For accurately assessing the structural load bearing capacity of corroded reinforced concrete(CRC) structures after exposure to high temperatures, such as in sudden fire accidents, there is an urgent need for a unified model to predict the bond strength of CRC under high-temperature conditions. However, the degradation mechanism of bond strength in CRC under high-temperature conditions is highly complicated. Traditional empirical models cannot account for the multivariate correlations among factors such as high temperature, corrosion, and material properties. Based on a large amount of existing experimental data, ML methods can effectively establish a regression relationship between input and output features through data analysis. This paper utilized five ML algorithms, namely ANN, SVM, DT, RF, and XGB, to establish an interpretable prediction method for the high-temperature bond strength of CRC. The model was trained and tested using 612 sets of experimental data on high-temperature CRC. The results show that the predictions of the ML model are in good agreement with the experimental results. Moreover, the calculation outcomes of the ML model were compared with those of three theoretical calculation formulas, and the ML model demonstrated clear advantages. Furthermore, to address the black-box issue inherent in ML algorithms, the SHAP method was employed to enhance the interpretability of the bond strength prediction process for high-temperature CRC. This model will provide a theoretical basis for accurately assessing the extent of damage to CRC buildings after exposure to high temperatures.
Read full abstract