Study On Tunnel Dynamic Deformation Detection And Analysis System

Deformation detection of the metro tunnel is essential to the train driving safety. In view of the fact that the traditional artificial static detection is not suitable for metro tunnel in operation, this thesis designed a vehicular tunnel dynamic deformation detection and analysis system based on the principle of laser scanning ranging, which can detect and analyze metro tunnel deformation effectively, and realize the monitoring and early warning of the deformation disaster.Firstly, the thesis introduces the working principle of the system and the overall design scheme; and then introduces the design of data acquisition system, including the selection of main sensor, the data acquisition of LMS500 and the data transmission of USB; then introduces the design of data processing algorithm, including multi-sensors relative positioning algorithm and tunnel modeling algorithm based on RBF neural network. Finally verify the feasibility of the system through experiments.In order to adapt to the train with high speed, the thesis adopts high-speed scanning laser rangefinders and high-speed laser displacement sensor to upgrade the original data acquisition system, and designs a new data acquisition module in the FPGA, which enhances the universality of the system. This thesis also uses USB interface to improve the speed of data transmission. The thesis achieves data acquisition of each sensors with real time and high speed, and consummate the original system.This thesis puts forward a set of tunnel deformation detection and analysis algorithm, including multi-sensors relative positioning algorithm and the tunnel modeling algorithm based on RBF neural network. Multi-sensors relative positioning algorithm uses existing encoder, laser scanning rangefinders and laser displacement sensor of the system to obtain the kilometer post, tunnel feature data and track feature data at no additional cost. The relative positioning algorithm achieves rough alignment positioning through kilometer post and tunnel feature data, and achieves precise alignment positioning through kilometer post arid track feature data. Finally the relative position with high precision, high reliability and high accuracy is finished. The algorithm solves the problem that the cyclical detection data in the same location does not match, which is caused by limited positioning precision of the single encoder, achieving a cyclical detection data fusion, thus lays the foundation for the establishment of subsequent tunnel model. Modeling algorithm based on RBF neural network through multi-cycle training fused 3D point cloud data of the tunnel, establishes multiple RBF neural network to optimize the tunnel model constantly, finally builds the tunnel lining model and special region model. The tunnel deformation detection and analysis algorithm compares the actual detection data with the exporting model data to complete the deformation analysis.To verify the feasibility of the system, experiments were conducted in metro tunnel and laboratory track respectively. Experimental results indicate that the system can effectively carry out the relative positioning and build the tunnel models, with the relative positioning accuracy of ± 10cm and deformation analysis precision up to ±10 mm. Finally it realizes the tunnel contour dynamic deformation detection and analysis, and reaches the expected requirements.