Study On Hidden Markov Model Based Rolling Element Bearing Performance Degradation Assessment And Life Prediction

With the rapid development of science and technology,modern industrial equipment turns to be high-speed,high-precision and with heavy-load.The operation conditions of mechanical equipment has become increasingly complicated and harsh at the same time.Especially that more and more fierce market competition makes the reliability and the efficiency of equipment more important.Abrupt breakdown of mechanical equipment greatly reduces the efficiency of industrial production and equipment reliability,and it can also cause a serious security problem even personnel casualties at the same time.Therefore,the real-time monitoring and health maintenance of mechanical equipment is still a focus of research work and practical application.Condition based maintenance as an emerging equipment maintenance method in recent years,includes two aspects of diagnosis and prediction.With the improvement of intelligent production process and the increase of reliability requirements,failure diagnosis as an after-event analysis has been unable to meet the needs of industrial production.Prediction techniques expects further development.As two important methods of prediction techniques,performance degradation assessment tries to assess the current state of degradation through real-time monitoring data,and life prediction is to estimate the equipment remaining useful life.The first one focus on assessing the deviation degree of the current state and the normal state,while another one focus on prediction of the time interval from the current state to the failure state.They have different focuses.However,results of the performance degradation assessment can assist the residual life prediction and the residual life prediction results can improve the forecast system.Hidden Markov model is a statistical model based on time sequences,which is very suitable for the equipment degradation process modeling.Therefore,study on the hidden Markov model based performance degradation assessment and life prediction is carried out,and the main contents include in several aspects:(1)From demands of the actual industrial production and the trend of the present research,the thesis expounds the origin and the significance of this research work.It summarized the research progress of performance degradation assessment and life prediction methods at home and abroad in recent years.The research status and practical applications of hidden Markov model are analyzed,and the research points of this thesis are established in view of the practical application.(2)Hidden Markov model is introduced in detail,which covers the basic theory,the basic algorithms and the common models.The diagnosis methods based on hidden Markov model are presented.These part is the foundation of the follow-up work.(3)Geared to the needs of performance degradation assessment and life prediction,in view of practical applications,improved algorithms are proposed for applications based on hidden Markov model.The nuisance attribute projection based transformation method is proposed.Nuisance attribute projection aims to eliminate redundant information of the feature space in the practical application and improves the target attributes in space.The state duration distribution of traditional hidden Markov model is fixed to be the geometric distribution,which limits its applications in life prediction.Extension model of hidden Markov model is proposed as hidden semi-Markov model.Improved algorithms of hidden semi-Markov model are introduced,which makes the model is more suitable for the life prediction.(4)The channel interference elimination algorithm based on nuisance attribute projection is introduced into the hidden Markov model based performance degradation assessment method.Nuisance attribute projection can eliminate the nuisance attribute information in feature space caused by channel interference,which can effectively improves the effectiveness of the performance degradation assessment method based on hidden Markov model.The accelerate life test data verifies the effectiveness of the proposed method.(5)The operation condition interference elimination algorithm is introduced into the hidden Markov model based performance degradation assessment method.The operation condition interference elimination algorithm is based on nuisance attribute projection.With this method,the redundant information influenced by the operation condition can be removed out from the feature space.The simulation data and the accelerate life test data verify that the proposed method is feasible and effective.(6)The remaining useful life prediction method based on nuisance attribute projection and the state remaining duration estimation is proposed.The method is based on the advantages of hidden semi-Markov model on the state duration distribution.The feature transformation algorithm based on nuisance attribute projection can improve the effectiveness of feature space.Combined nuisance attribute projection and hidden semi-Markov model,the method is more effective and accurate.The accelerate life test data of PRONOSTIA verify that the proposed method is feasible and effective.In summary,this thesis researches on the performance degradation assessment and life prediction of practical applications.The hidden Markov model based methods are summarized,improved and completed.The research work effectively promote the applications of hidden Markov model in equipment intelligent performance degradation assessment and life prediction.