Thermal Image-Based Fault Detection Using Machine Learning and Deep Learning in Industrial Machines: Issues-Challenges and Emerging Trends

Abstract

Thermal imaging has long been utilized across industries to maintain electrical equipment and detect faults in machines, ensuring their reliable operation. Infrared thermography, or thermal imaging, has emerged as a powerful, non-contact, non-destructive monitoring tool. Unlike RGB imaging, it offers a non-intrusive approach that respects privacy and does not rely on ambient lighting. As heat signatures vary based on defects or machine conditions, infrared imaging is essential for effective condition monitoring. This survey study introduces the principles and applications of thermal imaging in fault detection and a comprehensive analysis of recent developments and approaches in the field. The survey encompasses various machine types, including electrical systems, mechanical components, and industrial equipment. Multiple techniques, algorithms, and deep learning models for image processing, feature extraction, and fault classification are extensively discussed. The paper also includes an in-depth literature review on fault detection in distinct machines, such as induction motors, rotating machines, and transformers. The survey aims to identify open challenges and issues in fault detection for different machines, such as induction motors and transformers.