Data-driven efficient geotechnical reliability analysis in spatially variable soils

• Applicant: Yue Hu, Ph.D.
• Subject Area: Geotechnics, Hydraulic Engineering
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 568506959

Geotechnical engineering deals with highly heterogeneous geomaterials (e.g., soils and rocks) that exhibit significant spatial variability and uncertainty. These uncertainties critically affect the performance of engineered structures (e.g., slope stability, excavation deformations), posing potential risks to public safety. Geotechnical reliability analysis is essential to quantify and manage these risks. With the European Commission’s push towards Industry 5.0, integrating big data, machine learning, and advanced computational tools into geotechnical reliability assessment is becoming increasingly important. However, practical implementation faces key challenges. Spatial variability is one of the most influential sources of uncertainty in geotechnical engineering, yet site investigation data are often sparse and insufficient for accurate reliability assessment. Meanwhile, modern machine learning models require large-scale training datasets that adequately capture geotechnical spatial variability and performance metrics—datasets that are rarely available in engineering practice. This project aims to develop an efficient geotechnical reliability analysis method tailored to specific target sites with spatially variable soils. The core innovation is leveraging both limited site-specific investigation data and extensive geotechnical databases to enhance predictive modelling and decision-making. The methodology is based on the premise that a given site shares similarities with certain database sites in terms of spatially variable soil conditions, allowing its performance to be inferred from recorded performance at analogous sites. A novel multivariate spectrum model is introduced to capture complex spatial variability across multiple geotechnical parameters, serving as a benchmark to connect target sites with database sites. By bridging data-driven methods and engineering practice, this research facilitates the digital transformation of geotechnical engineering, contributing to a more sustainable and resilient industrial sector.

• DFG Programme: Research Grants