Electronic Theses and Dissertation

Gayo Arabica coffee is a leading Indonesian commodity known for its distinctive flavor, which is influenced by its roasting level (light, medium, and dark). The determination of roasting levels has traditionally relied on the subjectivity of the roaster and organoleptic methods, leading to variability in product quality. Analytical methods such as Gas Chromatography–Mass Spectrometry (GC-MS) can accurately identify volatile compounds but are costly, complex, and require specialized laboratories and trained personnel, making them inefficient for direct field applications. This study aims to design a roasting level classification device for Gayo Arabica coffee using gas sensors integrated with an Arduino microcontroller. The system utilizes four gas sensors (MQ-3, MQ-8, MQ-9, MQ-135) to detect volatile compounds from fine coffee powder samples in a sealed chamber. Sensor data are transmitted to a laptop and analyzed using Linear Discriminant Analysis (LDA) and Random Forest to develop a classification model, which is then implemented into the device. Testing results show a classification accuracy of 96.67%, with high precision, recall, and F1-score values across all three classes. This system has proven effective in distinguishing roasting levels of Gayo Arabica coffee and holds potential for application in small- to medium-scale industries as an affordable alternative to subjective methods and high-cost laboratory equipment such as GC-MS.

Kopi Arabika Gayo merupakan komoditas unggulan Indonesia dengan cita rasa khas yang dipengaruhi oleh tingkat penyangraian (light, medium, dan dark). Penentuan tingkat sangrai selama ini masih bergantung pada subjektivitas roaster dan metode organoleptik, sehingga menimbulkan variabilitas mutu. Metode analisis seperti Gas Chromatography–Mass Spectrometry (GC-MS) mampu mengidentifikasi senyawa volatil secara akurat, namun mahal, kompleks, dan memerlukan laboratorium khusus serta tenaga ahli, sehingga kurang efisien untuk aplikasi langsung di lapangan. Penelitian ini bertujuan merancang alat klasifikasi tingkat sangrai kopi Arabika Gayo berbasis sensor gas dan mikrokontroler Arduino. Sistem menggunakan empat sensor gas (MQ-3, MQ-8, MQ-9, MQ-135) untuk mendeteksi senyawa volatil dari bubuk kopi berukuran fine dalam chamber tertutup. Data sensor dikirim ke laptop dan dianalisis menggunakan Linear Discriminant Analysis (LDA) dan Random Forest untuk membentuk model klasifikasi, yang kemudian diimplementasikan ke dalam alat. Hasil pengujian menunjukkan akurasi klasifikasi sebesar 96,67%, dengan nilai presisi, recall, dan F1-score yang tinggi untuk ketiga kelas. Sistem ini terbukti efektif dalam membedakan tingkat sangrai kopi Arabika Gayo, serta berpotensi diterapkan pada skala industri kecil dan menengah sebagai alternatif dari metode subjektif maupun alat laboratorium berbiaya tinggi seperti GC-MS.