Sustainably produced concrete using weathered Linz-Donawitz slag as a fine aggregate Substitute: A comprehensive study with Artificial intelligence approach

Abstract

Continuous consumption of natural resources such as river sand and coarse aggregates due to increased industrialization and urbanization has impeded the movement towards green concrete in construction, i.e., utilizing waste materials in its production, steel slag waste being one of them. Accumulation of Linz-Donawitz slag (LD slag) has become cumbersome in India, posing a problem of landfills and environmental degradation, i.e., abating soil and water quality due to the leaching of toxic metals. Fewer experimental studies have examined whether weathered fine LD Slag can substitute for fine aggregates in concrete. Additionally, the research has yet to compare the behavior of concrete comprising weathered fine LD slag in Normal Strength Concrete (NSC) and High Strength Concrete (HSC) containing Metakaolin. The current study aims to establish an optimum replacement quantity for fine aggregates with fine LD slag. The slag replaced fine aggregates by volume at 25, 50, 75 and 100%. The findings that examined the effects of using LD slag on the mechanical behaviour, durability and microstructure of concrete are presented in this study. Furthermore, taking a total of 180 experimental data points, the compressive strength of concrete was predicted using artificial intelligence (AI) methods such as Artificial Neural Networks (ANNs), Decision Trees (DT) and Random Forests (RF). Coefficient of determination R2, Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) were computed to assess the performance of generated models. Besides, a sensitivity analysis technique was employed to determine the most influential parameter among cement, Metakaolin, Fine Aggregate, Coarse Aggregate, LD slag, Superplasticizer, water content and testing age on the compressive strength results. Experimental investigation showed that adding LD slag up to 50% for NSC and 25% for HSC was optimal and showed better results for tested properties than conventional concrete mixes. Among proposed AI techniques, the DT technique was the best-performing model with an R2 value of 0.973 and the least for MAE (3.534 MPa) and RMSE (4.409 MPa). Sensitivity analysis performed on the DT model showed that testing age was the most critical parameter for the prediction of compressive strength.