Study On Diagnosis Method For Bearing Faults Of In-wheel Motor Based On Fuzzy Artificial Hydrocarbon Networks

In-wheel motor has the advantages of high efficiency,fast response,and full-time wire control,which is an important basic technology for the future generation of electric vehicles and the promotion of industrial development.It is also an ideal power source for electric vehicles in the future.Drive system with in-wheel motors can effectively reduce vehicle energy consumption,improve vehicle performance,and optimize space layout,which has attracted extensive attention from the market.However,the special installation method of the in-wheel motor makes it difficult to install protective devices,and it is extremely easy to induce local faults.Once one or more in-wheel motors fail,the local output torque will fluctuate sharply,which will inevitably lead to the decrease of vehicle handling stability and driving comfort.In severe cases,it can cause traffic accidents and even life-threatening.Similar safety hazards have become a “blocker” in the marketing of electric vehicles driven by in-wheel motors.In order to give full play to the advantages of the in-wheel motor drive system and meet the "Technical Conditions for Safety of Motor Vehicle Operation"(GB 7258-2017),it is urgent to carry out research on the theoretical methods of fault diagnosis for in-wheel motor.In-wheel motor is a typical complex electromechanical system,and its fault types include mechanical faults and electrical faults.This thesis takes the common mechanical faults of in-wheel motor-bearing faults as the starting point,studies the fault feature extraction,data reduction and fault diagnosis methods under complex working environment and variable operating conditions,a fault diagnosis method based on fuzzy artificial hydrocarbon network is proposed to realize effective diagnosis of inwheel bearing faults.First of all,focus on the influence of speed and load torque on the vibration signal of the in-wheel motor under the actual operating conditions of the vehicle,a test bench for bearing faults of in-wheel motor was built.Faulty motors with common bearing damages including outer ring damage,inner ring damage and rolling element damage are customized,and corresponding test schemes are designed to collect monitoring signals of in-wheel motors at different speeds and load torques,laying the foundation for subsequent research work.Secondly,for the problem of difficulty in extracting fault information from the vibration signal of in-wheel motor bearing,a fault feature extraction method based on improved empirical mode decomposition is proposed.The original signal is divided into minimum units,and empirical mode decomposition is performed for each minimum unit signal.Through the similarity analysis of the intrinsic mode function for each unit in the frequency domain,fault information can be obtained.Then,for the problem that the symptom parameters calculated by the vibration signal have a large amount of data,high randomness and weak discrimination,a symptom parameter discretization method based on rough set theory is proposed.The symptom parameters can be roughly divided without destroying the algorithm’s resolution relationship,so as to speed up the running speed of the machine learning algorithm and improve the classification accuracy.Finally,for the shortcomings of the existing artificial hydrocarbon networks diagnosis model,a diagnosis method for bearing faults of in-wheel motor based on fuzzy artificial hydrocarbon networks is proposed combined with fuzzy theory.The membership function based on the output state value of the learned artificial hydrocarbon networks model is established as the follow-up diagnosis part of the model,and the fuzzy artificial hydrocarbon networks diagnosis system is constructed.By comparing with other classifier algorithms,the method proposed in this thesis has better generalization and robustness.