AI-Enabled Digital Twin Modelling of Dynamic Soil Plug Movement During Caisson Installation
Please login to view abstract download link
Suction caisson foundations, commonly used in offshore wind farms, face installation challenges due to soil plug movements like heave and uplift, which can hinder reaching target depths. Current monitoring methods, such as single-beam echosounders (Sparrevik et al., 2015), may fail to detect these issues effectively. This research introduces a novel AI-enabled sensor system that integrates an intelligent 3D ultrasound point scanner with a self-learning Bayesian optimization algorithm to capture the most informative measurements with minimal effort. Gaussian process regression is used to predict a 3D digital twin model of the soil plug surface from limited data. Tested in lab-scale experiments simulating typical soil plug movements, such as heave and uplift, the system accurately tracked dynamic surface movements and produced reliable real-time predictions (Figure 1), enabling a robust digital twin model of the soil plug surface.
