Pengembangan Model Hybrid Deep Learning dan CatBoost untuk Analisis Kelayakan Oli Mesin melalui Objek Dipstick pada PT Telkom Infrastruktur Indonesia

Miskiyah, Miskiyah and Izzati, Hikmia Sofia Nur (2026) Pengembangan Model Hybrid Deep Learning dan CatBoost untuk Analisis Kelayakan Oli Mesin melalui Objek Dipstick pada PT Telkom Infrastruktur Indonesia. Project Report. [s.n.]. (Unpublished)

Text 5025231119_5025231147-Project_Report.pdf - Accepted Version Download (845kB)

Abstract

Pengecekan kualitas oli mesin secara manual melalui dipstick merupakan proses yang bersifat repetitif dan memerlukan ketelitian visual yang tinggi. Kondisi ini menyebabkan inspeksi menjadi subjektif dan rentan terhadap kesalahan manusia (human error), terutama saat menghadapi beban kerja yang tinggi. Penelitian ini bertujuan untuk mengotomatisasi proses klasifikasi kualitas oli menggunakan teknologi Computer Vision dan Machine Learning. Metode yang diusulkan menggabungkan arsitektur YOLOv11 untuk instance segmentation area oli dan algoritma CatBoost Classifier untuk analisis fitur warna secara objektif. Melalui penerapan Gamma Correction dan ruang warna HSV, sistem mampu mempertahankan konsistensi deteksi pada berbagai kondisi pencahayaan. Hasil pengujian menunjukkan bahwa YOLOv11 memperoleh precision 0,8570 dan recall 0,9333, sedangkan CatBoost mencapai akurasi klasifikasi sebesar 83%, sehingga mampu menghasilkan penilaian kualitas oli yang lebih konsisten dan efisien.
================================================================================================================================
Manual inspection of engine oil quality using a dipstick is a repetitive process that requires a high level of visual accuracy. This condition makes the inspection subjective and prone to human error, particularly under high workloads. This study aims to automate the engine oil quality classification process using Computer Vision and Machine Learning technologies. The proposed method combines the YOLOv11 architecture for engine oil area instance segmentation with the CatBoost Classifier algorithm for objective color feature analysis. By applying Gamma Correction and the HSV color space, the system is able to maintain consistent detection performance under various lighting conditions. Experimental results show that YOLOv11 achieved a precision of 0.8570 and a recall of 0.9333, while the CatBoost Classifier attained a classification accuracy of 83%, enabling a more consistent and efficient assessment of engine oil quality.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	CatBoost, Computer Vision, Instance, Kualitas Oli, Segmentation, YOLOv11
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	HIKMIA SOFIA NUR IZZATI
Date Deposited:	09 Jul 2026 04:36
Last Modified:	09 Jul 2026 04:36
URI:	http://repository.its.ac.id/id/eprint/134553

Actions (login required)

View Item