Integration of robotic complexes into the processes of inspection and monitoring of the technical condition of underground structures

Abstract

The article is devoted to the implementation of robotic complexes in the inspection and monitoring of underground structures operating under conditions of aggressive soil environment, vibrational impacts, and limited accessibility, where manual control demonstrates insufficient discreteness of observations and vulnerability to subjective errors. The transition from episodic inspections to predictive maintenance based on continuous accumulation of heterogeneous data and construction of digital twins is considered. A system diagram of the application of mobile platforms of various classes (wheeled with enhanced cross-country capability, tracked with magnetic retention, unmanned in a protective frame) with a unified set of sensors, including high-resolution LiDAR, thermal imaging module, ground-penetrating radar with 400 and 900 MHz antennas, and a camera with circular illumination, is presented; the specifics of navigation in an environment without satellite signals and requirements for inertial modules and SLAM localization are analyzed. Based on inspections of 58 facilities with a total length exceeding 145 km, a dataset exceeding 24 TB (point clouds, thermograms, radargrams, photographic data) was formed and compared with parallel expert verification using manual non-destructive testing methods. A noticeable increase in the probability of detecting micro-cracks and hidden defects (including rebar corrosion and voids behind the lining) with acceptable metrological characteristics is shown, while the interpretation of ground-penetrating radar data remains sensitive to medium heterogeneity and requires more robust calibration procedures. The compromise between traversal speed and scanning density is quantitatively revealed: UAV ensures maximum productivity for geometric control, whereas the comprehensive tracked robot produces the most detailed models at the expense of a sharp increase in data volume and post-processing time. The economic model records an increase in direct costs for inspecting 1 km of collector with robotization but demonstrates an advantage in terms of cost per unit of validated information and a reduction in the need for field personnel. The influence of humidity on LiDAR data noise and the effectiveness of its compensation by filtering, as well as the synergy of photogrammetry and laser scanning that enhances 3D reconstruction accuracy, are established; forecasting of degradation using digital twins is characterized by higher reliability compared to expert assessments. Organizational barriers (regulatory legitimation of digital results, requirements for storage formats and evidentiary significance of the digital twin), engineering prerequisites for design for inspection , and the need for reference markers to correct positioning drift are discussed, enabling an assessment of the practical applicability of the proposed solutions in the operation of urban underground infrastructure.