Diversion Tunnel Defects Inspection and Identification Using an Automated Robotic System

Diversion tunnels servicing the hydropower plant, hydraulic concrete defects of the tunnels can affect the safety, durability, and performance. Relative to the conventional manual inspection, robotic systems can complete the inspection process safely with objective results and high efficiency. This paper proposed a crawler teleoperation robot system (CTRS) for diversion tunnel structural inspection and a high-precision defects classification method suitable for hydraulic diversion tunnel defects. CTRS collected fewer but effective dataset in the MRock hydropower plant located in the southwest of China. The dataset is expanded by data enhancement, and the data set is divided into three sub-datasets for training. GoogleLeNet as feature extraction layer was introduced to modify the number of full connection layers of the network.The results show that each type of intact, repair and seepage can be correctly predicted. The accuracy of the final test set is 94.3%.