SHM of RC Bridge Girder for Damage Identification and Localization Using Machine Learning

Abstract

Structural Health Monitoring (SHM) is essential for ensuring the safety of reinforced concrete (RC) bridge girders, which are key components of transportation infrastructure. Traditional methods like visual inspection and nondestructive evaluation can be expensive and may not cover all areas, making it hard to detect and localize damage comprehensively. This study uses Finite Element (FE) modelling in Abaqus software to create detailed computer models of scaled down RC bridge girder (as a RC beam) and simulate different localized shear crack scenarios through smear crack modelling to extract time history output with respect to acceleration data. These generated synthetic datasets are then used to train a multi-layer perceptron (MLP) based machine learning architecture for damage detection and localization. The MLP model is designed to automatically learn complex, non-linear relationships from a wide range of statistical, frequency domain, and wavelet features extracted from the acceleration data. This approach offers a cost-effective way to monitor bridges by reducing the need for extensive sensors and manual inspection, potentially improving safety and reduce maintenance costs. However, further validation with lab tested data and then scaling it to real-world data is needed to confirm its effectiveness.