Experimental and Machine Learning Analysis of High‐Strength Concrete Incorporating Waste Slag and Recycled Steel Fiber

The synergistic impact of waste slag and recycled steel fiber on high‐strength concrete (HSC) has been needed to produce eco‐friendly concrete in recent times. Consequently, this study aims to investigate the combined effects of waste slag and recycled steel fiber on the fresh, mechanical, and durability properties of HSC. Waste steel fibers (60 mm × 0.9 mm) were incorporated at 0.25%, 0.50%, and 0.75% by volume, levels selected based on practical ranges reported in literature and allowing evaluation of performance across incremental fiber additions, while slag replaced 15% and 30% of the cement by weight in the concrete mixes. Experiments assessed the workability, Kelly ball penetration, density, and compacting factor of fresh concrete, while mechanical characteristics (compressive, splitting tensile, and flexural strengths) were evaluated at 7, 28, and 90 days, and durability performance was tested through rapid chloride penetration, water absorption, sorptivity, and electric resistivity. Therefore, artificial neural network (ANN) and random forest (RF) were used as machine learning (ML) methodologies to predict the strengths of concrete. This research also explored the compressive strength to compare the nondestructive test with the destructive test results at 28‐ and 90‐day periods. The experimental outcome revealed that incorporating slag improved fresh workability (higher slump and compacting factor), while steel fibers slightly reduced it due to the interlocking effect. Additionally, the introduction of 30% slag and 0.75% steel fiber into HSC led to substantial improvements in compressive, tensile, and flexural strengths compared to the reference mix after 90 days. At 90 days, slag and steel fiber mixes showed up to 23% lower water penetration and slag‐only mixes achieved up to 82% higher electrical resistivity than the control, confirming improved durability. The SEM analysis of slag‐based concrete mixes revealed a denser and more homogeneous microstructure with reduced porosity, which correlates with the observed improvements in compressive strength, tensile strength, and durability performance. Building on these experimental insights, predictive modeling was performed, where the RF model showed better results than the ANN model for all three strength properties, with higher R 2 values (0.994, 0.992, and 0.996) compared to ANN (0.971, 0.932, and 0.948) and lower errors in terms of mean square error (MSE), root MSE (RMSE), and mean absolute error (MAE) than ANN.