Augmented intelligence framework for real-time ground assessment under significant uncertainty

Real-time assessment of unsaturated soils through deflection tests is challenging due to the complex effects of water and air in soil pores, which significantly impact test outcomes but are difficult to quantify, especially when key data like gravimetric water content and suction are incomplete or missing. While human expertise and intuition are valuable in high-pressure scenarios like ground assessment during soil compaction, they are prone to biases. AI-driven solutions excel at processing complex datasets but often require highly specialised inputs, which may not always be readily available. This paper aims to develop a robust and pragmatic approach to decision-support in ground assessment by combining human insight with AI's computational power and principles from unsaturated soil mechanics. This paper outlines key limitations of current ground assessment practices and discusses the challenges of developing reliable intuition when using deflection tests on unsaturated soils. To address these challenges, an augmented intelligence framework is introduced that leverages fuzzy human inputs for missing gravimetric water content information and incorporates a sophisticated self-improving mechanism to estimate missing suction data, based on insights gained during calibration. This framework significantly enhances ground assessment practices after validation using recent field trial data, particularly in highly uncertain unsaturated subsurface conditions. The study also demonstrates the framework's resilience in qualitative assessments, maintaining accuracy across a range of assumptions about missing gravimetric water content.