Khumaidi, Agus (2017) Klasifikasi Image Sequence Hasil Pengelasan Menggunakan Metode Convolution Neural Network (Cnn) Untuk Non Distructive Test. Masters thesis, Institut Teknologi Sepuluh Nopember.
Preview |
Text
2215205204-Master_Thesis.pdf - Published Version Download (3MB) | Preview |
Abstract
Visual inspection adalah salah satu metode Non Distructive Test (NDT) yang sangat penting untuk proses uji hasil kualitas pengelasan, visual inspection dilakukan sebagai seleksi tahap awal sebelum hasil pengelasan dilanjutkan ke proses Distructive Test (DT) (Jurandir Primo, 2012). Proses visual inspection ini masih menggunakan cara manual yaitu dengan menggunakan pengelihatan manusia, sehingga hasil pengujian masih sangat subjektif.
Pada penelitian ini proses visual inspection akan dilakukan melalui pengolahan citra pada image sequence dengan menggunakan metode Convolution Neural Network (CNN) untuk klasifikasi cacat pada hasil pengelasan. Terdapat 4 vektor output klasifikasi yaitu good, over spatter, porosity, dan undercut. Arsitektur CNN yang digunakan memiliki 10 feature map, 1 layer tersembunyi dengan 64 hidden Neuron, jumlah iterasi 60 dan laju pembelajaran α yaitu 0,2. Hasil akurasi menunjukkan CNN mampu mengklasifikasi kategori cacat las dengan tingkat akurasi untuk data validasi bernilai 70,85% dari total 24 data yang diujikan.
===============================================================================================
Visual inspection is one of the most important Non Distructive Test (NDT) methods to test the results of weld quality. This inspection method is performed as initial selection before Distructive Test (DT) process (Jurandir Primo, 2012). Nowadays, visual inspection process is still using the manual way by using human vision, so the test results are still very subjective.
In this research, visual inspection process will be done through image processing in image sequence by using method of convolution neural network (CNN) to classify the result of welding defect. There are 4 vectors of classification output, these are good, over spatter, porosity, and undercut. The CNN architecture used has 10 feature maps, 1 hidden layer with 64 hidden Neurons, 60 numbers of iterations and learning rate α of 0.2. The results show that CNN is able to classify the welding defects with the accuracy rate of 70.85% for the validation data from the total of 24 data tested.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Visual Inspection, Non Distructive Test, Convolution Neural Network, Cacat Pengelasan |
| Subjects: | Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5105.546 Computer algorithms |
| Divisions: | Faculty of Industrial Technology > Electrical Engineering > 20101-(S2) Master Thesis |
| Depositing User: | Agus Khumaidi . |
| Date Deposited: | 10 Nov 2017 08:18 |
| Last Modified: | 05 Mar 2019 06:22 |
| URI: | http://repository.its.ac.id/id/eprint/43078 |
Actions (login required)
 |
View Item |