Scholarlink

<p>Coffee is one of the most widely consumed beverages worldwide, and its high economic value makes it vulnerable to adulteration with low-cost materials. Detecting such fraud is particularly challenging when adulterants are present at low levels and do not cause obvious sensory changes. In this study, an electronic nose (e-nose) system combined with multivariate analysis and machine learning was evaluated for the detection of common adulterants in medium-dark roasted Arabica coffee. Coffee samples were adulterated with soybean, wheat, and barley flour at five weight percentages (10–50%). Samples were prepared using a laboratory extraction procedure designed to enhance volatile compound release for e-nose analysis, rather than to replicate standard brewing conditions. An e-nose equipped with eight metal oxide semiconductor sensors was used to record aroma fingerprints, which were analyzed using principal component analysis (PCA), linear discriminant analysis (LDA), and artificial neural networks (ANN). The results showed that ANN consistently outperformed LDA, achieving higher classification accuracy across all adulteration scenarios. These findings demonstrate the potential of combining e-nose technology with nonlinear machine learning models as a rapid and reliable approach for detecting coffee adulteration, contributing to improved quality control and consumer protection.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Purity identification of Arabica coffee against cereal flour adulteration using an electronic nose and machine learning

  • 期刊论文

  • 发表日期：

    2026年4月29日

查看全文
  • Saleh Azari,
  • Esmaeil Mirzaee-Ghaleh,
  • Hekmat Rabbani,
  • Hamed Karami

摘要

Coffee is one of the most widely consumed beverages worldwide, and its high economic value makes it vulnerable to adulteration with low-cost materials. Detecting such fraud is particularly challenging when adulterants are present at low levels and do not cause obvious sensory changes. In this study, an electronic nose (e-nose) system combined with multivariate analysis and machine learning was evaluated for the detection of common adulterants in medium-dark roasted Arabica coffee. Coffee samples were adulterated with soybean, wheat, and barley flour at five weight percentages (10–50%). Samples were prepared using a laboratory extraction procedure designed to enhance volatile compound release for e-nose analysis, rather than to replicate standard brewing conditions. An e-nose equipped with eight metal oxide semiconductor sensors was used to record aroma fingerprints, which were analyzed using principal component analysis (PCA), linear discriminant analysis (LDA), and artificial neural networks (ANN). The results showed that ANN consistently outperformed LDA, achieving higher classification accuracy across all adulteration scenarios. These findings demonstrate the potential of combining e-nose technology with nonlinear machine learning models as a rapid and reliable approach for detecting coffee adulteration, contributing to improved quality control and consumer protection.