Machine learning prediction of steel–concrete composite beam temperatures during hot asphalt paving

Abstract

In the bridge construction process, the temperature distribution within the steel–concrete composite beam (SCCB) under hot asphalt paving is not negligible in its impact on the structural performance. However, traditional static analysis methods for bridge temperature fields, such as temperature measurements and numerical simulations, are plagued by high workload and costly equipment requirements. Therefore, in this study, we explore a machine learning (ML) approach based on field measurements to predict the temperature field of SCCB during hot asphalt paving. The result showed that of the various ML algorithms tested, the K-Nearest Neighbors (KNN) algorithm provided the highest predictive accuracy for the temperature field of SCCB. Through feature selection and experimental analysis, we identify beam temperature (T_bt), hot asphalt temperature (T_s), ambient temperature (T_a), and box interior temperature (T_box) as key features for predicting the temperature of SCCB during hot asphalt paving. This study demonstrates that ML is an powerful tool for predicting the thermal behavior of bridge structures, with potential widespread application in identifying temperature evolution in bridge structures.