Deep Learning Enabled Optimization and Mass Transfer Mechanism in Ultrasound-Assisted Enzymatic Extraction of Polyphenols from Tartary Buckwheat Hulls.

Tartary buckwheat hulls, a phenolic-rich by-product of buckwheat processing, offer great potential for resource utilization. In this study, ultrasound-assisted enzymatic extraction with two temperatures (40 °C and 50 °C) was employed to obtain phenolics from Tartary buckwheat hulls. Compared with the traditional extraction method (207 mg/100 g), ultrasound-assisted enzymatic extraction increased the total phenolic yield by 91.3% at 50 °C. Numerical simulations based on Fick's law indicated that enzyme pretreatment concentration positively correlated with the effective diffusion coefficient (De), which increased from 9.15 × 10 to 2.00 × 10 m/s at 40 °C. Meanwhile, the neuro-fuzzy inference system (ANFIS) successfully predicted the extraction yield under various ultrasonic conditions (R > 0.98). Regarding quantitative analysis of phenolic compounds in extracts, the results revealed that catechins and epicatechins were the most abundant in Tartary buckwheat hull. Additionally, phenolic acids rapidly diffused at higher temperatures (50 °C), and flavonoids were highly sensitive to temperature and enzyme synergy. Phenolic extracts exhibit significant potential for value-added applications in food processing, particularly in improving antioxidative stability, prolonging shelf life. This study provides a theoretical basis for green, efficient phenolic extraction from plant residues.