This thesis aims at performing a probabilistic analysis of an offshore monopile foundation embedded in a spatially varying clayey soil. The probabilistic analysis of geotechnical structures involving spatially varying soil properties is generally performed using simulation methods as the conventional Monte Carlo Simulation MCS methodology or the variance reduction techniques. These methods are not suitable for the accurate assessment of the small values of the failure probability encountered in practice when considering a computationally expensive mechanical model as is the case in this thesis. The traditional Kriging-based approach AK-MCS is an Active learning approach combining Kriging metamodeling and MCS simulation technique. Within AK-MCS, Monte Carlo simulation is performed without evaluating the whole population. Indeed, the population is predicted using a Kriging meta-model which is defined on the basis of only a few points of the population, thus significantly reducing the computation time with respect to the crude MCS. Notice however that AK-MCS approach presents some issues related to (i) the strategy of selection of new training samples when performing the enrichment process of the Kriging meta-model and (ii) the corresponding stopping criterion. Three Kriging-based approaches (namely, GSAS, AK-MCSm and AK-MCSd) were developed in order to overcome these issues. GSAS is a Global sensitivity analysis-enhanced surrogate modelling approach, AK-MCSm is a multi-point enrichment approach and AK-MCSd is a Kriging-based approach that takes into account the dependency between the Kriging predictions. The proposed approaches were found to provide efficient tools for the computation of the small failure probability of complex offshore foundations within a reasonable computation time.
Keywords: spatial variability, Kriging metamodeling, probability of failure, monopile foundation.
Stakeholders or Phd/Writer name
- Abdul Kader EL HAJ