The drainage and energy dissipation building is an important aspect of the water conservation and hydropower project, and its structural safety is linked to the safety of the whole project. For a long time, the drainage and energy dissipation buildings have been subjected to the erosion of high pressure and high speed water, which will inevitably cause damage to the concrete structure. The apparent damage to underwater concrete structures is concealed due to the structure's uniqueness and diversity. The traditional method of diver inspection or manual inspection following cofferdam draining takes a long time and is expensive and dangerous. Using underwater robots for unmanned inspection reduces personal risk. However, the underwater robots' detection accuracy is limited due to the sediment accumulation on the bottom and poor visibility in the water, which makes it impossible to conduct timely investigations of defects and hidden dangers. This study has developed a desilting replacement module that is suitable for the conditions of sediment bottom and turbidity water and studied the mechanism design of the module, the efficiency of silt removal, and defect detection.

The desilting replacement module is built in this study by examining the starting condition of sediment deposition and the features of the submerged water jet. It consists mostly of the desilting mechanism, the replacement detection mechanism, and the lifting mechanism. The Euler multiphase flow model was used to create the continuity equation and momentum equation of water and sand, and the hydrodynamic influence of the desilting replacement module was investigated. A simulation model based on Euler water-sand two-phase flow was developed using computational fluid dynamics software to mimic the desilting detection process of the desilting replacement module in the underwater sediment environment. The thickness of the sediment is considered to be 100 mm in the simulation, and the height of the sediment deposited at the beginning distance of the replacement detection shell was used as a variable to evaluate the status of the desilting replacement module when the detection effectiveness is optimal. Finally, the simulation findings are compared and examined by combining them with the real experimental data.

This study verified the necessity of each mechanism in the desilting replacement module and concluded that when the initial height of sediment from the bottom of the replacement detection shell was 60 mm, the desilting detection efficiency was the highest, and the sediment of 100 mm thickness in the detection area could be removed to the remaining 10% within 1.56 s, and the total time of "desilting and detection" was 9.56 s.

The silt removal replacement module's novel design tackles the underwater detection problems caused by underwater sediment accumulation and turbidity. The desilting replacement module may be carried on most existing underwater detection robots, which can significantly improve the detection ability of underwater robots in turbidity water environments. It can achieve short-term efficient single point or long-term continuous visual image acquisition in the sediment environment, depending on the operation requirements, which has a great promotion role for the application of underwater robot detection technology in the water conservation and hydropower industries.

Underwater building safety inspection is a vital tool for ensuring reservoir dam operation safety of hydropower stations. Due to the high expense of traditional drainage inspections and the high safety risk of divers performing underwater inspections for areas that cannot be drained, underwater detection technology provides a new solution to underwater building safety inspection.

Using the Yalong River basin step operation hydropower stations as a typical scenario, the target water environment and structural boundary conditions were analyzed; combined with open water underwater inspection experience and technical differences, underwater robot inspection equipment and essential technology systems under various scenarios of hydropower stations were constructed. Through robot system integration and testing, a small-sized, powerful, multi-sensor fusion cable remote control submersible was developed for tunnel-like structures with lengthy cavern lines. A high-strength, zero buoyancy photoelectric composite umbilical cable with a small diameter was designed, and remote power supply and fiber optic real-time communication were carried out through high-frequency medium voltage transmission technology. The real-time monitoring system for underwater robot movement based on the virtual exercise platform was built, and the robot was directed to return autonomously by an adaptive control algorithm in the case of communication and power supply failure. The combined inertial navigation-based positioning technology was investigated, and the position information, such as structural seams and feature markings detected by sonar and camera, was used to calibrate the inertial guidance positioning information. A high-frequency three-dimensional real-time sonar system with 360° mobility, continuous scanning, and real-time generation of a three-dimensional point cloud model was jointly developed, which improved precise positioning capability and detection efficiency for large cross-sections and long-distance closed structures. The research offers a defect identification technology based on dynamic feature distillation to intelligently identify the inspection images and uses the joint analysis method of inspection information and a 3D digital model to conduct intelligent spatiotemporal correlation analysis of defect information. Intelligent unmanned vessel systems and RTK-GNSS combined positioning technology are introduced for underwater inspection in semi-open waters in high mountain canyon areas. A joint inspection scheme of multibeam sonar and underwater robots is proposed to make full use of the advantages of fast and high-density scanning of multibeam sonar and fine inspection of robots to improve inspection efficiency.

