Predictive Modeling and Durability Analysis of Low Carbon Concrete Incorporating Recycled Materials

Abstract

Increasing environmental demand for low carbon concrete materials has developed significant interest in the use of industrial by-products and recycled materials towards minimizing the environmental impact of concrete production processes. This dissertation investigates the use of recycled concrete aggregate (RCA), coal ash (CA), and quarry by-products (QB) as the substitute materials to produce cementitious composites using a hybrid approach of machine learning meta-analyses combined with experimental verification. In the first half of the dissertation, a meta-analysis of more than 750 experimental data available in the literature was performed to predict the compressive strength of concrete using RCA. Various machine learning models, such as individual learners and ensemble models, were developed and compared. Among them, the Light Gradient Boosting Machine (LightGBM) provided optimal predictive performance (R² = 0.94), and the most critical variables were related to age, water-to-cement ratio, and fine RCA content. The study also revealed that partially saturated RCA provided optimal strength results, followed by reduction in strength with fully saturated or oven dry RCA. These results provide data-driven insights into the optimization of the RCA concrete mixtures. The second half of the dissertation focuses on the simultaneous utilization of CA and QB in pastes, mortars, and concretes. Specifically, unconventional CA were considered, including a high sulfur fly ash and a fluidized bed combustion ash, relative to a commercially available coal fly ash. Granite and limestone QB were considered as replacements of commercially available limestone powder. A machine learning approach was used to determine optimal proportions of CA and QB in mortars. This method replaced traditional trial-and-error by leveraging existing data in the literature on using coal fly ash and limestone powder. Five different binder systems with different types and combinations of CA and QB were examined using isothermal calorimetry, flow tests, pore solution composition analysis, and compressive strength tests at different curing ages. Of the above, a ternary blend with portland cement, high sulfur fly ash, and granite QB was found to perform best with improved hydration kinetics, similar pore solution composition, and strength gain reproducibility relative to a control with a conventional coal fly ash and limestone powder. The study emphasizes the role of proportioning and chemical compatibility towards achieving sustainable mortars. These findings indicate that the proposed ML-assisted mix design approach can effectively identify high-performing mortar mixtures using industrial by-products. Though the model targeted for a compressive strength of 40 MPa, only the 100% cement OPC mix attained this value. All the other ML-optimized mixtures had a relatively lower strength, reflecting some compromise between performance and sustainability. In the final study, the synergistic interaction of CA and QB as partial replacements of cement in concrete was explored, specifically focusing on mechanical properties and durability performance. Constituent samples were evaluated for workability, mechanical properties, fracture properties, and durability properties. Testing proved that the combined usage of CA and QB enhanced the mechanical properties compared to conventional coal fly ash and control limestone powder as well as the long-term strength, in addition to eliminating the drawback of individual use. Compared to the control mix containing limestone, conventional coal fly ash in the CA-QB blends showed improved fracture toughness. Such synergy propels the development of structurally resilient and sustainable concrete mixes. The results confirm that combinations of CA-QB can provide a long-lasting and mechanically robust alternative to conventional cement, especially regarding long-term durability. This combination presents a scalable and circular economy solution for next-generation concrete infrastructure. The findings of this dissertation offer a multi-faceted solution combining predictive modeling and experimental verification to optimal utilization of recycled and secondary materials in concrete construction. This research facilitates sustainable construction by offering practical solutions to material optimization and performance improvement, opening the door to increased use of low carbon concrete in contemporary infrastructure. Furthermore, it demonstrates the practical potential of mix design with ML assistance and industrial waste recycling to produce sustainable concrete, but its practice application are limited by regional material variability and the lack of long-term field validation.