Displacement analysis of point cloud removed ground collapse effect in SMW by CANUPO machine learning algorithm

In this paper, a three-dimensional laser scanner was used to monitor the displacement of retaining structures for excavation, including a ring beam and a reinforced soil mixing wall (SMW) at an open excavation site. Eight scans of the retaining structures were taken before and after the excavation. Three-dimensional point clouds of the retaining structures produced with these scans are registered and analyzed to determine displacements of the retaining structures. Cloud to Mesh (C2M) method is used to identify displacement along the length of ring beam and depth of the SMW. The displacement obtained is then validated against displacement measured by the total station. The surface of the ring beam is flat so that the displacement of the ring beam can be estimated by the C2M. There was soil collapse and deposition on the surface of the SMW, which can affect the estimation of displacement in the cloud comparison. Therefore, a CANUPO machine-learning algorithm is applied to detect these areas and to remove the point clouds affected by soil collapse and deposition. The displacement is re-estimated based on the revised point clouds. The re-estimated displacement by the CANUPO method is more uniformly changed with the depth of the SMW than results without the CANUPO method, and the horizontal displacement due to excavation can be calculated.