Gas Sensors and Machine Learning for Quality Evaluation of Grape Spirits (Pisco)

Bolivar R.; Sarmiento-Calisaya E.; Alzamora G.S. · Communications in Computer and Information Science (2023) · DOI: 10.1007/978-3-031-35445-8_12

Resumen

Pisco is Peru’s national drink (The Peruvian state has the “Appellation of Origin”). It is distilled from grapes grown only in the regions of Lima, Ica, Arequipa, Moquegua or Tacna, and using methods that preserve the traditional principles of quality. Peru currently exports pisco to large-market countries (e.g. the European Union, the United States, Mexico, Canada, Australia, etc.); however, according to ADEX (The Peruvian Exporters Association), despite the fact that exports grow annually, losses have increased, mainly due to penalties and taxes for adulterated pisco. The adulterated pisco contains a high concentration of methanol and other types of alcohols, which are harmful to human health, and consequently it damages the economics of the beverage industry in Peru. This work addresses the design and development of a prototype tool to classify pisco and adulterated pisco, which is based on i) a gas sensor matrix that allows the identification of volatile compounds and congeners of pisco, ii) a data acquisition module based on a low-cost ARM (Advanced RISC Machine) micro-controller (Arduino), and iii) a classification model based on machine learning techniques. In the evaluation, pisco samples were used; and the results showed 97.29% of accuracy for the problem of classification and 7.43 s for the problem of training time. Therefore, it provides insights that the prototype is useful, low-cost, easy to use and fast. Thus, its development continues.

Instituciones: Universidad Nacional de San Agustín de Arequipa; Universidad Nacional del Altiplano