TEMPERATURE-INDUCED STRAIN PREDICTION FOR THE LONG-SPAN STEEL TRUSS ARCH RAILWAY BRIDGE USING THE GRU

The mapping for temperature-induced strains is a reliable approach to characterize bridge structures. However, the temperature field of bridge structures is intrinsically time-varying due to the time-dependent nature of the solar radiation and surrounding environments. In addition, long-span steel truss arch railway bridges consist of thousands of structural members. Capturing the accurate relationships between bridge temperature and temperature-induced strains is challenging. To explore an accurate mapping for temperature-induced strains in bridges, this study systematically investigates the temperature distributions and temperature effects on a long-span steel truss arch railway bridge based on field monitoring data. Accordingly, the relationships between temperature-induced strains and spatial temperature distributions are investigated using the principal component analysis (PCA) and gated recurrent unit (GRU), which are applied to calculate structural strains under ambient excitation using field temperature measurements. Particularly, the GRUs with different combinations of inputs, including temperature increments, principal components of temperature increments, and the recorded time of the temperature, are trained. The predictions from different GRUs are compared with the field monitoring data. Results show that temperature-induced strains are profoundly affected by the temperature field of the bridge. The temperature-induced strains can be predicted accurately using GRU in cooperation with field temperature measurements from multi-measurement points. Note that the recorded time of the bridge temperature can be employed to represent the characteristics of the temperature field during the prediction. The results of this study can improve the performance assessment methods for bridge structures, which can be utilized for the abnormal detection of bridge structures.