Research On Prediction Of Bearing Degradation Trend And Fault Pattern Recognition Based On Gradient Boosting Decision Tree

In winter,snowfall occurs in most areas of China,especially in the north.The snowfall lasts for a long time,which brings great inconvenience to people’s normal travel.Especially at the airport,the flight delay is affected by the snowfall weather.The three-in-one snow removal vehicle is one of the quick snow removal tools of the airport.How to ensure the safe operation of the snow removal vehicle and ensure the normal take-off and landing of the flight is very important.As a key component of the snow removal part of the snow removal vehicle,the speed reducer is easily damaged under frequent and severe impact conditions.Among them,rolling bearing inner and outer ring failure and rolling element failure are the main reasons for its damage.If the fault of the bearing is predicted in time,the spare parts in advance can not only save costs,but also ensure the replacement time of parts and improve the efficiency of snow removal.Therefore,it is crucial to predict the degradation trend of rolling bearings and fault pattern recognition.The thesis focuses on the prediction of bearing degradation trend and fault pattern recognition.The main research contents include:(1)Singular value decomposition(SVD)noise reduction.The problem of other signal components is also included in the collected bearing vibration signal.Based on the singular value decomposition and noise reduction theory,a lot of research on the effective rank determination is carried out,and the singular value fitting error minimum principle is used to determine and compare by method.Verify the effectiveness and superiority of the method.At the same time,SVD noise reduction and wavelet noise reduction are compared to further illustrate the effectiveness of the method.(2)Variational mode decomposition(VMD)study.Aiming at the problem of insufficient sample characteristics in bearing degradation trend prediction,the VMD decomposition method is adopted,which can effectively avoid modal aliasing and make the obtained features independent.In terms of modal aliasing,it is compared with empirical mode decomposition(EMD)to verify the effectiveness and superiority of VMD decomposition.At the same time,for the problem that the parameters in the VMD method are difficult to determine,the particle swarm optimization algorithm is used to optimize the parameters,thus avoiding the occurrence of modal aliasing and providing guarantee for the subsequent research.(3)Research on bearing degradation trend prediction and fault pattern recognition based on gradient boosting decision tree(GBDT).In order to further improve the generalization ability of the model and alleviate the over-fitting phenomenon of the model,the regularization of GBDT is improved,and the regular term is introduced and verified.In terms of pattern recognition,in addition to some commonly used time domain and time-frequency domain features,wavelet packet decomposition and sample entropy are combined to achieve high accuracy of fault identification.(4)Development of bearing degradation trend prediction and fault pattern recognition system.Determine the overall development framework of the system,and develop the user login module,degradation trend prediction module,fault pattern recognition module,and information management module.