For over a decade, INES Ingenieros Consultores has closely partnered with the Ministry of Transport and Sustainable Mobility (MITMS) in managing bridges and structures under various framework contracts, alongside its strategic partner Geocisa. Seeking to enhance the current asset management approach, which has traditionally focused on periodic condition monitoring, INES has proposed integrating advanced analytics and machine learning techniques to enable future condition predictions and transition towards a predictive maintenance model. This shift will support proactive intervention planning, optimizing both the safety and efficiency of each asset throughout its lifecycle.

MITMS oversees and maintains more than 12,000 km of roadways and around 25,000 bridges. As part of its asset management policy, the ministry maintains a comprehensive inventory database detailing the typology, dimensions, materials, and specific features of each bridge’s components. Moreover, in accordance with Standard 6.3-IC, MITMS has conducted regular inspections of these structures for over 15 years, with frequencies adapted to each type: annual basic inspections and more detailed inspections approximately every 5–10 years. The ministry also maintains records of incidents, emergency actions, historical interventions, and operational data such as traffic volumes. This extensive historical data set represents a highly valuable resource, and, when analyzed using advanced analytical tools, can be leveraged to predict structural reliability and implement predictive maintenance.

To achieve this, INES has developed a solution based on machine learning algorithms capable of predicting the future condition of bridges managed by MITMS. This system uses gradient boosting algorithms fed with all available descriptive data to predict each structure’s degradation rate and calculate its future condition index. The selection of a gradient boosting algorithm reflects both the regression challenge of the project and the type and quality of the existing data. Although this algorithm requires substantial data and poses complexity in interpretability, it delivers highly accurate performance.

The results achieved represent a 40% improvement in accuracy over the baseline model, with mean absolute errors around 6 points, providing a continuous insight into the condition of structures and overcoming the limitations of periodic assessments, which, at best, offer only a static “snapshot” every five years.

The project’s success opens up new possibilities for optimizing bridge management by anticipating maintenance needs, thereby increasing safety and resource efficiency for MITMS-managed infrastructure. Additionally, it creates new avenues for refining bridge evaluation methodologies and integrating predictive insights into decision-making processes, steering asset management towards a predictive maintenance model.