SAMUDERA, MUHAMMAD ADJI (2024) RANCANG BANGUN PROTOTIPE PROGRAM DAMAGE DETECTION PADA FUSELAGE PESAWAT BERBASIS IMAGE PROCESSING. S1 thesis, Universitas Mercu Buana Jakarta.
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (HAL COVER)
01 Cover.pdf Download (577kB) |
|
Text (ABSTRAK)
02 Abstrak.pdf Download (328kB) |
|
Text (BAB I)
03 Bab 1.pdf Restricted to Registered users only Download (343kB) |
|
Text (BAB II)
04 Bab 2.pdf Restricted to Registered users only Download (553kB) |
|
Text (BAB III)
05 Bab 3.pdf Restricted to Registered users only Download (451kB) |
|
Text (BAB IV)
06 Bab 4.pdf Restricted to Registered users only Download (789kB) |
|
Text (BAB V)
07 Bab 5.pdf Restricted to Registered users only Download (243kB) |
|
Text (DAFTAR PUSTAKA)
08 Daftar Pustaka.pdf Restricted to Registered users only Download (243kB) |
|
Text (LAMPIRAN)
09 Lampiran.pdf Restricted to Registered users only Download (328kB) |
Abstract
Inspection is an important part of aircraft maintenance process. With the wide area of inspection objek, lack of licensed, and authorized engineer resource, then will be an obstruction in the maintenance process. One of the detection’s object is fuselage that visually inspected. The example of damage on fuselage are nick, dent, and scratch. This research aims to design a prototype of damage detection program that able to detect types of damage on aircraft fuselage to contribute to inspection process. The model used in this damage detection design is YOLOv5 algorithm. This research is carried out in two steps. The first step is using 600 dataset from internet source and self-taken. The second steps is using 232 dataset that selected and have good quality. The datasets divided into training dataset and testing dataset with 9 : 1 comparison. The datasets are trained with variation of batch size and image size hyperparameter which aims to get the best performance. The best performance result will be tested and simulated in google colaboratory platform. The performance’s result the precision of 0.88, recall of 0.917, F1 score of 0.898, and accuracy of 0.986 with the variety of batch size value of 4, image size of 640 pixel, 500 epoch from the second step of training. Then, the best permormance model tested and simulated in google colaboratory platform with 23 testing datasets. The prototype of damage detection program has accuracy of 88%. Keyword : Object Detection, Hyperparameter, YOLOv5 Inspeksi merupakan bagian penting dalam proses maintenance pesawat. Dengan cakupan area inspeksi yang luas dan terbatasnya engineer berlisensi yang memiliki otoritas melakukan inspeksi, maka menjadi hambatan pada proses maintenance pesawat. Salah satu objek deteksi yakni fuselage pesawat yang diinspeksi secara visual. Jenis damage pada fuselage pesawat di antaranya adalah nick, dent, dan scratch. Penelitian ini bertujuan membuat sebuah program mampu mendeteksi jenis – jenis damage pada fuselage sehingga dapat membantu proses inspeksi. Model yang digunakan pada perancangan damage detection ini adalah algoritma YOLOv5. Adapun penelitian dilakukan dengan dua tahap. Tahap pertama adalah training menggunakan 600 dataset gambar dari internet dan pengambilan gambar sendiri. Tahap kedua menggunakan 232 dataset gambar terpilih dengan kualitas baik. Dataset dibagi menjadi training dataset dan testing dataset dengan perbandingan 9 : 1. Dataset gambar dilatih dengan variasi hyperparameter batch size dan image size untuk memperoleh performa pelatihan terbaik. Performa hasil pelatihan terbaik diuji dan disimulasikan pada google colaboratory. Performa hasil pelatihan terbaik mendapatkan nilai precision 0.88, recall 0.917, f1 score 0.898, dan accuracy 0.986 dengan variasi nilai batch size 4, image size 640 pixel, dan 500 epoch yang diperoleh dari tahap pelatihan kedua. Kemudian model yang memiliki performa terbaik diuji dan disimulasikan pada google colaboratory dengan 23 testing dataset. Prototipe program damage detection yang dibangun memperoleh nilai accuracy sebesar 88%. Kata Kunci : Deteksi Objek, Hyperparameter, YOLOv5
Actions (login required)
 |
View Item |
