Probabilistic prediction of spudcan peak penetration resistance based on parameter estimation and sectionalized adaptive linear simplification

Punch -through failure is widely acknowledged as the most common accident during jack -up preload. Evaluating the possibility of punch -through failure is essential for reducing risks in practical operations. However, the predictions of peak penetration resistance are often based on deterministic analyses, where uncertainties related to the soil and model are not taken into account. A new parameter estimation technique, derived from the Kalman filter and local least square fitting, has been developed. This technique combines real-time monitored data with sectionalized adaptive linear simplification of the profile. The prediction of peak penetration resistance is refined by attributing all prediction -related uncertainties to the parameter and re -predicting with the updated parameter. The precision, robustness, and reliability of the parameter estimation method are measured through accurate measurements obtained from geotechnical centrifuge tests and field applications. The results demonstrate that predictions can be effectively refined for sand -over -clay and three -layer clay deposits by incorporating monitored data, regardless of the initial model. The introduced technique facilitates real-time optimization in evaluating spudcan penetration, offering a significant improvement in predicting and mitigating punch -through failure risks during jack -up preload operations.