Session Description:

This session will explore innovative approaches to probabilistic site characterization that leverage multi-source heterogeneous data. A wide range of datasets, such as cone penetration test (CPT) logs, standard penetration test (SPT) results, geological maps, images of soil profiles, geophysical data, tunnel boring machine (TBM) operational data, and remote sensing data, can be integrated to quantify the soil stratification or spatial variability of soil properties. By treating these diverse data as a unified system, the session will demonstrate how to quantify, model, and simulate the uncertainties of subsurface soil properties and stratigraphy, and apply these insights to the design and analysis of underground geotechnical structures. The session will cover the following topics:

• Data harmonizing techniques for multi-source data with different scales and formats
• Global or local database development for various soil or rock properties
• Machine learning for digital geological and ground modelling
• Spatial variability estimation and efficient simulation of large-scale random fields
• Optimization method of site investigation
• Probabilistic reconstruction of 2D/3D sparse measurement sites
• Applications of the probabilistic site characterization method to geotechnical structures such as slope, tunnel, and retaining wall
• Geological anomaly forecasting in TBM tunnelling by fusing multi-source data
• Practical case studies demonstrating the benefits of probabilistic site characterization for geotechnical engineering projects