Artificial neural network-assisted process modeling and glow discharge plasma modulation of cashew nut protein concentrate properties

The production of cashew nut protein concentrate was investigated using response surface methodology, artificial neural networks, and genetic algorithms, followed by glow discharge plasma as a post-processing strategy to modify the ingredient. The models showed that refrigerated resting times affected protein concentration and yield differently. A practical condition, rather than the mathematical optimum, was selected from this trade-off (RT 1 = 0 min; RT 2 = 10 min) and experimentally validated, producing 68.58 ± 0.35% protein concentration and 54.74 ± 0.52% protein yield. The optimized concentrate (OC) was treated by glow discharge plasma at synthetic air flow rates of 10 and 30 mL/min. Powder-flow measurements indicated reduced wall friction after optimization and plasma exposure, while cohesion was highest in OC-GDP30. Quantitative analysis of the amide I region revealed treatment-dependent changes in relative β-sheet and β-turn contributions, while scanning electron microscopy showed greater particle-surface heterogeneity after plasma exposure. Differential scanning calorimetry indicated distinct residual thermal profiles, although these results were interpreted descriptively. Plasma generated distinct functional profiles rather than a uniform improvement across properties. OC-GDP10 showed the strongest emulsion-related performance, with an emulsifying activity index of 57.56 m 2 /g and emulsion stability of 114.45 min, whereas OC-GDP30 showed the highest solubility (56.16%) and foaming capacity (20.00%). Multivariate analysis separated the treatments according to their functional, colorimetric, and electrokinetic responses. These results indicate that glow discharge plasma can selectively modify cashew nut protein concentrate, with the two flow rates producing distinct profiles associated with emulsion stabilization or with greater solubility and foaming capacity.