Adaptive GeoBIM frameworks for excavation. Integrating construction machine data into an uncertainty-based model for the underground

Accurate modelling of subsurface conditions is essential for efficient and sustainable geotechnical excavation, yet current practices remain constrained by the sparse and point-specific nature of conventional site investigations. Boreholes, soundings, and laboratory tests provide critical soil information during design but, when interpolated across heterogeneous terrain, they introduce uncertainty that can compromise excavation control, productivity, and cost.

To address these limitations, this dissertation develops and evaluates an adaptive GeoBIM framework that integrates machine-sensed excavation data with probabilistic uncertainty modelling to refine soil boundary predictions continuously during construction.

The research comprises three interlinked studies. The first examines whether hydraulic-cylinder pressure measurements collected during excavation can reliably indicate soil boundaries. Field trials showed strong agreement between pressure-derived boundaries and static–dynamic penetration test (SDPT) results, demonstrating that standard excavators retrofitted with low-cost sensors can function as geotechnical investigation instruments.

The second study introduces a probabilistic uncertainty-modelling method combining Kriging interpolation with Monte Carlo simulation. Applied at a controlled test site, the method assimilated new sensing data iteratively to update the soil boundary model. Excavation initiated in high-uncertainty areas accelerated model convergence and reduced volumetric uncertainty by more than 80%, while systematic excavation sometimes increased uncertainty when uncharted zones remained.

The third study embeds adaptive uncertainty modelling within a GeoBIM environment and evaluates it across three excavation projects under typical Nordic conditions. The framework delivers a continuously updating most-probable boundary surface to machine control systems, improving prediction accuracy, reducing over-excavation, and lowering costs compared with TIN-based and operator-guided methods.

Overall, linking on-board sensing with probabilistic modelling transforms GeoBIM from a static design tool into a dynamic, measurement-based decision-support system. The adaptive approach is compatible with existing machine fleets and establishes a foundation for uncertainty-aware digital twins and more sustainable geotechnical construction.