Research On The Diagnosis Of The Global Wheel Irregularities Of Urban Rail Trains Based On The Track Vibration Signal

Urban rapid development is inseparable from the urban rail transit,urban rail transit plays an increasingly important role in the development of the city,so the safety of the urban rail transit train also receives more and more attention.Traveling system is one of the key components of the urban rail train,and the wheelset is an important of the traveling system,it supports the whole weight of the train and keep the train on the track.Therefore,it is of great significance to carry out on-line testing of the urban rail train wheel failure.Based on the research of predecessors,a method of urban rail train wheel out-of-round(OOR)failure detection based on track vibration signals is proposed,and the main content of this paper is as follows:Firstly,the main components and functions of the wheelsets of the urban rail train are introduced.The wheel out-of-round failure and its formation mechanism are especially introduced.EMD decomposition algorithm is also introduced.The acquisition system is briefly introduced as the data source.Secondly,the urban rail train dynamics model is accomplished.The model includes vehicle model,track model and wheel-rail dynamics model.The math expressions of OOR wheel and track irregularity are studied and input as model excitation,the track vibration response is obtained.Thirdly,aiming at the problem that the original vibration signal contains noise,an adaptive morphological filtering algorithm is proposed to filter the vibration signal.In order to decompose the intrinsic mode function components containing the fault information in the original vibration signal,an improved EMD decomposition algorithm is proposed and the mode mixing problem of the EMD algorithm is improved.The model simulation signal and on-site test signal are decomposed to verify the effectiveness and correctness of the decomposition algorithm.Finally,this paper focus on the fault diagnosis of wheel out-of-round failure based on LM-BP neural network.The method extracts multiple eigenvalues of the intrinsic mode functions as the input vectors of the LM-BP neural network and trains them.The simulation signal of the model and the on-site test signal verify that the proposed method can correctly identify the wheel out-of-round failure correctly.