Analysis of the characteristics and environmental benefits of rice husk ash as a supplementary cementitious material through experimental and machine learning approaches

Rising cement consumption over the past few decades has become a huge environmental concern since Ordinary Portland Cement emits huge amounts of atmospheric CO2. Thus, governments and environmental organizations have required strict emission reduction measures by switching to greener cement additives. The significance of this experiment was to assess the implication of varying concentrations of rice husk ash (RHA) on mechanical characteristics with machine learning-based prediction, microstructural analysis and environmental impact assessment of concrete. This environmental impact investigation takes into account every stage of the life cycle of RHA concrete, from the extraction of raw materials and RHA manufacture to the final disposal and inclusion of RHA into concrete. A range of RHA weight percentages (0, 5, 10, 15, 20, 25 and 30%) was used to partly replace the cement. The most beneficial compressive and tensile strength results were shown in concrete and mortars that used 15% RHA as an alternative to cement. Additionally, incorporating RHA resulted in the most advantageous in terms of microstructural assessment and led to higher environmental resilience. Finally, the environmental impact assessment includes metrics such as the reduction of embodied CO2 equivalent (CO2-eq, measured in kg CO2) per MPa for RHA concrete, showcasing a considerable positive impact on the lowered global warming potential (GWP). Meanwhile, the mechanical characteristics are explored through the use of two distinct approaches to machine learning, extreme gradient boosting (XGBoost) and k-nearest neighbors (KNN), demonstrating the feasibility of precisely foreseeing the characteristics of mortar without the need for rigorous experimental work. XGBoost is chosen for its adaptability and excellent accuracy predictions for analyzing numerous data types and operations. In contrast, KNN is picked for its ease of detecting specific correlations and trends within the dataset. With a larger R2 value, XGBoost ultimately performs better than KNN.