EFFECT OF RBF BASIS FUNCTIONS ON STRUCTURAL RELIABILITY ESTIMATES

Tóm tắt

Traditional approaches such as the Monte Carlo simulation and the finite element method are widely used for structural reliability analysis but usually demand excessive computational resources. To address this limitation, response surface methods employing radial basis functions (RBFs) have been introduced as efficient surrogates for approximating implicit limit state functions. Nevertheless, the accuracy of RBF-based reliability analysis strongly depends on the choice of parameters and basis functions, while systematic guidelines for their selection remain insufficient. This study investigates how different RBF types—Gaussian (GA), Multi-Quadric (MQ), Inverse Multi-Quadric (IMQ), Thin Plate Spline (TPS), Cubic, and Linear—and their parameter settings influence reliability index assessment (RIA). The surrogate models are constructed using Latin Hypercube Sampling (LHS) to approximate limit state functions, and the Hasofer–Lind–Rackwitz–Fiessler (HL-RF) algorithm is employed to compute both the reliability index and failure probability. The findings provide insights into the sensitivity of structural reliability estimates to RBF configuration, offering useful guidance for selecting appropriate surrogate modeling strategies in engineering applications.