Multi-source onboard data-driven method for intelligent identification of subway track irregularities
To address the shortcomings of high-cost and poor-timeliness of track irregularity detection,an intelligent track irregularity identification method with multi-source vehicle data drive based on the correlation between vehicle dynamic responses and track irregularities was proposed.Firstly,a subway vehicle system dynamic model was established to obtain the data of vehicle vibration and motion attitude.Secondly,strong correlation data were selected through correlation analysis algorithm to make the dataset of network model.Finally,a network model was established through the combination of the convolutional neural network(CNN)and the long short-term memory network(LSTM)to detect track irregularities.A particle swarm optimization(PSO)method was used to optimize the neural network parameters.The results show that the vertical accelerations of the vehicle have a stronger correlation with the track irregularities compared to the lateral accelerations of the vehicle.At the same time,the motion attitudes of the car body,such as the angular velocity of pitch,have an significant correlation with the track irregularities.The proposed PSO-CNN-LSTM model has good performance in identifying track irregularities,and the determination coefficient of vertical and lateral track irregularities identification is 0.92 and 0.76,respectively.Moreover,the proposed method,namely PSO-CNN-LSTM,has better accuracy and timeliness than the classical fully connected neural network(FCNN)and support vector machine(SVR).
国家科技期刊平台
NSTL
万方数据
