Research Projects

The behaviour of geotechnical structures is governed by the variability and complexity of soil and rock properties, shaped by geological processes across spatial and temporal scales. These properties are affected by uncertainty from both aleatory sources, representing inherent natural variability, and epistemic sources, reflecting limited or imprecise information. In practice, acquiring reliable data is often restricted by sample disturbance, high testing costs, site-specific constraints, and instrumentation limitations. As a result, conventional reliability methods, such as FORM/SORM, Point Estimate Methods, and Monte Carlo Simulations (MCSs), may yield biased or overly conservative results when applied to problems involving uncertain or incomplete inputs. This research aims to develop hybrid reliability methodologies for geotechnical structures that address mixed aleatory and epistemic uncertainties through integration of precise and imprecise probabilistic frameworks. The central hypothesis is that combining imprecise models, such as intervals, fuzzy sets, convex models, and probability-boxes, with conventional probabilistic methods enables more realistic and robust assessments of structural safety and design under data-limited conditions. The study begins with a review and classification of uncertainty modelling techniques into non-probabilistic, precise probabilistic, and imprecise probabilistic categories. Relevant geotechnical parameters will be compiled from laboratory tests, field investigations, and databases, and characterised to distinguish aleatory from epistemic uncertainties. Aleatory uncertainty will be represented using probability distributions, while epistemic uncertainty will be modelled using intervals, fuzzy sets, convex models, and probability-boxes. These will be embedded in conventional reliability frameworks, enabling safety evaluation without full probabilistic input definitions. The methodologies will be implemented through computational tools and applied to real geotechnical applications, including slopes, tunnels, and foundations. Their performance will be evaluated through comparison with deterministic and traditional probabilistic methods, and validated against observed in-situ behaviour where available. Global sensitivity analyses using variance-based and moment-independent indices will identify key influencing parameters, guiding data collection and model refinement. This research aims to establish unified frameworks for reliability analysis of geotechnical structures under mixed uncertainties. By enabling realistic treatment of imprecise input data, the developed methodologies will enhance the efficiency, transparency, and robustness of geotechnical design and safety assessments. The outcomes aim to advance the theoretical foundation of uncertainty quantification and support informed decisions in data-constrained environments.