Please use this identifier to cite or link to this item:
https://repository.uksw.edu//handle/123456789/30839
Title: | Deteksi Cacat pada Isolasi Trafo secara Visual Menggunakan Algoritma Convolutional Neural Network (CNN) |
Authors: | Faudisyah, Alfendio Alif Hartomo, Kristoko Dwi |
Keywords: | Isolasi Trafo;Klasifikasi Gambar;CNN |
Issue Date: | 13-Jun-2023 |
Abstract: | Isolasi trafo adalah bahan dielektrik yang memiliki fungsi untuk memisahkan dua atau lebih penghantar listrik yang bertegangan. Kerusakan pada isolasi trafo akan menyebabkan gangguan kinerja trafo sehingga dapat membuat trafo mengalami kegagalan operasi atau bahkan kerusakan. Penelitian ini membangun suatu sistem yang dapat mengklasifikasikan gambar isolasi trafo cacat dan normal. Metode Convolutional Neural Network diimplementasikan dalam pembuatan model. Metode penelitian dimulai dengan melakukan perencanaan penelitian, pengumpulan dataset, preprocessing data, pembangunan model klasifikasi, training model, serta testing dan evaluasi. Berdasarkan hasil uji dengan standardisasi data ukuran 180 x 180 x 3 piksel menghasilkan accuracy 0.9913 untuk training, 0.9884 untuk testing, dan 1.00 untuk evaluasi. Hasil uji dengan standardisasi data ukuran 240 x 240 x 3 piksel menghasilkan accuracy 0.9798 untuk training, 0.9651 untuk testing, dan 0.94 untuk evaluasi. Berdasarkan penelitian yang telah dilakukan menunjukkan bahwa perbedaan standardisasi data dapat memengaruhi hasil dari performa model. Transformer insulation is a dielectric material that has the function of selling two or more voltage electrical conductors. Damage to the transformer insulation will cause interference with the performance of the transformer so that it can cause the transformer to experience operational failure or even damage. This research builds a system that can classify defective and normal transformer insulation images. The Convolutional Neural Network method is implemented in model building. The research method begins with conducting research planning, dataset collection, data preprocessing, classification of development models, training models, as well as testing and evaluation. Based on the test results with standardized data size 180 x 180 x 3 pixels, it produces an accuracy of 0.9913 for training, 0.9884 for testing, and 1.00 for evaluation. Test results with standardized data size 240 x 240 x 3 pixels produce an accuracy of 0.9798 for training, 0.9651 for testing, and 0.94 for evaluation. Based on the research that has been done, shows that differences in data standardization can affect the results of the model performance. |
URI: | https://repository.uksw.edu//handle/123456789/30839 |
Appears in Collections: | T1 - Informatics Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
T1_672019222_Judul.pdf | 1.07 MB | Adobe PDF | View/Open | |
T1_672019222_Isi.pdf Until 9999-01-01 | 579.83 kB | Adobe PDF | View/Open | |
T1_672019222_Daftar Pustaka.pdf | 423.56 kB | Adobe PDF | View/Open | |
T1_672019222_Formulir Pernyataan Persetujuan Penyerahan Lisensi Tugas Akhir dan Pilihan Embargo.pdf Restricted Access | 415.72 kB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.