Deep Migration Learning-Based Detection of Structural Diseases in Railway Tunnel Lining Structures

It is difficult to quickly and accurately detect defects such as voids, voids, and non compactness in the lining structure of formed railway tunnels. We have conducted in-depth research on the propagation law of ground penetrating radar electromagnetic waves in tunnel lining structures and intelligent detection methods based on deep learning. The finite difference method was applied to numerically simulate the propagation process of electromagnetic waves in tunnel lining structures with different diseases. B-scan radar images were obtained, and the characteristics of voids, voids, and non dense images were summarized. On this basis, a sample library composed of numerical simulation and real radar images was established, and a deep migration method for detecting railway tunnel lining structural defects was proposed. Automatically learning complex disease features and updating network model parameters overcomes the limitations of limited real disease sample sets for tunnel lining structures. The experimental results show that the network after transfer learning can accurately detect defect bodies in actual tunnel lining structures, verifying the effectiveness and feasibility of the proposed method.