Bearing performance and failure-mode prediction of geocell-reinforced foundations

The plate-load tests of geocell-reinforced foundations were simulated using the Particle Flow Code in Two Dimensions (PFC2D) to investigate bearing performance under varying geocell geometries, layout configurations, and eccentric loading conditions. The results show that geocell reinforcement markedly alters foundation failure modes and increases bearing capacity by factors of 1.39–5.65. The optimum burial-depth ratio and reinforcement-layer number were u / B = 0.1 and N = 2, respectively, yielding the maximum bearing-capacity improvement. A soft-voting ensemble integrating five optimized base classifiers was subsequently developed through weighted averaging to predict failure modes. The ensemble achieved an accuracy and F1-score of 0.81, a Kappa coefficient of 0.71, and a Matthews correlation coefficient (MCC) of 0.72, outperforming all individual base classifiers. Shapley additive explanations (SHAP) analysis identified geocell height, soil relative density, and load eccentricity as the dominant factors influencing failure-mode evolution. The proposed model enables direct prediction of failure modes in geocell-reinforced foundations based on soil, geocell, and loading parameters, demonstrating practical engineering relevance.