Mathematical and artificial neural network modeling for describing the infrared drying process of Moringa oleifera leaves and evaluation of product quality

Moringa oleifera leaves were used in the infrared drying method for powder production. The moisture ratio datasets during drying at different temperatures were fitted with eight thin-layer drying kinetics and analyzed by an artificial neural network (ANN). The goodness of fit was evaluated using the value of the coefficient of determination (R²), the chi-square (χ²), and the root mean square error (RMSE). Results indicated that drying time was between 40 and 95 min at a temperature of 55 to 70 °C. Among the mathematical drying models used, the Wang and Singh model best described the drying kinetics of Moringa leaves. But comparing with the ANN model—a machine learning-based model—it showed higher prediction capacity than the mathematical model did. For Moringa leaves dried at temperatures between 55 and 70 °C, the R², χ², and RMSE values for this model ranged from 97.85 to 99.59%, 0.0007 to 0.0029, and 0.0228 to 0.0503, respectively. Effective moisture diffusivity (D_eff) values varied between 1.908 × 10⁻¹¹ and 3.875 × 10⁻¹¹ m²/s, with an activation energy of 43.92 kJ/mol. The drying temperature also influenced the bioactive compounds in Moringa leaves. The vibrant color of the powder was produced by drying Moringa leaves at 65 °C for 50 min. The powder had 5.85% moisture, 31.97% protein, 61.05 mg/100 g β-carotene, 62.82 mg QE/g total flavonoid content, and 1789.65 mg/100 g calcium content.