FATIGUE ASSESSMENT OF STEEL-CONCRETE COMPOSITE BRIDGES USING NONLINEAR FINITE ELEMENT ANALYSIS AND NEURAL NETWORK
TL;DR
A data-driven framework for fatigue assessment of steel-concrete composite bridges that integrates nonlinear finite element analysis (FEA) with artificial neural networks (ANN) is presented, enabling efficient exploration of load paths and cycle-dependent deterioration.
Abstract
Fatigue loading is a primary degradation mechanism in transport infrastructure, where repeated traffic actions progressively reduce structural capacity. This paper presents a data-driven framework for fatigue assessment of steel-concrete composite bridges that integrates nonlinear finite element analysis (FEA) with artificial neural networks (ANN). Nonlinear FEA is used to generate quantities of interest—peak stresses and deflections—across varied load positions and accumulated cycles. The resulting dataset is then used to train and validate an ANN surrogate that rapidly predicts these parameters for unseen loading scenarios. The framework is demonstrated on the BEAST® experiment, a large-scale study featuring high-cycle moving loads and pronounced steel-concrete interaction effects. Across the examined cases, the surrogate closely reproduces the nonlinear FEA results while substantially reducing computational cost compared to full nonlinear analysis, thereby enabling efficient exploration of load paths and cycle-dependent deterioration. The methodology facilitates accelerated parametric studies, supports scenario analysis, and informs maintenance planning for composite bridge systems.