Hybrid Swin Transformer approach for hydrophobicity monitoring of outdoor polymeric insulators in transmission lines

Akash, R., Sarathi, R. and Haddad, Manu ORCID: https://orcid.org/0000-0003-4153-6146 2025. Hybrid Swin Transformer approach for hydrophobicity monitoring of outdoor polymeric insulators in transmission lines. IEEE Transactions on Dielectrics and Electrical Insulation 10.1109/tdei.2025.3548969

Preview

PDF - Accepted Post-Print Version Download (821kB) | Preview

Abstract

Hydrophobicity is one of the vital properties of outdoor polymeric insulators which prevents the water accumulation on the insulator’s surface. Due to extreme environmental conditions, polymeric insulators tend to lose its hydrophobicity property. Accurate hydrophobicity classification is essential to understand the condition of the insulators in operation within the power system network. In this study, a Hybrid Swin Transformer (HST) is adopted to enhance the hydrophobicity classification accuracy. Traditional image classification methods struggle with variation in droplet shapes, sizes and surface patterns which can complicate the classification process but the model adopted in this study integrates hybrid shifted windows with advanced vision transformer techniques to capture both short-range and long-range dependencies in images, providing a more robust understanding of complex visual patterns associated with different hydrophobicity levels. This Hybrid approach has been evaluated with laboratory generated dataset (according to IEC Standard 62073) and a online available dataset. Extensive experimental and analytical results demonstrate that the hybrid model outperforms existing state-of-the-art techniques. Additionally, an android application for image classification was developed with a simple graphical user interface (GUI) to enhance the maintenance of the insulators.

Item Type:	Article
Date Type:	Published Online
Status:	In Press
Schools:	Schools > Engineering
Additional Information:	License information from Publisher: LICENSE 1: URL: https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html, Start Date: 2025-01-01
Publisher:	Institute of Electrical and Electronics Engineers
ISSN:	1070-9878
Date of First Compliant Deposit:	21 March 2025
Date of Acceptance:	3 November 2024
Last Modified:	21 Mar 2025 12:30
URI:	https://orca.cardiff.ac.uk/id/eprint/176954

Actions (repository staff only)

Edit Item