An Integrated Machine Learning Framework for Structural Health Monitoring of Bridges: A Case Study on Soan Bridge

Abstract

The failing of bridges in develop and developing world requires AI based monitoring systems to ensure safety, prolonged existence, and economic sustainability. This research proposes an integrated Structural Health Monitoring (SHM) framework that uses machine learning (ML) to check in real-time and predict when the maintenance of bridge is required. The proposed system uses data from sensors (accelerometers, strain gauges, LVDTs) along with visual images captured using cameras. A complete model which consists of data acquisition, preprocessing, Finite Element Model (FEM) validation, and use of multiple ML models such as Support Vector Machines (SVM), Convolutional Neural Networks (CNN), and Linear Regression. The proposed framework is implemented and validated through a step-by-step procedure and using the Soan Bridge, Pakistan as implementation bridge. The results obtained through evaluations show that SVM model achieved 96% accuracy in damage classification, however, the CNN was observed to be successful in the surface cracks. Further, linear regression model forecasted an alarming 79% loss in moment capacity by the year 2035, if no precautions measures are taken. The study concludes that the proposed ML-driven SHM system offers a strong, practical, and scalable solution for bridges or other infrastructure management, ultimately enhancing safety and optimizing maintenance resources.