Integration of modal strain energy-based damage index and artificial neural networks for damage severity prediction in single-span steel bridges

Abstract

Structural health monitoring (SHM) is essential for ensuring the durability and safety of steel bridges, which are considered as critical part of modern infrastructure. Ensuring the integrity of bridges is vital for maintaining their functionality and safety. Considerable research has been conducted on predicting damage location in multi-span steel bridges, while relatively limited attention has been given to the identification of damage in single-span steel bridges, and there exists a notable gap in the literature regarding the quantification of damage based on severity. This study presents a novel approach combining Modal Strain Energy-based damage indices with extensively trained data-driven Artificial Neural Networks (ANNs) to achieve highly accurate damage location and quantify this damage in terms of severity in single-span steel bridges. The research integrates a finite element (FE) model as the case study of the Alamosa Canyon Bridge in Sierra County, New Mexico. The validated FE model was used to compute damage indices based on the first two bending mode shapes. The damage index (DI) of each mode for different damage scenarios simulated on the FE model is then combined using the absolute value method to pinpoint damage locations. ANNs are employed to predict severity based on these damage indices, leveraging a comprehensive dataset generated through cubic spline interpolations and FE simulations. The DI was used as an input parameter for the ANNs to estimate damage. This methodology highlights the reduced computational effort using a streamlined method for structural health monitoring that maintains high accuracy in severity prediction and damage detection. This study enhances the infrastructure safety and maintenance strategies by offering a practical framework for steel bridges.