Session Description:

The digital era is reshaping uncertainty quantification and reliability analysis in geotechnical engineering. This session showcases methods that integrate machine learning and big-data pipelines with physics-based and probabilistic models to quantify, propagate, and reduce uncertainties in geotechnical analysis, design, construction, and operation. Emphasis is on approaches that are transparent, data-efficient, and deployable in practice—e.g., real-time reliability updating with sensor networks and digital twins, Bayesian inference and active learning for targeted tasks, and hybrid AI–mechanics models for complex ground–structure interaction. Aligned with 5ISMLG and ISSMGE TC309, the session connects methodological advances to field applications in geotechnical infrastructure risk assessment and management, highlighting how the digital era drives the development of paradigms, methods, and applications for geotechnical uncertainty quantification and reliability analysis, enabling safer, more sustainable, and cost-effective geotechnical-related decisions. This session will cover the following related topics, including, but not limited to, the following:

• ML-assisted UQ for soil/rock properties and spatial random fields
• Physics-informed and surrogate models (e.g., PINNs, GP, ensembles) for UQ & sensitivity
• Integration of digital twins and sensor data for real-time uncertainty reduction
• Bayesian methods and optimization techniques in geotechnical reliability
• Probabilistic modeling and simulation using AI for geotechnical reliability analysis
• Big-data/IoT-driven monitoring and real-time reliability updating
• Digital twins for dynamic reliability- and risk-based decision-making
• Case studies on applying new paradigms and methods to infrastructure resilience and sustainable geoengineering