Advances in prognostics and health management of light emitting diodes: A comprehensive review

Abstract

Energy efficiency, longevity, and environmental benefits have made light emitting diodes (LEDs) indispensable in modern lighting and display applications. However, degradation mechanisms influenced by thermal stress, electrical overstress, and environmental conditions mean that their reliability remains a significant challenge. Prognostics and Health Management (PHM) has emerged as a promising approach for monitoring and predicting LED failures, enabling predictive maintenance whilst optimizing operational efficiency. This review comprehensively explores PHM methodologies for LEDs, encompassing physics-of-failure (PoF) models, data-driven approaches, and hybrid techniques that integrate both methodologies. While PoF models offer insights into physics-based failure, data-driven methods leverage statistical analysis, machine learning (ML), and deep learning (DL) for predictive analytics. Hybrid PHM frameworks combine these approaches to enhance prediction accuracy and robustness. The integration of Internet of Things (IoT)-enabled real-time monitoring, digital twins, and edge computing has further improved LED PHM capabilities. Despite these advances, challenges persist in sensor placement limitations, variability in LED architecture, data availability issues, and high computational costs. Overcoming these challenges through standardization, the development of adaptive hybrid models, and the application of advanced Artificial Intelligence (AI)-driven analytics will be essential for enabling the widespread adoption of PHM in LED applications across various industrial sectors. This review highlights key advances, current limitations, and future research directions to improve LED reliability and extend operational life through PHM strategies.