Probabilistic analysis of skirted foundations under combined loading in soils modelled by non-stationary random fields

Skirted foundations are widely used in offshore engineering. The soils that skirted foundations are sited in commonly exhibit inherent spatial variability due to complex formation processes and stress history. In this study, probabilistic undrained capacities are investigated for skirted foundations with varying embedment ratios in spatially variable soils characterised by both stationary and non-stationary random fields using lower bound random finite element limit analysis combined with Monte Carlo simulation. Results show that the spatial variability of soil tends to decrease the uniaxial capacities. The coefficient of variation of soil strength, among the investigated parameters, could significantly affect stochastic uniaxial bearing capacities. Both embedment ratio and soil heterogeneity index affect the size and the shape of probabilistic failure envelopes, while statistical characteristics of soil strength including coefficient of variation and autocorrelation distance affect the size of probabilistic failure envelopes but have a minor influence on the shape. A larger embedment ratio is recom-mended to be adopted for skirted foundations sited in spatially variable soils from the perspective of load -carrying capacity. A series of expressions are proposed to describe the deterministic and probabilistic normal-ised failure envelopes for skirted foundations under combined vertical, horizontal and moment loadings.