High-temperature bond strength evaluation of concrete overlays with industrial by-products: Experimental and analytical approaches using machine learning

This paper presents a comprehensive study on the bond strength of concrete overlays incorporating industrial by-products such as fly ash and ground granulated blast furnace slag under high-temperature conditions. The research involved 540 concrete specimens exposed to temperatures of 25 °C, 100 °C, 300 °C, 500 °C, and 700 °C and cured for 14, 28, 60, and 90 days. Results demonstrated the beneficial roles of fly ash incorporation in overlays in improving the thermal stability and bond strength between concrete substrates and repair material overlays at high temperatures. Utilizing artificial intelligence techniques, a set of advanced machine learning models was created to accurately predict bond strength properties, specifically compressive, tensile, and shear strengths. The models evaluated included gradient boosting, extreme gradient boosting, light gradient boosting, and categorical boosting. Additionally, a multilayer perceptron (MLP) neural network was used as a baseline for comparison. The categorical boosting model demonstrated superior performance, achieving a root mean square error of 2.570, mean absolute error of 1.792, and R² of 0.989 for predicting compressive strength in the test dataset. It also achieved similarly high accuracy for tensile and shear strength, with R² values of 0.927 and 0.942, respectively. The SHapley Additive exPlanations analysis highlighted the significant impact of the predictive models on temperature, curing time, density, and porosity. Monte Carlo simulations further ensured robust model predictions and insights. The findings, accessible through a user-friendly web application (https://high-temperature-opc-ibpc-cs-ts-ss-prediction-by-javid-toufigh.streamlit.app/), provide valuable insights into optimizing material compositions and enhancing the durability of repaired concrete structures in extreme environments. This study demonstrates the efficiency and reliability of artificial intelligence in bond strength evaluation of heat-damaged concretes.