Failure Detection Method of Vibration Isolator of Long-size Floating Slab Track Based on Structure Vibration Response
At present,the failure identification of vibration isolator of steel spring floating slab track mainly depends on manual inspection or visual imaging system,but such kinds of methods are inefficient,slow in detection,and costly.In order to improve the current unfavorable situation of vibration isolator failure detection,this paper proposed an automatic detection method.The method uses the residual learning idea and the basic theory of convolutional neural network to build a data classifier,which identifies the failure of vibration isolators for an in-service common long-size floating slab track through the structural dynamic response caused by the train dynamic load.Firstly,based on the vehicle-track vertical coupling dynamics theory,a vertical coupling dynamic simulation analysis model of subway vehicle and floating slab track considering different service states of vibration isolators is established.Then a data set under different operating conditions were generated the network training and performance testing.Furthermore,the influence of different sensor deployments on the detection performance is investigated and the appropriate sensor deployments are determined.Finally,an experiment was carried out in a actual subway line to validate the proposed method.The results indicate that the detection accuracy of the ResNet model can be significantly improved by optimizing the sensor deployment.The reasonable set of the number of sensors can also improve the detection performance of the network model.By properly selecting the sensor deployment,the proposed vibration isolator failure detection method based on the ResNet can achieve an accuracy of 98.99%on simulation data,and 96.33%on field data.This method has high feasibility and application potential,which can provide a technical reference for intelligent operation and maintenance of floating slab vibration isolators in subway transportation,and is expected to be applied in the future for automatic detection of vibration isolators failure.
NETL
NSTL
万方数据
