of wind turbine accurately and timely can take measures to avoid it,which is significant for maintaining normal operation of wind turbine.Taking the rolling bearing of wind turbine as the research object,aiming at sensitive and accurate diagnosis of bearing condition,this paper puts forward three kinds of diagnosis methods for bearing condition of wind turbines based on hidden Markov model(HMM).1.A condition diagnosis method based on fuzzy scalar and discrete hidden Markov model quantization is proposed.Based on fuzzy theory,the method improves the discrete data method of discrete HMM and adds membership degree in scalar quantization,so as to weaken the influence of data discretization information loss on discrete HMM condition recognition.The experimental results show that the method can weaken the influence of data loss of data discretization and make accurate bearing condition recognition.2.A condition diagnosis method based on stacking integration algorithm and continuous hidden Markov model is proposed.According to the stacking algorithm in the ensemble methods,the method integrates the continuous HMM recognition results of various features to reduce the influence of the bias of feature extraction on the diagnosis results.The results show that this method can avoid discretization of information loss and bias influence of feature extraction,and make it more sensitive and accurate.3.A condition diagnosis method based on genetic algorithm and coupled hidden Markov model is proposed.The method uses genetic algorithm to optimize coupled HMM parameters,reduce the computation of coupled HMM parameter optimization,and get better parameter combination in a relatively short time to make sensitive and accurate condition recognition.The results show that the method can fuse multi-channel information and can take into account the accuracy and sensitivity of fault condition recognition.