Development of a low-cost e-nose to assess aroma profiles: An artificial intelligence application to assess beer quality

Abstract The assessment of aromas in beer is critical to assess its quality since it could be used as an indicator of contamination or faults, which will directly influence consumers’ acceptability. Traditional techniques to evaluate aromas are time-consuming, require special training, costly equipment, and trained personnel. Therefore, this study aimed to develop a portable, low-cost electronic nose (e-nose) coupled with machine learning modeling to predict aromas in beer. Nine different gas sensors were used i) ethanol, ii) methane, iii) carbon monoxide, iv) hydrogen, v) ammonia/alcohol/benzene, vi) hydrogen sulfide, vii) ammonia, viii) benzene/alcohol/ammonia and ix) carbon dioxide. Output data were assessed for significant differences using ANOVA and least significant differences as post hoc test (α = 0.05). Two artificial neural network (ANN) models were also developed to predict i) the peak area of 17 different volatile aromatic compounds (Model 1) obtained from gas chromatography–mass spectroscopy (GC–MS) and ii) the intensity of ten sensory descriptors acquired from a sensory session with 12 trained panelists. Results from the ANOVA showed that there were significant differences between the samples used, which showed that the e-nose was able to discriminate samples. The resulting ANN models were highly accurate with correlation coefficients of R = 0.97 (Model 1) and R = 0.93 (Model 2). The combined method using the developed e-nose and the ANN models could be used by the industry as a low-cost, rapid, reliable and effective technique for beer quality assessment within the production line. This may also be calibrated for its use in other foods and beverages.