Predictive analysis and performance assessment of coal bottom ash in recycled aggregate concrete under elevated...

This study investigates the viability of utilizing coal bottom ash (BA) as a replacement for natural fine aggregate (NFA) in concrete containing recycled coarse aggregates (RCA). Although recycled materials often diminish mechanical properties, BA's potential as a fine aggregate in recycled aggregate concrete is under-researched. This study investigates various replacement levels (0%, 50%, and 100%) of NFA and natural coarse aggregates (NCA) with BA and RCA, respectively, to develop sustainable concrete. Physical and mechanical behavior assessments were conducted to evaluate their practical applicability, particularly under elevated temperatures, affecting key properties: compressive strength (f_c) and splitting tensile strength (f_sts), crucial for fire resistance. Cylindrical specimens were cast and exposed to temperatures spanning from 27 °C to 400 °C in a high-temperature furnace. These specimens were then tested using non-destructive and destructive testing techniques to assess their performance under thermal conditions. Results indicated that a compressive strength equivalent to that of the control mixture can be achieved with a 100% RCA replacement. However, non-destructive and destructive test results indicated that elevated temperatures cause significant damage to concrete containing BA (either with natural aggregates or RCA). Additionally, three predictive models, including Artificial Neural Network (ANN), Decision Tree (DT), and Extreme Gradient Boosting (XGBoost), were developed to predict f_c and f_sts of concrete after exposure to elevated temperatures and at various curing periods (28, 90, and 180 days). Model performance was evaluated using metrics such as mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), and correlation coefficient (R²). The XGBoost model outperformed ANN and DT, achieving the lowest MSE (after training:testing as 0.152:0.345 for f_c and 0.007:0.009 for f_sts), RMSE (as 0.39:0.588 for f_c and 0.085:0.094 for f_sts), MAE (as 0.301:0.451 for f_c and 0.068:0.079 for f_sts), and the highest R² (up to 0.997 for both f_c and f_sts). Furthermore, eXplainable Artificial Intelligence (XAI) techniques provided insights into the factors significantly impacting concrete strength predictions, with temperature emerging as the most influential feature. This study's findings offer a theoretical basis for enhancing fire-resistant concrete design and advancing sustainable construction materials.