Coupling Soil-Structure Interaction (SSI) PDEs with live jacking load data to estimate resistance in real-time.
Read Time: 14 min read | Published: 2025-03-25
Punch-through incidents are catastrophic; in one documented sand-over-clay case, failure occurred at 53-56 MN despite prediction models estimating capacity at only 20-34 MN (Jack-Up Conference). Such methods often underpredict failure loads by 50%+. This risk is compounded by stiff-over-soft clay layering (Polish Maritime Research, 2020) and the lack of site-specific spudcan design for mobile units.
We utilize Soil-Structure Interaction (SSI) PINNs to solve the inverse problem, utilizing domain-decomposition multi-network architectures to handle discontinuities at soil interfaces (arXiv:2212.08306). Unlike basic Random Forest models, this physics-constrained approach integrates jacking load measurements to infer the soil resistance profile ahead of the spudcan tip in real-time.
While PINNs have demonstrated success in identifying parameters for pile-soil interaction (Acta Geotechnica, 2024), validating spudcan penetration involves complex large-deformation mechanics. Distribution shift remains a challenge when moving between sites, though PPI-based uncertainty quantification offers a path to robust per-site calibration.
This must remain a decision-support tool. Given the complexity of soil failure mechanisms (Jack-Up Conference, 2015), 'Human-in-the-loop' certification is essential to enhance, not replace, mandatory site-specific evaluations.
In the domain of structural health monitoring, particularly for heavy machinery components like ladders and spuds, static analysis is no longer suffic...
Pipelay Tensioner Slip PredictionPhysics-Informed Friction Models using existing pressure/speed data to predict slip risk with a 30-minute horizon....
Fleet-Wide Regulatory ComplianceExplainable PIML (XAI) that outputs uncertainty bounds rather than simple alerts to satisfy classification societies....