The robot system successfully passed through the access channel with only 2.1 m in diameter, the diversion tunnel with 12.0 m in diameter and 2.30 km in length, and the pressure pipeline with a 250.0 m vertical drop in the practice of normalized underwater inspection of multi-water scenes at Yalong River Basin Hydropower Station. The flaw detection accuracy reached the millimeter level, the plane positioning precision reached 10 cm, and the axial positioning accuracy was better than 1‰ of the traveled distance. The intelligent recognition rate of common defects in concrete structures, such as broken or exposed bars and cracks, reached 96.5%. The intelligent unmanned vessel successfully inspected various types of semi-open waters, such as tailwater channels of hydropower stations with a flow speed of 2 m/s and turbulent flows, obtaining three-dimensional topographic maps of underwater environments and identifying the distribution characteristics of underwater defects with a horizontal positioning accuracy better than ±8 mm+1 ppm and a vertical orientation better than ±15 mm+1 ppm.

The underwater intelligent inspection system features a high degree of innovation and integration, a large number of practical cases, safe and reliable inspection, and accurate and intuitive results. The system has effectively guided the safe operation and maintenance of underwater structures at the Yalong River Basin Hydropower Station and significantly reduced the cost of traditional inspection, which has important significance for industry promotion.

Headrace tunnels are key structures of major projects characterized by long tunnel lines, large tunnel diameters, high water pressure, and complex surrounding rock geology. Typical defects, such as cracks, landslides, and exposed reinforcement, will occur during long years of operation. If they are not prevented, the safe operation of the project will be seriously affected. Long cycles, high safety risk, high leak rate, and insufficient information are all issues with traditional manual inspection. Given the urgent need for regular inspection of large-diameter and super-long headrace tunnels in super-large water conservancy and hydropower projects, this study solved key scientific issues, such as the adaptability of robot underwater environment tasks, the active detection of super-long headrace tunnel apparent defects, and the safety risk assessment of tunnel structures based on robot inspection data. The key technology breakthroughs include the sub-parent cooperation of complex underwater environments, the fine operation of load manipulator, ultra-long distance underwater high-voltage power supply, umbilical cable safe release and recovery, ultra-long distance human-machine cooperative control, special environment adaptation of underwater robots, active defect detection and identification based on multi-sensor fusion. Structural safety classification, risk analysis and evaluation, and virtual drills were also carried out. The developed underwater robot inspection system was successfully applied to large-diameter and long headrace tunnels for comprehensive verification.

The application performance of underwater robots in special environments has improved due to breakthroughs in key technologies such as remote power supply, cooperative operation, intelligent patrol inspection, defect identification, and safety assessment of robots in complex underwater environments including water turbidity, high water pressure, adhesion and siltation, and local accessibility difficulties. The safety classification and risk assessment of the headrace tunnel structure are completed through the research and development of the multi-function "sub-parent" underwater robot system, and the whole process integration of "inspection, inspection, control, diagnosis, and use" of the underwater robot is realized, which has been demonstrated and verified in the eastern route of the South-to-North Water Transfer Project, Jinping Ⅱ Hydropower Station, and other major national projects, to improve the intelligent degree of the inspection of the headrace tunnel of large water conservancy and hydropower projects and support the safe operation of large projects.

The research findings can significantly improve the accuracy of the headrace tunnel inspection, reduce the headrace tunnel inspection cost, and improve the guaranteed rate of the safe operation of large water conservancy and hydropower projects; promote the interdisciplinary integration of artificial intelligence and water conservancy disciplines to form interdisciplinary advantages; promote the application of robots in special environments, especially in the inspection of headrace tunnels, and guide the development of robots in special environments; promoting the application of artificial intelligence and intelligent management of water conservancy projects, as well as improving the level of technology and equipment in relevant fields in China and cultivating a large number of versatile talents, will have significant social, economic and scientific values.