Advanced data analytics for predictive maintenance of bridges on the Spanish national highway network

Predictive maintenance is a strategy aimed at anticipating asset deterioration by forecasting future condition, enabling proactive maintenance scheduling. Advanced analytics and machine learning allow for the identification of patterns in historical data, leading to more accurate forecasts and better-informed maintenance planning. In civil engineering, this approach is particularly relevant for managing complex assets such as bridge structures. This article presents a study on 24 517 bridges in Spain’s national highway network, managed by the Directorate-General for Highways under the Ministry of Transport and Sustainable Mobility. The objective was to model the evolution of the condition index (CI) over time. Historical inspection data were combined with traffic volumes, climate variables and geotechnical characteristics. Due to the absence of key variables such as the year of construction and maintenance history, the deterioration rate between inspections was modelled first and subsequently used to forecast future CI values. An iterative process was used to develop a predictive model with gradient boosting algorithms. The model improved the error metric by 40% compared to a naïve baseline and provided a continuous estimate of structural condition. This data-driven approach supports more efficient bridge management by optimising maintenance planning and identifying anomalies in inspection data.