Automation in dynamic analysis and generative design of prestressed concrete railway bridge infrastructures

This study presents an innovative method for the dynamic analysis and generative design of high-speed ballasted railway bridges subjected to High-Speed Locomotive Multiple Articulated (HSLM-A) train loads. Compliant with Eurocode standards, a comprehensive database of over 4 million data points was generated, including maximum vertical displacement and acceleration data for more than 10,000 bridges affected by ten HSLM-A models at speeds ranging from 150 to 350 km/h. The key contribution of this research lies in a novel surrogate model that incorporates semantic search and advanced decoding techniques, significantly enhancing the calculation time and accuracy of dynamic behaviour predictions for single-span high-speed railway bridges. The performance of the developed model was verified through case studies on existing 30 m and 50 m span bridges, evidenced by an R 2 value of 0.999, highlighting the model's precision and rapid prediction capabilities. Additionally, the research introduces a cutting-edge framework for optimising the cross-sectional geometry of prestressed concrete railway bridges. A case study was then conducted for a typical box girder bridge to identify 25 feasible solutions better than the original design in terms of mass per unit length. This research showcases the synergy between advanced technology and structural optimisation, and it opens new avenues for future studies in this field. • A surrogate model with semantic search achieves R²= 0.999 for bridge dynamics. • 4 M+ data points enable 0.1 s analysis vs 22 min traditional finite element computation. • Multi-objective optimisation yields 25 designs with 30 % mass reduction for bridges. • Fast KD-Railway software provides real-time dynamic analysis. • A generative design framework compliant with Eurocode for PSC railway bridges.