Abstract
The quality of tunnel excavation is evaluated by comparing the excavated tunnel and the design model. Terrestrial laser scanning (TLS) provides surveyors with dense and accurate three-dimensional (3D) point clouds for excavation model reconstruction. However, sufficient attention has not been paid to incorporating design models for tunnel point cloud processing. In this paper, a technical framework that combines TLS point clouds and the design model for tunnel excavation evaluation is proposed. Firstly, the point clouds are sliced into cross-sections and the feature points are accordingly extracted. Then, considering the structure of the design model, feature point deficiencies are repaired by topological and parametric model interpolation. Finally, the excavation quality is evaluated in terms of the deviation of centerlines and 3D models. This method is validated in the case study. Experiments show that the deviation of centerline azimuth is acceptable but there remain considerable overbreak and underbreak, which respectively account for 20.6% and 11.2% of the design excavation volume.
Introduction
In tunnel construction, maintaining fidelity to the design model is critical for safety and cost-efficiency. Engineers must constantly monitor overbreak (excessive excavation) and underbreak (insufficient excavation), alongside deviations in the tunnel's centerline. While Terrestrial Laser Scanning (TLS) captures high-density 3D data, it faces significant challenges such as data gaps from occlusions (caused by equipment or personnel) and geometric ambiguity during surface reconstruction. Traditional methods often treat the design model only as a final reference for comparison. This research proposes integrating the design model early in the data processing pipeline to guide feature extraction and structural completion.
Technical Framework
The proposed framework consists of a three-stage pipeline designed to leverage the geometric priors of the design model to "fill in the blanks" caused by sensor occlusions.
- Cross-Section Feature Point Extraction: Point clouds are registered to the design coordinate system and sliced into cross-sections. Points are categorized by their proximity to specific design segments (floor, walls, ceiling). To filter noise from internal clutter like machinery, only the outermost points relative to the design model are selected as "feature points."
- Structure-Aware Feature Point Completion: To address occlusions, the system repairs missing data using the design model's logic. Planar interpolation is used for floors and walls, while a RANSAC-based parametric model fits the arched ceiling, ensuring the reconstructed model is topologically sound.
- Excavation Evaluation: Centerlines are fitted using a customized residual function with a smoothness term to ensure seamless transitions. Finally, Boolean operations between the completed excavation mesh and the design solid model quantify overbreak and underbreak volumes precisely.
Experimental Results and Discussion
The framework was validated at the extension project of the Wuqiang Creek Hydropower Station in Hunan, China, using a Leica ScanStation P40. The study area included an 820-meter tunnel section featuring straight lines and circular arcs.
The analysis revealed high accuracy in directional alignment but significant physical excavation deviations:
- Centerline Accuracy: The azimuth bias was remarkably low (0.017° for straight segments and 0.002° for curves). The curvature radius deviation was only 0.05 meters.
- Excavation Volume: The 3D comparison revealed that overbreak accounted for 20.6% (1730.9 m³) of the design volume, while underbreak accounted for 11.2% (943.1 m³).
The "structure-aware" completion successfully modeled the tunnel even in sparse data regions, proving that design-led reconstruction effectively handles noisy and incomplete point clouds.
Conclusion
This research demonstrates that integrating design models into the TLS processing workflow significantly enhances the reliability of tunnel excavation evaluations. By using engineering knowledge as a geometric prior, the method robustly filters noise and completes occluded data. Future work will focus on automating parameter selection and adapting the framework for mobile LiDAR systems to further improve on-site efficiency.
Summary updated