In this study, a simple and cost-effective method was used to synthesize red iron oxide pigments from a mill scale, which is a waste material generated in the iron and steelmaking industries. Mill scale and sulfuric acid were reacted to produce hydrated ferrous sulfate (FeSO 4 ·XH 2 O), which was then oxidized with varying concentrations of sodium nitrate and then calcined at various temperatures and times. X-ray diffraction (XRD), scanning electron microscopy (SEM), dynamic light scattering (DLS), and ultraviolet-visible spectroscopy (UV-Vis) were used to characterize the resulting powders. The morphology of Fe 2 O 3 was influenced by the concentration of the oxidizing agent. As the amount of the oxidizing agent increased, the structure changed from rod shape to round. The findings also indicated that increasing the calcination time and temperature facilitates the growth of particles from 127 to 524 nm, resulting in the production of various shades of red while maintaining a consistent Fe 2 O 3 crystalline phase. Additionally, optical analysis showed that the samples possessed a bandgap in the range of 2.05-2.25 eV, which changed with the temperature, time, and oxidizing agent concentrations. To predict pigment color (L*, a*, b*) from process parameters, six machine learning models; Linear Regression (LR), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RF), Support Vector Regression (SVR), and Multilayer Perceptron (MLP), were developed and evaluated. Among the models tested, SVR demonstrated the best performance based on cross-validated prediction errors, with an average MAPE of 0.126 across the three-color parameters (L*, a*, b*), MAE = 2.79, RMSE=3.42 and R 2 = 0.62, providing improved predictive performance relative to linear kernels within the studied parameter range. SHAP (SHapley Additive exPlanations) analysis indicated temperature as the dominant factor controlling color, with time and oxidizing agent influencing outcomes through nonlinear, interaction-dependent effects. This integrated experimental–computational approach enables efficient synthesis of tunable red iron oxide pigments from industrial waste and reliable prediction of their color properties. • Synthesis different shades of red iron oxide pigments from waste material: mill scale • Investigation of the effects of calcination time, temperature, and oxidizing agent on the color features of synthesized red iron oxide pigments • Prediction of color parameters using artificial intelligence techniques
Ghazitabar et al. (Sun,) studied this question.