Modeling Deformation Induced by Thermal Loading Using Long-Term Bridge Monitoring Data

An accurate correlation model between thermal loading and deformation is required for facilitating a reliable deformation-based condition evaluation in bridge service periods. In this paper, a general approach for modeling closed-form thermal correlation of deformation based on monitoring data is proposed and applied in a long-span arch bridge. First, samples of all available thermal variables and deformation induced by thermal loading are prepared by averaging preprocessed monitoring records at a 10-min interval. Then these available thermal variables are reduced to several predominant thermal variables, each of which represents a cluster of thermal variables with statistical similarity and has the strongest relationship with the thermal deformation in concern. Finally, the model of thermal deformation is formulated as a weighted sum of fitted functions of predominant thermal variables. The weighted coefficients are calculated by the back-propagation neural network technique combined with the mean impact value method and the fitted functions are estimated by the nonlinear least-squares method. The proposed approach is applied to 1 year of monitoring data obtained from a sophisticated structural health monitoring system deployed on the Jiubao Bridge. The proposed method is validated and a closed-form thermal correlation model of vertical deformation in the Jiubao Bridge is established.