An analytical and machine learning model for SPD estimation and its prediction for lumen and chromaticity shift based LED lifetime performance analysis

• Published in: Physica scripta 2024-07, Vol.99 (7), p.76018
• Main Authors: Lokesh, J, Padmasali, AN, Mahesha, MG, Kini, S G
• Format: Article
• Language: English
• Subjects: lifetime estimation; light emitting diodes; machine learning
• ISSN: 0031-8949; 1402-4896
• DOI: 10.1088/1402-4896/ad55e7

The LED lifetime is commonly estimated by manufacturers using an exponential model to evaluate L70 criteria. However, it ignores colour characteristic variation and does not explain the root cause of LED failure. In this paper, a spectral power distribution (SPD) based approach is proposed to estimate lifetime performance of cool white LED considering both colour characteristics and lumen maintenance, as all the lighting performance parameters are extracted from SPD. The exponential model does provide the lifetime using only lumen data and does not explain the colour characteristics. As an alternate to the exponential model, a quadratic polynomial, and machine learning (ML) models with hours and temperature as input factors, is proposed to determine SPD for experimental conditions as well as to predict for other operating conditions. Further, the lifetime performance analysis is performed for reliability assessment conditions through both lumen and colorimetric performances. The outcomes of all the models are analysed and it is found that the results are comparable. As ML models are simpler than analytical models for more than two inputs, further it is used to predict SPD at different temperatures and the LED performance is validated. Further analysis shows that a decrease in blue light is the primary cause of the overall decrease in light output and decrease in yellow emission due to phosphor degradation is the reason for chromaticity shift.