The ANFIS-RSM based multi-objective optimization and modelling of ultrasound-assisted extraction of polyphenols from jamun fruit (Syzygium cumini).

Abstract

Given their potential as natural substitutes for artificial additives and their health advantages, the extraction of bioactive substances like polyphenols from plant sources is becoming more and more significant. Nevertheless, it is still difficult to achieve effective extraction with minimal time and energy. In order to optimize polyphenol extraction from ripe jamun fruit pulp, including traditional and ultrasound-assisted methods, this study assessed the prediction power of response surface methodology (RSM) and adaptive neuro-fuzzy inference systems (ANFIS). It examined how temperature, process time, solvent type, and extraction method affected the yield of extracted polyphenols. Analysis of variance (ANOVA) indicated that solvent type (F-value = 292.15) was the most significant factor influencing polyphenol extraction. Numerical optimization identified optimal conditions for maximizing phenolic compound extraction: a process temperature of 45 °C, a duration of 65 min under ultrasound, using methanol as the solvent (desirability of 0.935 and a realization rate of 95 % of the maximum possible). Imposing minimum temperature and process time conditions will yield the same optimal process parameters as before, achieving 89 % of the maximum possible while significantly reducing the process time from 65 min to just 5 min (desirability 0.953). For each of the six process-solver conditions, optimal ANFIS models were determined by analyzing the number and type of input membership functions, the output membership function, and the selected optimization and defuzzification methods, based on the highest correlation between actual and predicted data, along with the lowest error rates. Statistical analysis confirmed the effectiveness of both RSM and ANFIS in modeling polyphenol extraction from ripe jamun fruit. Error indices demonstrated that ANFIS (R² = 0.8490-0.9989) outperformed RSM (R² = 0.9265) in predictive capability, underscoring the relative superiority of ANFIS.