“STRUCTURAL SYSTEM-BASED EVALUATION OF STEEL GIRDER HIGHWAY BRIDGES AND ARTIFICIAL NEURAL NETWORK (ANN) IMPLEMENTATION FOR BRIDGE ASSET MANAGEMENT”

Bridge asset management can be aided by implementing more rigorous, less conservative 3D analysis methods compared to conventional line-girder analyses. However, doing so requires an investment of time and cultivation of technical expertise. Machine learning can provide an intermediate measure between routine and rigorous methods to supplement decision-making in bridge management. This study examined the potential benefits of leveraging artificial neural networks (ANNs) in decision-making. ANNs can help decide when to employ detailed structural system-based evaluation and capacity assessment through more rigorous analysis methods or load testing in bridge asset management. Parametric analyses, performed using detailed 3D FEM-based modeling of representative bridges, highlighted the complexities of structural system-behavior that are ignored or obscured in the 1D-line girder approach, but have the potential to inform and improve bridge asset management. Both single-best-network and committee networks (CN) ANN approaches demonstrated a practical capability to predict refined capacities for existing steel girder bridges, mapped from geometric and material structural properties. The demonstrated efficacy of an FE-based load rating methodology integrated with a properly trained network model of optimized complexity for an existing steel-girder bridge population in Nebraska suggests its potential to supplement decision-making diversely in civil infrastructure management.