Electric Machine Load Estimation Using Deep Learning Based Infrared Thermography: a Review Paper

Accurate load estimation in electric machines is essential for ensuring operational efficiency, safety, and predictive maintenance. Traditional methods rely on intrusive sensors and transducers to measure electrical and mechanical parameters such as current and torque. Although effective, these techniques often require physical integration and modification of existing systems. Infrared thermography (IRT) presents a non-contact alternative by capturing heat emission patterns, enabling indirect assessment of machine load based on thermal signatures. With recent advancements in artificial intelligence, deep learning-especially Convolutional Neural Networks (CNNs)-has emerged as a powerful tool for analyzing thermal images and extracting complex patterns associated with load variations. This review explores the integration of IRT and deep learning for electric machine load estimation, highlighting key theoretical foundations, recent methodologies, performance benchmarks, and practical challenges. Additionally, the paper discusses the role of explainable AI and edge computing in enhancing real-time diagnostics, scalability, and transparency of model predictions.