Self-healing concrete can repair microcracks and enhance durability, but traditional methods are costly and complex. Using recycled coarse aggregate (RCA) as a substitute for natural aggregates can reduce costs, improve self-healing properties, and support sustainable construction. This study applied machine learning (ML) to predict the impact of RCA on concrete's self-healing performance. A database of 173 datasets was established, with eight key indicators as input variables and the self-healing rate as the output. An optimized NRBO-XGBoost model was developed and compared with four ML models and two optimization methods. The NRBO-XGBoost model outperformed all others, achieving the highest accuracy (R² = 0.9569, RMSE = 7.1800, MAE = 4.9575). Sensitivity analysis using the Shapley method identified crack width as the most influential factor in self-healing performance, while RCA had minimal impact within the studied range. However, RCA's low cost and environmental benefits highlight its potential for practical applications. This study provides theoretical insights into self-healing recycled concrete and introduces an optimized ML approach for performance prediction. The findings offer valuable guidance for sustainable construction practices.
As a novel and sustainable building material, self-healing concrete technology has garnered significant interest from both the scientific and engineering communities due to its ability to autonomously repair microcracks in concrete, thereby enhancing structural durability. However, the high cost and complexity of producing these cement-based composite materials have largely confined their application to experimental research stages. There are various types of self-healing concrete, including natural self-healing concrete, capsule-based self-healing concrete, fiber-reinforced self-healing concrete, and bacterial self-healing concrete. The self-healing performance varies among different combinations of these materials. To improve self-healing capabilities and reduce the high production costs, some researchers have combined self-healing concrete with recycled aggregates. Using RCA as a raw material for self-healing concrete not only reduces economic costs and extends the service life of building materials but also increases the survival rate of bacteria that promote concrete self-healing. Additionally, it lowers carbon emissions, thus contributing to the sustainable and healthy development of green construction. Liu et al. reviewed the applications of self-healing concrete derived from waste materials, including fly ash, blast furnace slag, silica fume, and recycled aggregates. Ali et al. studied the performance of concrete made from RCA and explored optimized preparation methods to enhance concrete properties. Wang et al. found that adjusting the aggregate gradation and sand ratio in concrete could improve its self-healing ability. Deepasree et al. attempted to use construction waste as recycled fine aggregate to prepare self-healing mortar, discovering that adding 25% recycled fine aggregate resulted in optimal strength. Sabri et al. used low-calcium fly ash and partially substituted RCA to produce concrete with strong self-healing capabilities, revealing the potential of self-healing recycled concrete. Muhammad et al. compared the effects of different immobilizing agents and their sizes on RCA concrete embedded with Bacillus, finding that bacteria immobilized with RCA exhibited the highest crack repair efficiency at 28 days after pre-cracking. When initial cracks were within a certain range, the combination of microbes and RCA was able to heal most of the cracks within a week, with healing behavior observed in both crack width and depth, and a certain degree of recovery in compressive strength.
Furthermore, artificial intelligence (AI) and ML technologies have been widely applied to predict concrete properties. These methods enable modeling based on existing experimental data, significantly reducing the time and cost associated with experimental testing. Wassim et al. reviewed the application of ML models in predicting the mechanical strength of concrete, providing a detailed discussion and analysis of the performance of various models, including artificial neural networks, support vector machines, decision trees, and evolutionary algorithms. Mohammad et al. further elaborated on the relationships among AI, ML, and deep learning (DL), as well as their applications in predicting the mechanical properties of concrete. Harshit et al. delved deeper into the techniques for enhancing the strength and improving the microstructure of recycled aggregate concrete by incorporating fibers and Bacillus. Ahmad et al. studied the impact of Bacillus subtilis on the strength characteristics of RCA and used ML for prediction. Kennedy et al. compared various intelligent optimization algorithms in predicting concrete's mechanical properties and identified the best-matching algorithms for different prediction indicators. In addition, Rupesh et al. used optimization algorithms in conjunction with machine learning as a way to improve the accuracy of engineering material property predictions and optimize structural design. Suraj et al. focus on the combination of optimization algorithms with machine learning models such as XGBoost to solve key problems in materials engineering and provide technical support for sustainable building materials approaches. Despite the extensive research on predicting the mechanical properties of concrete, there are relatively few studies on predicting the self-healing capacity of composite concrete materials. Even fewer studies focus on predicting the self-healing capacity of cement-based composite materials by integrating factors such as RCA, bacterial content, and concrete preparation parameters. Studies that combine the effects of multiple factors, including RCA and bacteria, on the self-healing properties of concrete and combine them with ML modeling to predict the performance of composites are extremely rare and important. This comprehensive study not only considered the economic and environmental advantages of RCA, but also explored the mechanism of its influence on the self-healing properties of bacterial concrete, providing new ideas and methods for developing high-performance and sustainable self-healing concrete.
Therefore, this study aimed to accurately predict the self-healing performance of concrete by examining the influence of various factors and exploring the potential application of RCA in the development of self-healing concrete. A comprehensive dataset of experimental results on concrete self-healing performance from recent years was compiled. Based on correlation analysis and research objectives, eight key indicators were selected as input variables, with the concrete self-healing rate serving as the output variable. The study primarily focused on evaluating the accuracy and reliability of the NRBO-XGBoost ensemble model in predicting concrete self-healing performance, comparing its performance against various single models and other ensemble models. The NRBO-XGBoost model used in this study not only performs well in hyper-parameter tuning, but more importantly, it combines the Newton-Raphson iterative method with the XGBoost algorithm, which is able to deal with the nonlinear problem more efficiently, and more accurately capture the relationship between the performance of the self-healing concrete and a variety of complex factors, so that it can have higher accuracy and reliability in predicting the efficiency of the self-healing concrete. Additionally, the Shapley explanation method was utilized to analyze and rank the significance of each factor, identifying the most sensitive variables influencing concrete self-healing performance. The analysis concluded that RCA has a minimal impact on the self-healing performance of concrete. However, its strong negative correlation with bacterial content suggests that RCA can provide space for bacteria to survive, thereby increasing bacterial survival and reducing the amount of bacteria required. Moreover, an evaluation of economic costs and other considerations affirmed the potential of RCA in self-healing concrete applications, providing a theoretical and practical foundation for future research on self-healing recycled concrete.