Modeling and optimization of pectin extraction from banana peel using artificial neural networks (ANNs) and response surface methodology (RSM)

Abstract

In the present study, the extraction of pectin from banana peel (Musa sp.) was optimized using artificial neural network and response surface methodology on the yield and degree of esterification obtained using microwave-assisted extraction methods. The individual, quadratic and interactive effect of process variables (temperature, time, liquid–solid ratio and pH) on the extracted pectin yield and DE of the extract were studied. The results showed that properly trained artificial neural network model was found to be more accurate in prediction as compared to response surface model method. The optimum conditions were found to be temperature of 60 °C, extraction time of 102 min, liquid–solid ratio of 40% (v/w) and pH of 2.7 and within the desirable range of the order of 0.853. The yield of pectin and degree of esterification under these optimum conditions were 14.34% and 63.58, respectively. Temperature, time, liquid–solid ratio and pH revealed a significant (p < 0.05) effect on the pectin yield and degree of esterification. Based on the value of methoxyl content and degree of esterification the extracted pectin was categorized as high methoxyl pectin. Generally, the findings of the study show that banana peel can be explored as a promising alternative for the commercial production of pectin.