Sistem Pendeteksi Clip Roof Mobil Pada Line Sealer Metode Convolutional Neural Network Untuk Mengatasi Kelalaian Manpower

Alfarabi, Deanova Rakhazl (2023) Sistem Pendeteksi Clip Roof Mobil Pada Line Sealer Metode Convolutional Neural Network Untuk Mengatasi Kelalaian Manpower. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 2040201045_Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (9MB)

Abstract

PT. Toyota Motor Manufacturing Indonesia (TMMIN), dalam memproduksi setiap unit mobil, membagi menjadi 4 proses utama yaitu pembentukan body frame, pengelasan, pengecatan, dan perakitan. Salah satu proses yang membutuhkan keteletian tinggi adalah proses pengecatan. Pada proses ini, meliputi banyak tahap, mulai dari pemberian lapisan electrodeposition, sealer, lapisan primer, lapisan base, dan lapisan clear. Pada tahap pemberian sealer, terdapat pemasangan clip roof pada model Innova Reborn, Innova Zenix dan Fortuner. Jumlah clip roof yang dipasang pada setiap model tidak tergantung pada tipe model, sehingga pada jumlah clip roof yang terpasang terdapat keselahan yang disebabkan kelalaian manpower. Untuk mengatasi permasalahan tersebut, maka pada Proyek Akhir ini dibuat sistem berbasis Convolutional Neural Network (CNN) sebagai mendeteksi objek. Algoritma yang digunakan dalam mendekteksi objek secara akurat dan efisien salah satunya You Only Look Once (YOLO) dengan versi delapan. Sistem deteksi dapat mencegah kesalahan pada proses produksi dalam mendeteksi jumlah clip roof yang terpasang. Ketika jumlah clip roof yang terpasang tidak sesuai dengan jumlah yang sudah ditentukan, maka akan muncul peringatan berupa alarm dan menghentikan proses produksi, untuk memberi peringatan jika ada clip roof belum terpasang atau jumlah yang tidak sesuai. Hasil dari pengujian sistem deteksi YOLO dengan 200 sampel gambar mendapatkan score precision 0,992; score recall 0,853; score specificity 0,98; F1-score 0,917; score mean avarage precision 0,847; score akurasi 0,885 dan pengujian sistem keseluruhan dengan 300 unit mobil mendapatkan true dari semua sampel pengujian. Dengan begitu, sistem deteksi dapat digunakan sebagai alat evaluasi kinerja dari manpower ketika proses pemasangan clip roof.
===================================================================================================================================
PT Toyota Motor Manufacturing Indonesia (TMMIN), in producing each car unit, divides it into 4 main processes, namely body frame formation, welding, painting, and assembly. One of the processes that requires high accuracy is the painting process. This process includes many stages, starting from the application of electrodeposition layer, sealer, primer layer, base layer, and clear layer. At the sealer stage, there is the installation of roof clips on Innova Reborn, Innova Zenix and Fortuner models. The number of clip roofs installed on each model does not depend on the type of model, so that the number of clip roofs installed is a mistake caused by manpower negligence. To overcome these problems, in this Final Project, a Convolutional Neural Network (CNN) based system is made to detect objects. The algorithm used in detecting objects accurately and efficiently is You Only Look Once (YOLO) with version eight. The detection system can prevent errors in the production process in detecting the number of clip roofs installed. When the number of clip roofs installed does not match the predetermined number, an alarm will appear and stop the production process, to warn if there is a clip roof that has not been installed or the number that does not match. The results of testing the YOLO detection system with 200 sample images get a precision score of 0.992; a recall score of 0.853; a specificity score of 0.98; an F1-score of 0.917; a mean avarage precision score of 0.847; an accuracy score of 0.885 and overall system testing with 300 car units getting true from all test samples. Therefore, the detection system can be used as a performance evaluation tool for manpower during the clip roof installation process.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Convolutional Neural Network, Deteksi Clip Roof, Kelalaian Manpower, Sistem Deteksi Objek, You Look Only Once, Convolutional Neural Network, Clip Roof Detection, Manpower Negligence, Object Detection System, You Look Only Once
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7888.3 Digital computers T Technology > TS Manufactures > TS174 Maintainability (Engineering) . Reliability (Engineering) T Technology > TS Manufactures > TS176 Manufacturing engineering. Process engineering (Including manufacturing planning, production planning) T Technology > TS Manufactures > TS183 Manufacturing processes. Lean manufacturing.
Divisions:	Faculty of Vocational > 36304-Automation Electronic Engineering
Depositing User:	Deanova Rakhazl Alfarabi
Date Deposited:	19 Aug 2024 06:31
Last Modified:	19 Aug 2024 06:31
URI:	http://repository.its.ac.id/id/eprint/115455

Actions (login required)

View Item