Mechanical and durability assessment of marble dust–fiber concrete supported by ML prediction

This research investigates the need for sustainable, high-performance concrete by blending waste marble dust (MD) with recycled polypropylene fiber (PF) to reduce cement consumption and enhance concrete’s mechanical properties. An extensive experimental Programme was designed: 25 mix designs with varying MD levels (0–20%) and PF dosages (0–1.0%) were performed, and the experimental study was further aided by ML models for predicting and improving concrete properties. Fresh, durability and mechanical properties, with slump, density, compressive strength, flexural strength, split tensile strength, water absorption, permeability, and acid resistance, were measured. The compressive strength increased from 51.62 MPa for the control mix to 57.7 MPa at 10% MD replacement and 1.0% PF addition. However, mixes containing 10% MD with 0.6–0.8% PF exhibited comparable compressive strength while demonstrating better overall performance in terms of strength, durability, and workability. The split tensile strength increased from 3.236 MPa for the control mix (A0) to the maximum value of 4.249 MPa at 10% MD and 0.8% PF, which is an improvement of about 31%, and similarly, the flexural strength increased by 25% (to 5.54 MPa) compared to the conventional mix. Durability showed considerable improvements, with the amount of water absorbed decreasing from 3.42 to 2.84% and a 30% reduction in permeability (from 9.42 × 10–12 to 6.64 × 10–12 m/s). The ANN models yielded R2 values > 0.95 and thus demonstrated high predictive accuracy. One of the novel aspects of this study is the interaction among the integration of waste MD and secondhand PF, machine learning-based prediction and optimization, which enable the precise determination of the optimal mix proportions with respect to strength, durability, and sustainability, surpassing traditional empirical design methodologies.