STUDY AND MODELING OF THE HYDRATION KINETICS OF TARWI (LUPINUS MUTABILIS)

The hydration behavior of tarwi (Lupinus mutabilis Sweet) grains was characterized through an integrated approach combining gravimetric measurements, three-dimensional (3D) morphometric reconstruction, near-infrared (NIR) spectroscopy, and kinetic modeling. Moisture uptake followed the typical non-linear hydration pattern observed in legumes, with a rapid initial absorption stage (0–15 min) driven by capillary forces, followed by a slower diffusion-controlled phase that approached equilibrium after approximately 45 min, reaching a final moisture content of 226.5 kg water kg⁻¹ dry solids at 570 min. Three-dimensional digital twins generated by photogrammetric reconstruction revealed irreversible structural deformation during hydration, with grain volume and surface area increasing by 115% and 65%, respectively, confirming the importance of considering real geometric changes during water absorption. Effective moisture diffusivity, estimated using Fick's second law with realistic oblate spheroidal geometry, was 2.23 × 10⁻⁹ m² s⁻¹, consistent with values reported for protein-rich legumes. NIR spectra collected between 900 and 1700 nm successfully captured hydration-induced molecular changes, particularly around 1450 nm, corresponding to the O–H stretching overtone associated with water absorption. A hybrid kinetic–spectroscopic model significantly outperformed the classical Peleg equation, improving predictive performance from R² = 0.945 (RMSE = 0.1690) to R² = 0.994 (RMSE = 0.0551) through Random Forest-based correction of NIR spectral information. Color measurements in the CIELAB space revealed increases in lightness and chroma, indicating a cream-to-yellowish transition likely associated with alkaloid migration and seed coat modification. Water activity increased from 0.60 to 0.94, reflecting progressive moisture uptake and the formation of a hydrated seed coat–pericarp interface. Overall, the integration of digital morphometry, NIR spectroscopy, machine learning, and diffusion modeling provides a robust and highly accurate framework for describing tarwi hydration, with potential applications in process optimization, quality control, and digital food engineering.