Main content area

Determining degree of roasting in cocoa beans by artificial neural network (ANN)‐based electronic nose system and gas chromatography/mass spectrometry (GC/MS)

Tan, Juzhong, Kerr, William L
Journal of the science of food and agriculture 2018 v.98 no.10 pp. 3851-3859
chocolate, cocoa beans, electronic nose, farmers, flavor, gas chromatography-mass spectrometry, headspace analysis, neural networks, prediction, roasting, texture, water content
BACKGROUND: Roasting is a critical step in chocolate processing, where moisture content is decreased and unique flavors and texture are developed. The determination of the degree of roasting in cocoa beans is important to ensure the quality of chocolate. Determining the degree of roasting relies on human specialists or sophisticated chemical analyses that are inaccessible to small manufacturers and farmers. In this study, an electronic nose system was constructed consisting of an array of gas sensors and used to detect volatiles emanating from cocoa beans roasted for 0, 20, 30 and 40 min. The several signals were used to train a three‐layer artificial neural network (ANN). Headspace samples were also analyzed by gas chromatography/mass spectrometry (GC/MS), with 23 select volatiles used to train a separate ANN. RESULTS: Both ANNs were used to predict the degree of roasting of cocoa beans. The electronic nose had a prediction accuracy of 94.4% using signals from sensors TGS 813, 826, 822, 830, 830, 2620, 2602 and 2610. In comparison, the GC/MS predicted the degree of roasting with an accuracy of 95.8%. CONCLUSION: The electronic nose system is able to predict the extent of roasting, as well as a more sophisticated approach using GC/MS. © 2018 Society of Chemical Industry