This paper presents an efficient threefold methodology for using random field finite-element limit analysis (RFELA), designed to be applied in probabilistic failure analysis of geotechnical structures under undrained conditions, explicitly accounting for the spatial variability and correlation of soil strength. The proposed approach extends the results of RFELA, conducted on a small yet representative subset of random spatial distributions of soil properties, to a much larger set. For each new spatial distribution, the method predicts strict upper bounds for the collapse load with good accuracy, without requiring additional finite-element analyses. The paper also explores the application of this approach to estimate non-strict lower bounds. This methodology facilitates the use of Monte Carlo simulations on large collections of random samples with minimal computational cost. To demonstrate its effectiveness, the method is validated through three numerical applications, assessing both its accuracy and computational efficiency.
Silva et al. (Thu,) studied this question.