With the development of inland waterway and sea-river interconnection projects, underwater borehole blasting is essential to channel improvement, harbor basin dredging, and cofferdam demolition. In waterway engineering projects, over and under excavation affect excavation quality, dredging workload, and slope trimming. Compared with tunnel blasting, underwater blasting is more difficult to measure and control because the blasting face is concealed, rock mass and water depth exhibit spatial variability, and blasting effects are influenced by drilling accuracy, charge structure, delay timing, subdrilling depth, and construction management. Therefore, a refined management approach integrating sensing, modeling, calculation, visualization, and feedback control is needed. Taking the Pinglu Canal waterway excavation project as the background, this study establishes a digital twin-based framework for controlling over and under excavation in underwater borehole blasting.

Based on the five-dimensional digital twin model, the Pinglu Canal digital twin system was structured into the physical, perception, transmission, virtual, and application layers. The physical layer includes blasting equipment, survey instruments, monitoring devices, construction personnel, and the site environment. The perception layer acquires geospatial, geological, construction-process, safety-management, and blasting-performance information. The transmission layer is responsible for data cleaning, storage, and exchange, while the virtual and application layers support model reconstruction, over and under excavation calculation, three-dimensional visualization, parameter optimization, and construction quality management. To establish the digital twin data foundation, high-precision scanning equipment was used to acquire terrain and bathymetric point-cloud data in the blasting area. An unmanned aerial vehicle equipped with LiDAR collected above-water slope and ground data, while an unmanned surface vessel equipped with a single-beam echo-sounding system collected underwater channel-bed data. The raw point-cloud data were processed according to the characteristics of waterway excavation projects. First, obvious outliers caused by mechanical vibration, floating objects, dust, or construction equipment were removed through denoising. Then, an octree-based thinning method was adopted to reduce data density while preserving the main terrain features, thereby improving processing efficiency. Finally, moving least-squares smoothing was applied to reduce local irregularities and improve surface continuity. On this basis, ContextCapture was used to generate a realistic three-dimensional scene model, and MATLAB was employed to construct a Delaunay triangulation model suitable for volume calculation. A mesh-based automatic algorithm was developed to calculate over and under excavation volumes. For each triangular element, the relative position between the measured terrain surface and the design surface was determined. Elements above the design surface were classified as under excavation, whereas those below it were classified as over excavation. Mixed elements intersecting the design surface were further decomposed into geometric bodies so that positive and negative excavation volumes could be calculated separately. The results were visualized in three dimensions, enabling engineers to identify the spatial distribution and concentration zones of over and under excavation after each blasting cycle. In the control stage, the digital twin platform stored over and under excavation results together with geological conditions, hole spacing, row spacing, charge per hole, subdrilling depth, hole depth, hole inclination, construction teams, and other key process parameters. Manual parameter adjustment was first conducted during the early blasting cycles to establish a training dataset. Artificial-intelligence-assisted analysis was then introduced to optimize multiple blasting parameters simultaneously and feed the updated results back into the dataset, forming a closed-loop process of measurement, evaluation, parameter adjustment, construction, and re-evaluation. The platform also supported refined management by recording borehole quality and construction responsibility by zone and team, which helped distinguish parameter-induced deviations from systematic deviations caused by construction management. The proposed scheme was applied to the underwater borehole blasting project of the Pinglu Canal. The results show that the digital twin system improved the quantification, visualization, and feedback efficiency of blasting quality. Compared with the initial construction stage, the over and under excavation volume was reduced by 80.25% after digital twin-supported parameter optimization and refined management were implemented. This study demonstrates that digital twin technology provides a feasible pathway for refined underwater blasting control in large-scale waterway engineering projects and offers a practical reference for intelligent blasting construction and quality management.