Apri Junaidi, Siti Zaiton Mohd Hashim, Sofyan Sofyan, M.P. Lukman, Usman Usman, Dharma Aryani, A.R. Idris, Ahmad Yani, Fathahillah, Ade Rahmat Iskandar, Haris Suhendar
This study addresses the challenge of detecting low- to medium-level defects in ceramic insulators through deep learning-based classification, employing the MobileNet architecture. The ceramic insulator's susceptibility to various forms of damage necessitates an accurate and efficient detection system for ensuring the reliability of power transmission systems. In this research, an initial class imbalance was addressed by obtaining 22 images for the defective class and 6 images for the normal class, with subsequent data augmentation resulting in a balanced dataset of 6,000 images per class. The MobileNet transfer learning model was employed, achieving an impressive accuracy rate of 92 percent, surpassing prior studies using CNN architectures. The comprehensive training configuration, including a seed value of 42, batch size of 32, 50 epochs, and the use of the Adam optimizer with a categorical crossentropy loss function, contributed to the robust performance of the model. Future work could focus on refining data augmentation strategies, investigating the model's resilience to real-world conditions, and staying abreast of evolving deep learning advancements to further enhance model performance. This research establishes a promising foundation for advancing defect detection in ceramic insulators through deep learning methodologies. © 2026 Author(s).
Faculty of Computing, Universiti Teknologi Malaysia, 81310 UTM Johor Bahru, Johor, Malaysia; Faculty of Electrical Engineering, Universiti Teknologi Malaysia, 81310 UTM Johor, Johor, Malaysia; Depatment of Electrical Engineering, State Polytechnic of Ujung Pandang, Makassar, 90245, Indonesia; Faculty of Engineering, Universitas Negeri Makassar, Makassar, Indonesia; Telecomunication Engineering, Intitute Teknologi Telkom Jakarta, Jakarta, Indonesia; Departement of Physics, Universitas Negeri Jakarta, Jakarta, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock