A study in Scientific Reports shows how waste marble dust can replace up to 32% of the cement in grouts used for semi-flexible pavement, with machine learning guiding the mix design. Researchers built a dataset of 72 grout mixtures, tested compressive strength at 7 and 28 days, and trained five algorithms to predict performance. A decision tree did the best job, with an R-squared of 0.994 and an average error of 0.266 MPa. The model also showed that both extra water and higher marble dust content weaken the grout, so the substitution only works in a specific window. At a water-to-cement ratio of 0.40 with a small superplasticizer dose, the optimized mix kept enough strength while cutting cement content. That translates into lower embodied carbon for bus lanes, intersections, and industrial yards that use semi-flexible pavement. The wider takeaway is the workflow. A small lab campaign, a transparent ML model, and explainability tools can map a mix design space that would be expensive to explore by trial and error. For anyone tracking low-carbon construction or circular economy materials, this is a practical example of waste streams replacing high-emission inputs.
