Research On Fault Diagnosis Algorithm For Rolling Bearing Of Urban Rail Train Running Gear

Abstract:As an important way of public transportation, urban rail transit has the characteristics of large capacity and high speed, the operating safety of which is directly related to people’s life. The rolling bearing of urban rail train running gear is not only a key component, but also a component which easily causes faults. Therefore research on rolling bearing fault diagnosis of urban rail train has important significance. Existing studies of rolling bearing fault diagnosis algorithm are mainly only based on the fixed condition, having certain limitations. In this thesis, a rolling bearing test bench is used to simulate rolling bearing operating conditions for a variety of experiments, and collects vibration acceleration signals for algorithm research. On the basis of the above, this thesis puts forward a bearing fault diagnosis algorithm which based on EMD and GNN-AdaBoost, and verifies effectiveness and availability of the algorithm, which can meet the demand of field urban rail train fault diagnosis.The main researches of the work are as followings.(1) Rolling bearing vibration signal features of simulating urban rail train under different operation condition are extracted through EMD. In order to collect rolling bearing vibration signals under different working conditions, including acceleration, deceleration, high speed, low speed, heavy load, light load, and different states, including fault-free, inside circle fault, outside circle fault, rolling body fault, many kinds of the experiment working condition scheme is designed. Vibration signal is decomposed into several IMF adaptively to extract the IMF energy moment. The fault identification characteristic parameters are constituted by combing the IMF energy moment and time domain parameters.(2) GNN-AdaBoost fault identification algorithm is proposed, which uses genetic neural network algorithm as sub-classifier of the boosting algorithm. Due to combining the advantages of genetic neural network and the AdaBoost algorithm, GNN-AdaBoost algorithm can identify rolling bearing fault state under different working conditions more accurately.(3) The algorithm based on the EMD and GNN-AdaBoost is used to diagnose rolling bearing fault under different conditions. On the working condition of load, speed factor change, comparing and analyzing the fault diagnosis results of GNN-AdaBoost algorithm and genetic neural network algorithm to prove that the GNN-AdaBoost algorithm has higher fault identification ability than the traditional genetic neural network under different working conditions.