Modeling for assessment of long-term behavior of prestressed concrete box-girder bridges

Large-span prestressed concrete (PC) box-girder bridges suffer excessive vertical deflections and cracking. Recent serviceability failures in China show that the current Chinese standard modeling approach fails to accurately predict long-term deformations of large box-girder bridges. This hinders the efforts of inspectors to conduct satisfactory structural assessments and make decisions on potential repair and strengthening. This study presents a model-updating approach that aims to assess the models used in the current Chinese standard and improve the accuracy of numerical modeling of the long-term behavior of box-girder bridges, calibrated against data obtained from a bridge in service. A three-dimensional finite-element model representing the long-term behavior of box-girder sections is initially established. Parametric studies are then conducted to determine the relevant influencing parameters and to quantify the relationships between those and the behavior of box-girder bridges. Genetic algorithm optimization, based on the response-surface method (RSM), is used to determine realistic creep and shrinkage levels and prestress losses. The modeling results correspond well with the measured historic deflections and the observed cracks. This approach can lead to more accurate bridge assessments, which result in safer strengthening and more economic maintenance plans.