A machine learning approach for identifying vertical temperature gradient in steel-concrete composite beam under solar radiation

Abstract

The traditional research methods for the temperature field of bridge under solar radiation suffer from issues such as high workload and high costs. The temperature field of steel-concrete composite beam (SCCB) is studied in this paper using the ANSYS finite element software and MATLAB software. Firstly, a finite element temperature field model of SCCB is established based on measured meteorological data. Furthermore, the accuracy of the finite element temperature field model of SCCB is validated by collecting a small amount of temperature measurement data. The temperature sample database of SCCB was expanded based on this. Finally, a large amount of historical meteorological data was collected. The ANSYS software and Genetic Algorithm Back Propagation (GA-BP) hybrid model were used for calculation, and the representative temperature differences T_d1 and T_d2 of SCCB were obtained separately. The measured values are in good agreement with the finite element analysis results, showing consistent trends over time with a maximum difference not exceeding 1.6 °C. The GA-BP hybrid model proposed in this study, characterized by ‘structural features, temporal features, environmental features—node temperatures’, exhibits a high degree of nonlinear mapping capability. It has been demonstrated that the GA-BP hybrid model also possesses a high level of accuracy through verification. The SCCBs’ maximum vertical positive temperature differences (T_v), computed using ANSYS software and the GA-BP hybrid model, follow Generalized Extreme Value (GEV) distributions with parameters (-0.2722, 12.8715, 1.4105) and (-0.2855, 12.813, 1.3714), respectively. The representative values (T_d) of the maximum vertical positive temperature differences of SCCB, calculated by ANSYS software and the GA-BP hybrid model, are 17.613 °C (T_d1) and 17.2 °C (T_d2), respectively. The proposed temperature field calculation model for SCCB is based on meteorological parameters and the GA-BP hybrid model. It can accurately calculate the temperature field of SCCB in Guangdong region and improve computational efficiency.