Fault Diagnosis Of The Train Wheelset And Crack Parameters Identification

Wheelset is a core part of railway trains,and its status is vital for the safe operation of trains.The wheelset is in the bad working condition over a long period,continuously bearing the influence of alternating load,and therefore it is very easy to germinate fatigue cracks and other wheelset faults.If the wheelset cracks are not detected and eliminated in time,the continuous crack propagation may make the axles broken,thereby endangers the safety of trains,and causes casualties and economic losses.Accordingly,during the operation of the trains,online detection of wheelset cracks and other faults are carried out to realize the diagnosis of fault types and the quantitative identification of crack parameters,which is of great significance to improve the safety performance of railway trains.A EMU wheelset is taken as the research object in this thesis.Firstly,a wheelset model with a breathing crack under the excitation of random track irregularities is established and the influence of the different parameters on the vibration characteristics of the wheelset is studied based on the finite element method.Then the Light-GBM algorithm is used to diagnosis the fault status of the wheelset,and this method is verified through experiment.Finally,a method for identifying wheelset crack parameters based on Kriging surrogate model and NSGA-III is proposed.The multi-body dynamics of the whole vehicle is simplified and Simpack is used to establish a vehicle-track dynamics model,and calculate the characteristic curve caused by random track irregularities.A finite element model of the cracked wheelset is established based on Abaqus,and the contact relationship is used to simulate the breathing mechanism of cracks.The correctness of the model was verified by comparing the free mode shapes and natural frequencies,and the vibration response of the cracked wheelset under the excitation of track irregularity is simulated.The effects of track irregularity excitation,crack depth,crack position and wheelset operating speed on the vibration response characteristics of the wheelset are studied.The results show that the frequency spectrum of the vibration signal of the cracked wheelset includes 1X,2X,3X and other harmonic components.The random irregularity excitation will increase the amplitude of the superharmonic components.The 1X component can accurately reflect the change of crack depth and be used for diagnosing the degree of crack damage.Furthermore,cracks near the wheel will cause a strong wheelset vibration response,and the 1X and 2X components are sensitive to the position parameters of the cracks.Finally,the higher the running speed of the wheelset,the more conducive to the diagnosis of cracks.When the running speed of the wheelset is greater than 300km/h,the degree of crack damage can be diagnosed through 1X and 2X harmonic components.A dynamic model of wheel flat and wheel out-of-round is established,and the influence of two types of faults on the axle crack diagnosis is analyzed.The Light-GBM algorithm is used to diagnose and classify the fault status of the wheelset.The results show that wheel flat and wheel out-of-round can cause significant changes in the harmonic components of the vibration signal,interfering the diagnosis of cracks based on the vibration signal.The wheelset fault diagnosis method based on the Light-GBM algorithm improves the diagnosis efficiency,and the fault recognition rate reaches97.4%.The accuracy of the method is verified by the wheelset fault test bench.Aiming at the identification of wheelset crack parameters in operating conditions,the Kriging surrogate model is used to establish the mathematical relationship between the crack position,depth parameters and the amplitudes of the 1X harmonic components of the wheelset vibration response.The problem of identifying wheelset crack parameters is converted into an improved multi-objective optimization problem,and the NSGA-III is used to find the optimal crack parameters.The results show that the accurate prediction of wheelset vibration response through the surrogate model replaces the time-consuming finite element simulation process.The improvement of the fitness function optimizes the crack parameter identification process in terms of speed and accuracy.Through verification,the crack location can be accurately identified based on this method,and the identification accuracy of the depth parameters can reach 96.84%.The work in this thesis provides new ideas and theoretical basis for the online detection of wheelset faults,and it can realize wheelset fault status diagnosis and crack parameters identification in actual operating conditions,which has certain engineering significance.