Application of machine learning tools in road bridge weigh-in-motion systems

Abstract This study explores the application of machine learning (ML) techniques in road and bridge weigh-in-motion (B-WIM) systems, focusing on their potential to improve the accuracy and efficiency of load identification processes. The study begins with an introduction to the fundamentals of ML and artificial intelligence, providing an overview of key algorithm families and their applicability in bridge civil engineering. Subsequently, the concept of B-WIM systems is discussed, emphasizing the role of ML in vehicle classification, detection, and load estimation. A novel system architecture is proposed, integrating the You Only Look Once algorithm for real-time vehicle detection and tracking, and autoencoders for analysing the structural response signal to estimate axle loads. Proof-of-concept case studies, implemented within a numerical simulation environment (finite element software SOFiSTiK), are presented to illustrate the feasibility of such approaches. These examples demonstrate key concepts in a simplified form, supporting the proposed methodology without requiring full-scale experimental validation. Results demonstrate the advantages of ML tools in handling dynamic loads and diverse traffic conditions, significantly outperforming traditional methods regarding accuracy and scalability. The study concludes by highlighting the broader implications of using ML in structural monitoring and its potential to revolutionize load identification in bridge engineering.