Machine learning powered design of eco-friendly prestressed concrete sleepers

• We proposed an innovative machine learning powered engineering design of railway prestressed concrete sleepers. • The ML-powered design tool is capable of analysing the embodied carbon footprint of the sleepers. • The ML-models were based on over 3,000 full-scale datasets to assess the structural performance. • The tool can automate the optimized concrete structural design of railway concrete sleepers towards low-carbon footprint. Prestressed concrete sleepers are integral to structural safety of railway infrastructures. Industry challenges have been encountered in reducing the carbon footprint of this vital railway component. This research is therefore the first to establish machine learning (ML) techniques to design and optimise embodied carbon (EC) of prestressed concrete railway sleepers. To achieve this, over 3,000 datasets from industrial design sources was collected, through a combination of experimental predictions with EN 13230 compliance, and design data. Advanced ML models (Bayesian ridge, Random Forest and Deep learning) have been established to predict and optimize both capacity and embodied carbon impact of eco-friendly prestressed concrete sleepers. The designed machine learning models exhibit excellent outcome for both capacity prediction and carbon prediction. Our results reveal that Bayesian Ridge (R 2 =1.0000) displays the optimum performance for carbon prediction. Bayesian ridge and random forest models appear better for sleepers’ capacity and carbon predictions. The insight offers new reliable tools for the capacity design of railway sleepers while reducing environmental impact, practically driving decarbonization in the railway industry and potentially leading to time and cost savings.