Development and optimization of a cocoa extraction treatment by means of the response surface methodology (RSM) and artificial neural networks (ANN)

Abstract

Cocoa is a well-known source of bioactive polyphenols, especially flavanols, which are reported to have a potential health benefit for human beings. For this reason, it is of great importance that a high yield of flavanols can be achieved using treatments during the extraction of cocoa and its by-products, leading to extracts that can be used for all kinds of (food) applications. Most prominent for recent years is the cold-extraction process developed by the Swiss company Oro de Cacao AG to enable a better chocolate quality ( Chetschik et al., 2019 ). The aim of the present study was to develop and optimize aqueous cocoa extracts as a prospective approach to recover more flavanols during the extraction. Treatment optimization was developed using response surface methodology (RSM) and artificial neural networks (ANN). It was found that the content of cocoa flavanols in an aqueous extract could be increased by acidic pH conditions. In fact, using Box-Behnken design and associated analysis of variance, it was determined that a change in pH has the largest influence on the final flavanol content in the extract, while temperature exerts a secondary influence on the content. Furthermore, models were developed using RSM and ANN methods to predict the content of flavanols in the extract with the parameters pH modification , temperature , and bentonite addition , which is used for purifying the extract. When comparing the two models, ANN had a higher and very well predictive performance. • Extraction of cocoa needs to be optimized for yielding more phenolic compounds. • Acidic pH conditions enabled increasing flavanol content in the cocoa extract. • Artificial neuronal network models can predict extraction success.