Prediction of Damage Severity in Steel Bridges Using Natural Frequency Data and Artificial Neural Networks

Abstract

Steel bridges are essential to transportation systems and must be regularly inspected to guarantee their robustness and safety. Conventional methods of identifying damage are subjective and time-consuming, frequently depending on manual inspections. This study aims to overcome these limitations by using Artificial Neural Networks (ANNs), which are sensitive to changes in mass and stiffness caused by damage, to predict the severity of damage in steel bridges based on natural frequency data. This research proposes a new method for detecting damage in steel girder bridges by integrating natural frequency data with ANNs. This vibration-based fault detection method addresses the shortcomings of traditional approaches by employing natural frequency as a dependable indicator of structural irregularities. An ANN model was trained and validated using an extensive dataset that included natural frequency data from experimental modal analyses conducted under various damage conditions. To assess the model's accuracy in predicting the severity of damage, its performance was studied on a different dataset. The findings indicate that ANNs can effectively analyze frequency data to accurately predict the severity of damage, reducing the reliance on manual inspections. In addition to improving the effectiveness of structural health monitoring systems, this strategy supports the robustness and safety of steel bridge infrastructure. Overall, the findings demonstrate that ANNs trained using modal curvature data provide a reliable and effective solution for early damage detection in steel girder bridges, substantially improving safety and operational performance. This novel approach advances the field of structural health monitoring and offers a valuable means of preserving the integrity of critical infrastructure, including steel girder bridge systems.