A method and system for identifying subway track geometry diseases

Abstract

The application discloses a subway track geometric disease identification method and system, and relates to the field of subway track disease identification.The method comprises the following steps: obtaining vehicle body vibration data and vehicle body information of a target subway track; segmenting the vehicle body vibration data of the target subway track, converting each segment of data based on a root mean square value method, and obtaining multiple equal-interval data segments of the target subway track; obtaining detection data of the target subway track according to the multiple equal-interval data segments of the target subway track and the vehicle body information of the target subway track; inputting the detection data of the target subway track into a track geometric disease identification model, and determining whether the target subway track has track geometric diseases and the disease level when the track geometric diseases exist; wherein the track geometric disease identification model is obtained by training a two-dimensional convolutional neural network based on a differentiable architecture search method using training data.The application can improve the identification accuracy and efficiency of subway track geometric disease identification.

Images