Study On Recognition Methods For Bearing State Of In-wheel Motor Based On BN And Improved DST

With the development of automobile industry and highway network,automobile has become one of the indispensable means of transportation.However,traditional fuel vehicles have the disadvantages of high energy consumption and low energy efficiency.Therefore,the state is vigorously promoting the development of electric vehicles.Compared with the centralized drive of transmission fuel vehicle,in-wheel motor drive system of electric vehicle has the advantages of flexible control and high efficiency,which is the ideal driving mode of electric vehicle.In-wheel motor drive system integrates driving,braking and load-bearing devices into the hub,which greatly simplifies the mechanical part of the vehicle.However,the variable driving conditions and complex road environment of the automobile not only induce the bearing failure of in-wheel motor,but also increase the difficulty of the bearing state identification of inwheel motor.Once in-wheel motor bearing fails,it will inevitably lead to the performance degradation.Light will increase the energy consumption of the vehicle,and the weight will affect the driving safety of the vehicle,even threaten the life of the driver.Therefore,it is necessary to study the status identification of in-wheel motor bearing,which will lay a theoretical foundation for the safety assessment of in-wheel motor drive system.In order to effectively identify the bearing state of in-wheel motor,this paper takes common bearing faults as the starting point,and studies the methods of signal processing,feature dimension reduction and state recognition.A new method of bearing state recognition is proposed based on Bayesian Network and improved Dempster Shafer Theory.Firstly,aiming at the problem that it is difficult to quantify the operating conditions and road environment in the real vehicle test,based on the real operating conditions of electric vehicles,a test bench for in-wheel motor bearing fault is built,and in-wheel motor with common bearing faults(such as rolling element fault,inner ring fault and outer ring fault)is customized.The data acquisition provides data support for the subsequent method of bearing state identification of in-wheel motor.Secondly,in order to solve the problem of difficulty to extract the state characteristics of in-wheel motor bearings under the strong interference environment,the paper proposes a method of state feature extraction based on Ant Colony Optimization to optimize the state feature extraction of resonance sparse decomposition.According to the impact characteristics of low resonance components and the smoothing characteristics of high resonance components,the ratio of kurtosis and smoothing index of high resonance components is taken as the objective function.Ant Colony Optimization is used to select superior-quality and low-quality factors,and then the bearing experimental data of Case Western Reserve University of America are decomposed by resonance sparse decomposition,which verifies the effectiveness of the method to extract the bearing fault state feature information.Then,in order to reduce the redundancy of data and improve the response speed of recognition,a feature dimension reduction method based on Pearson correlation coefficient and Gini coefficient is proposed to reduce the bearing fault feature dimension of in-wheel motor effectively.Based on signal characteristic parameters after dimension reduction,a Bayesian Network Based Model for bearing state identification of in-wheel motor is built.Based on the bearing experimental data of Case Western Reserve University of America and the fault test data of in-wheel motor bearings,the validity of the model is verified.Finally,in order to improve the accuracy of bearing state recognition,a new method based on Bayesian Network and improved Dempster Shafer Theory is proposed.The method combines Bayesian Network state recognition results based on vibration signal and noise signal,and further verifies the bearing state of in-wheel motor through K-fold cross validation.The superiority of the method is verified by the comparison of other methods.