Characterising the model uncertainty of ISO methods for punch-through capacity prediction

Abstract

The model uncertainty is explored for two ISO methods to estimate the peak penetration resistance in an offshore spudcan foundation punch-through incident. The model factor, defined as the ratio of measured resistance to calculated resistance, is introduced in the characterisation of the model uncertainty. Model uncertainty of a coupled Eulerian–Lagrangian (CEL) method is first characterised, wherein the model factor of the CEL method (MCEL) is found to be a random variable following a lognormal distribution, with a mean of 1.00 and coefficient of variation of 0.15. Based on results from the CEL method, regression analyses are performed, and conventional methods are modified by multiplying the results of the ISO method by a correction factor f obtained from statistical analysis of the CEL results. The model factors of modified ISO methods (M′) for centrifuge tests are calculated for further verification. Results show that, with the modifications, the systematic part of uncertainty is removed from both ISO methods, and M′ can be directly regarded as a lognormally distributed random variable. Finally, to illustrate the workability of the proposed modified ISO method, a case history is conducted. Results show that the ISO estimated punch-through capacity is improved by that modification.