As-Built Inventory and Deformation Analysis of a High Rockfill Dam under Construction with Terrestrial Laser Scanning

The construction of large earth/rock fill dams, albeit its remarkable progress, still relies largely on past experiences. Therefore, a comprehensive yet dependable monitoring program is particularly beneficial for guiding the practice. However, conventional measurements can only produce limited discrete data. This paper exploits the potential of the terrestrial laser scanning (TLS) for an accurate inventory of as-built states of a concrete-faced rockfill dam under construction and for a full-field analysis of the 3D deformation pattern over its upstream face. For the former, a well-designed 3D geodetic system, with a particular consideration of the topography, promises a regulated acquisition of high-quality and blind-zone-free point cloud at field and also eases the cumbersome data registration process while maintaining its precision in house. For the latter, a problem-tailored processing pipeline is proposed for deformation extraction. Its core idea is to achieve a highly precise alignment of the point clouds with Iterative Closed Point algorithms from different epochs in datum areas that displays a featured, undeformed geometry at stable positions across epochs. Then, the alignment transformation matrix is applied to the point clouds of respective upstream face for each epoch, followed by pairwise comparisons of multiple adjusted point clouds for deformation evaluation. A processing pipeline is used to exploit the peal scene data redundancy of the GLQ dam acquired at six different epochs. Statistical analysis shows that satisfactory accuracy for deformation detection can be repeatably achieved, regardless of the scanner's positioning uncertainties. The obtained 3D deformation patterns are characterised by three different zones: practically undeformed, outward and inward deformed zones. Their evolutions comply well with real construction stages and unique 3D valley topography. Abundant deformation results highlight the potential of TLS combined with the proposed data processing pipeline for cost-efficient monitoring of huge infrastructures compared to conventional labor-intense measurements.