Sistem Deteksi Kelengkapan Stut Bolt dan Hole dengan Metode Convolutional Neural Network (CNN) untuk Inspeksi Panel Dash pada Industri Manufaktur Otomotif

Prayogi, Aldan (2026) Sistem Deteksi Kelengkapan Stut Bolt dan Hole dengan Metode Convolutional Neural Network (CNN) untuk Inspeksi Panel Dash pada Industri Manufaktur Otomotif. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040221034-Undergraduate-Thesis.pdf Restricted to Repository staff only Download (40MB) | Request a copy

Abstract

Industri manufaktur otomotif memberikan kontribusi signifikan terhadap perekonomian Indonesia. Data nilai ekspor manufaktur pada tahun 2024 mencapai 6.760,5 juta dolar Amerika, dengan pertumbuhan sektor sebesar 4,46% pada tahun 2023. Kondisi ini mendorong perusahaan yang bergerak pada manufaktur otomotif ini untuk menjaga kualitas produk sekaligus meningkatkan efisiensi proses produksi. Pada lini Under Body Check, khususnya pos Criple and Model yang merupakan stasiun pemeriksaan bentuk dan bodi kendaraan, operator menghadapi beban kerja tinggi saat memeriksa panel dash kendaraan. Jumlah titik inspeksi mencapai 18.492 titik untuk tipe A, 11.776 titik untuk tipe B, dan 9.240 titik untuk tipe C. Posisi kerja yang tinggi menyebabkan operator bekerja dengan postur kurang ergonomis selama sekitar 7 jam, sehingga berpotensi menurunkan ketepatan pemeriksaan dan meningkatkan risiko human error. mengatasi permasalahan tersebut, penelitian ini mengembangkan sistem inspeksi otomatis berbasis Convolutional Neural Network menggunakan algoritma YOLO untuk mendeteksi titik inspeksi kendaraan. Hasil pengujian sistem deteksi menggunakan 200 sampel menunjukkan bahwa model YOLOv8 memperoleh precision sebesar 92,09%, recall 98,89%, F1-score 95,37%, dan mAP50-95 sebesar 91,07%. Pengujian keseluruhan sistem menggunakan 127 sampel menghasilkan akurasi klasifikasi total sebesar 93,70%. Selain itu, sistem otomatis berhasil menurunkan rata-rata waktu inspeksi dari 46,46 detik menjadi 2,38 detik per unit, yang setara dengan peningkatan efisiensi waktu sebesar 94,87%.
=====================================================================================================================================
The automotive manufacturing industry plays a significant role in Indonesia's economy. In 2024, the manufacturing export value reached USD 6.76 billion, while the sector recorded a growth rate of 4.46% in 2023. These conditions encourage companies such as a company engaged in automotive manufacturing to maintain product quality while improving production efficiency. On the Under Body Check (UBC) line, particularly at the Criple and Model station where the vehicle body shape is inspected, operators face a high workload during dashboard panel inspections. The number of inspection points reaches 18,492 for Type A, 11,776 for Type B, and 9,240 for Type C vehicles. The elevated working position requires operators to maintain non-ergonomic postures for approximately seven hours per shift, potentially reducing inspection accuracy and increasing the risk of human error. To address these challenges, this study developed an automated inspection system based on a Convolutional Neural Network (CNN) using the YOLO algorithm to detect vehicle inspection points. Experimental results obtained from 200 test samples showed that the YOLOv8 model achieved a precision of 92.09%, a recall of 98.89%, an F1-score of 95.37%, and an mAP@50–95 of 91.07%. Furthermore, the overall system evaluation using 127 samples achieved a total classification accuracy of 93.70%. The proposed system also reduced the average inspection time from 46.46 seconds to 2.38 seconds per unit, representing a 94.87% reduction in inspection time.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	inspeksi otomatis, CNN, YOLO, panel dash, industri otomotif.
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing. T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Aldan Prayogi
Date Deposited:	10 Aug 2026 02:24
Last Modified:	10 Aug 2026 02:24
URI:	http://repository.its.ac.id/id/eprint/144174

Actions (login required)

View Item