ABSTRACT
During bridge deck construction, overhang brackets are commonly utilized to resist the weight of fresh concrete and the loads from heavy construction equipment. These loads transfer through the overhang bracket to the bridge’s girder system, resulting in significant torsional moments, particularly on the exterior girder. Under certain circumstances, such torsional moments have been observed to induce excessive transverse rotation in bridge exterior girders. This rotation can potentially give rise to safety and maintenance concerns throughout the entire bridge’s service life. A detailed finite element analysis is typically recommended to mitigate these isto evaluate torsional effects on bridge girders during deck construction. However, this analysis, though thorough and comprehensive, can be tedious and time-consuming. In this study, several deep-learning models were developed to accurately predict the maximum rotation of the exterior girder. Over 10,000 finite element models were generated using SAP2000 to gather enough data to develop these deep learning models. The accuracy of these deep learning models was also scrutinized and compared. The results suggest that deep learning-based prediction models can effectively and efficiently assess the rotation of the exterior girder during bridge deck construction, thereby providing a powerful tool for engineers and construction professionals in the bridge design and construction field.
