부산대학교 도서관

This study addresses the critical need for efficient and accurate seismic analysis of high-speed railway bridges, which are increasingly prevalent in extensive railway networks. Traditional seismic models of these bridges are accurate but computationally intensive due to their complexity. To overcome this, we introduce an innovative parameter updating method for bidirectional simplified finite element models, which enhances computational efficiency without sacrificing accuracy. This method leverages the combined strengths of artificial neural networks (ANN) and an enhanced particle swarm optimization (PSO) algorithm, which together streamline the updating process and ensure precision. The ANN effectively maps the relationship between model parameters and seismic responses, while the PSO algorithm, with its dynamic adjustments, optimizes parameter updates. Application of this method to a four-span high-speed railway bridge demonstrated a significant increase in computational efficiency—over twice that of traditional models—while maintaining high fidelity in replicating both static and dynamic seismic behaviors. Our findings reveal that the updated model not only accelerates seismic analysis but also provides a reliable means of assessing the performance state of the bridge's components. This method holds promising potential for advancing regional seismic design and intelligent route selection in high-speed railway projects.