Copula/random field/probability density evolution method analysis of pile–slab structures with soil spatial variability

Reliable seismic assessment of pile–slab structures in complex geological regions requires the joint consideration of soil-parameter dependence, spatial variability and stochastic ground motions (GMs). An application-oriented Copula/random field (RF)/probability density evolution method (PDEM) framework for displacement-based seismic reliability analysis was established in this work. Based on 42 paired measurements of cohesion and internal friction angle, uncertainties in sampling and model selection were assessed using paired bootstrap analysis, the Bayesian information criterion and alternative model sensitivity comparisons. The Weibull distribution, truncated normal distribution and Frank copula were adopted as the best-fitting baseline probabilistic model. Cross-correlated RFs with an exponential spatial correlation structure were then generated and mapped element-by-element into a multi-scale finite-element model. Using a physical stochastic model, 100 weighted stochastic GMs were produced and the PDEM was used to evaluate first-passage serviceability reliability under five pile-head displacement criteria and three uncertainty representations. Convergence analyses and comparison with 1000 Monte Carlo simulations supported the adopted computational settings. Deterministic analyses identified the connection pier as the critical location of displacement response, with a peak displacement of 68.4 mm at 0.6g, 18.6 mm larger than that of the continuous pier. Under the cross-correlated RF model, reliability decreased from 98.97% at a 1.2% pile-head displacement angle threshold to 27.15% at 0.4%. At the strictest criterion, the reliability was 18.62 percentage points lower than that of the model considering only seismic randomness, indicating that neglecting soil spatial variability may overestimate displacement-based serviceability reliability under the adopted assumptions.