Numeric and machine learning modeling of mass transfer in radish brining with pulsed electric field and plasma-activated water pretreatments

Abstract

Brining requires a high NaCl concentration and long duration, complicating quality management and necessitating accurate prediction models. This study explores the synergistic effects of low-intensity pulsed electric field (PEF) and plasma-activated water (PAW) pretreatments on radish brining in 5 % and 10 % NaCl solutions, predicting mass transfer using numeric and machine learning models while evaluating microstructural changes by the pretreatments. The combined PEF (0.2–0.4 kV/cm) and PAW (oxidation-reduction potential: 496–531 mV) pretreatment reduced the brining rate by up to 42 % compared the rate obtained without the treatment during the initial stage of brining. This effect was attributed to enhanced cell permeability, as evidenced by increased electrical conductivity and Z-index. The Ensemble model, a machine learning model, provided competitive prediction on the moisture loss and NaCl uptake rate compared to numeric models, with a root mean square error of 0.04 for moisture loss and 0.01 for NaCl uptake, and a higher R² (0.99 for moisture loss, 0.99 for NaCl uptake). PEF and PAW individually enhanced moisture loss (up to 43.81 %) and NaCl uptake (up to 8.17 %), with combined treatments showing synergistic effects, particularly at 10 % NaCl due to greater osmotic gradients. The combined application of PEF and PAW pretreatments increased the hardness and decreased the cutting force of radish. These findings suggest that PEF and PAW pretreatments enhance mass transfer efficiency in the brining process and the process can be adequately predicted using the machine learning model.