IDENTIFICATION OF FERMENTATION STAGES OF COCOA BEANS USING A LOW COST ELECTRONIC NOSE

The cocoa bean is one of the most demanded agricultural commodities worldwide, and its fermentation is a key step that determines the sensory and chemical quality of chocolate. Traditionally, cocoa fermentation is evaluated through the cutting test, which involves cutting at least 200 beans per batch to assess their degree of fermentation and possible defects such as mold or insect damage. However, this method is destructive, time-consuming, and dependent on trained personnel. In addition, there is the influence of aroma on the fermentation of beans, which makes the need for trained specialists more important. Therefore, there is a growing need for fast, non-destructive, and reliable analytical methods to monitor fermentation. Recent advances in sensor technologies and data analysis have enabled the development of systems such as electronic noses (e-noses), which allow rapid assessment of food quality. In this study, a low-cost e-nose was used to monitor the fermentation degree of cocoa beans. Hybrid cocoa bean samples were collected from Bahia, Brazil, and fermented for 1 to 7 days under controlled humidity and temperature conditions. After fermentation, samples were sun-dried for two days and stored at 4 °C until analysis. The e-nose system consisted of a sampling chamber equipped with eight metal oxide semiconductor (MOS) gas sensors: MQ2, MQ3, MQ4, MQ6, MQ8, MQ9, MQ135, and MQ138, each with distinct sensitivities to volatile compounds such as alcohols, methane, carbon monoxide, ammonia, and aromatic compounds. For each analysis, three beans were placed in the chamber for two minutes, followed by two minutes of flushing with fresh air. All measurements were performed in triplicate. Data obtained from the sensors were processed using chemometric techniques with the PLS Toolbox for Matlab. Principal Component Analysis (PCA) was applied to visualize sample distribution along the scores plot and detect outliers. The PCA score plot revealed a temporal separation of cocoa samples according to fermentation time, indicating progressive biochemical transformations. PC1 (53.97%) captured the main variance related to fermentation progress, while PC2 (20.92%) described secondary variations, possibly due to variations between replicates or environmental factors. The exclusion of outliers improved class separation and revealed a coherent fermentation trajectory. These findings demonstrate that a low-cost electronic nose combined with chemometric analysis can effectively differentiate cocoa beans at different fermentation stages. This approach represents a promising, rapid, and non-destructive alternative for quality control and selection of cocoa beans in the chocolate industry, supporting more efficient and objective purchasing decisions.