Sensitivity Analysis of Concentric-Arches Model Parameters for Triangular Pile-Supported Embankments Using OAT and Monte Carlo Methods

Abstract

The Concentric-Arches (CA) model is widely adopted to predict load transfer in Geosynthetic-Reinforced Pile-Supported Embankments (GRPSE), yet its parameter sensitivity for triangular pile arrangements remains poorly quantified. This study performs a combined One-At-a-Time (OAT) and Monte Carlo Sobol global sensitivity analysis of six CA parameters: the fill internal friction angle, foundation cohesion (c), geogrid tensile stiffness (J), pile spacing (s), embankment height (H), and fill unit weight. Eleven-level OAT screening (66 runs) and Latin Hypercube Monte Carlo (n = 2000) were executed on a closed-form CA formulation extended to triangular geometry. Calibration was conducted against two field cases reported in a high-speed-railway embankment dataset: a time-dependent load-sharing series (500 d, four soil-model variants) and a height-dependent stress-ratio series (0.5-8 m, two field instruments). The results show that H has the greatest first-order influence (Si = 0.63, STi = 0.49), followed by s (STi = 0.21), J (STi = 0.11), the friction angle (STi = 0.1), and c (STi = 0.09). After tuning, the CA model reproduces the load-sharing time series with a Root Mean Square Error (RMSE) of 4% and a coefficient of determination (R-squared) of 0.84, and the height-dependent stress ratio with an RMSE of 5.2% and an R-squared of 0.84, representing reductions of 86% and 42% in RMSE relative to the literature-default parameter set. The calibrated parameter envelope provides design guidance for triangular pile layouts and identifies geogrid stiffness as the most controllable lever among the model inputs.