Probabilistic analysis of excavation-induced retaining wall responses considering parameter cross-correlation based on Archimedean Copula modelling

To investigate the influence of soil parameter dependence on excavation-induced retaining wall responses, a random finite element framework incorporating Archimedean Copula-based cross-correlation is developed. The spatial variability of soil parameters is modelled using random field theory, while the inter-parameter dependence is characterised using a bivariate Archimedean Copula. The framework is implemented in PLAXIS through the PLAXIS–Python API and Monte Carlo simulation, with the soil behaviour described by the Hardening Soil model with small-strain stiffness (HSS). The method is applied to an excavation case to examine the effects of dependence models and correlation levels between the secant stiffness E50ref and internal friction angle φ on retaining wall horizontal displacement and bending moment. The results show that parameter dependence affects the statistical distribution of wall responses, especially the response variability, high-percentile responses, and the probabilities that the responses exceed prescribed control limits. Compared with the Pearson linear correlation model, the Gumbel Copula model produces different estimates of response dispersion and the probability of exceeding prescribed control limits because of its nonlinear and asymmetric dependence characteristics. These findings indicate that Copula-based dependence modelling can provide additional insight into the probabilistic assessment of retaining wall responses when nonlinear or asymmetric dependence may exist between soil parameters.