Study On The Decoupling,Diagnosis And Evaluation Method Of Wheelset-Bearing Faults Based On The Vibration Spectrum Characteristics Of Axle Box

Bogie is a key equipment that integrates the core functions of high-speed trains,including load-bearing,guidance,vibration reduction,traction,and braking,directly determining the operational quality and safety reliability of trains.Due to the continuous high speed,long duration,and wide geographical range of high-speed trains in China,as well as the complex and variable operating environment of the line,the service performance of key rotating components of the bogie deteriorates under strong alternating and large amplitude continuous working load conditions,leading to bearing fatigue wear,polygonal wear of wheel sets,and other faults.Periodic wheel rail excitation can exacerbate train vibration and structural fatigue damage,seriously affecting the smooth comfort and safety reliability of high-speed train operation.At present,the health status monitoring of rotating components of bogies mostly focuses on a specific component,and there is little systematic evaluation method research on the health status and development trends of different key rotating components in service.Therefore,this article conducted research on the identification and evaluation method of wheelset bearing health situation based on the spectral characteristics of bogie axle box vibration signals through a combination of theoretical modeling and line testing.The main work conducted in this thesis include:1)A multi-index fusion clustering technology for accurate recognition of multiple fault modes of axle box bearings is proposed for the health status assessment of bogie axle box bearings.Using multi feature fusion to construct a set of health evaluation indicators for axle box bearings,clustering unknown samples through data dimensionality reduction methods such as principal component analysis and kernel principal component analysis,achieving accurate recognition of multiple fault modes for axle box bearings;Based on the machine learning method combining HMM and GP,the error caused by feature vector quantization is suppressed within a controllable range.The fusion signal and probability are used as aging indicators for bearings to establish a fusion model to predict bearing life.The experimental results show that the prediction model is more effective than single feature prediction.Predicting accuracy of different bearing health and fault extent reaches 91%.2)In order to identify and evaluate the polygonal faults in the wheelsets of the bogie,a high-order polygon attitude evaluation method for the wheelsets of the bogie was established based on the mapping between the vibration characteristics of the axle box and the geometric shape of the wheelset.A theoretical model(mapping)was established by conducting line tracking experiments to evaluate the amplitude of axle box vibration characteristics and the degree of polygonal fault under a single operating speed level.High-order polygonal depth fault test points(grab points)were captured based on the speed level of the line test,and the polygonal fault level such as non-roundness and roughness was located using the theoretical model of vibration characteristic amplitude and polygonal fault degree evaluation.Wheelset polygonal axis bench tests at multiple operating speeds were conducted to obtain vibration characteristic amplitudes(migration)corresponding to the same level of fault.A highorder polygonal fault state evaluation method of axle box vibration characteristic amplitude for different operating routes and speed levels was established(reconstruction).The method was applied and verified through line testing experiments and whole vehicle rolling tests of 25th-order highorder polygons.The validation proves that the relative error of resonance feature recognition between proposed method and vehicle test is less than6.75%.Based on the principal component cluster analysis,the degradation index model for the first wheelset overhaul cycle polygonal state and its revision over the entire overhaul cycle was established.Based on the evaluation and prediction model of wheelset polygons,predictive maintenance recommendations via axle box acceleration were proposed.3)A multi-source fault decoupling identification and hierarchical evaluation method for bogie wheelset bearings is proposed to address the diagnostic evaluation problem under strong coupling of multiple faults in wheelset bearings.A multi fault coupling dynamic model for wheel set axle box bearings was established,and the influence of different polygonal grinding depths and bearing outer ring peeling degrees on the vibration characteristics of the axle box was studied.The narrowband influence characteristics of different levels of bearing faults on the resonance response characteristics of the axle box of the wheel polygon were analyzed.A physical characterization model of multi-level bearing outer ring peeling faults with the grinding depth of the wheel polygon was established based on the envelope demodulation feature of the axle box vibration signal;The corresponding order polygonal grinding depth is determined based on the measured resonance characteristic amplitude of the axle box on the line,and the demodulation feature threshold of the bearing fault level is further determined through this grinding depth,achieving graded evaluation of bearing faults.The effectiveness of decoupling between the wheel set polygon and the bearing outer ring peeling fault was verified through the rolling test of the entire vehicle bench.The recognition error of polygon grinding depth under the combined fault of the axle box bearing outer ring peeling and the wheel set polygon is within 6%,and the fault degree of the bearing relative to the calibrated faulty bearing can be accurately identified based on the divided demodulation feature threshold.