Data-driven analysis of soil consolidation with prefabricated vertical drains considering stratigraphic variation

In coastal cities such as Hong Kong, rapid reclamation using prefabricated vertical drains (PVDs) is preferred as it can accelerate land supply to meet the urgent demand for houses. A robust PVD design relies on the correct identification of permeable soil layers and accurate delineation of their stratigraphic connectivity with surrounding drainage boundaries. The current engineering practice often ignores the physical locations of minor drainage boundaries (e.g., sand lenses) in the subsurface stratigraphy and might lead to a false interpretation of potential drainage and consolidation mechanisms. In this study, a data-driven analysis framework that takes stratigraphic uncertainty into consideration is proposed to investigate the spatiotemporal consolidation of PVDimproved ground using sparse site investigation data often encountered in engineering practice. The method adaptively develops multiple geological cross-sections from limited measurements and prior knowledge that is reflected by a single training image. The resulting multiple geological realizations serve as the input for PVD analysis. The proposed data-driven framework allows for a probabilistic evaluation of soil spatiotemporal behavior in terms of the degree of consolidation and future ground settlement. More importantly, the proposed method accurately predicts the spatial distribution of stratigraphic boundaries with quantified uncertainty, which significantly influences the consolidation mechanism.