Color-Based Lifetime Estimation of LEDs Using Spectral Power Distribution Prediction Through Analytical and Machine Learning Models

LEDs have seen widespread adoption across various applications in recent years. However, reliability continues to be a major concern. Conventional methods estimate LED lifetime using lumen data and an exponential model, which are suitable for lumen-based evaluations but inadequate for applications where color stability is equally critical. To address this limitation, this paper presents a reliability analysis of LEDs by evaluating performance metrics such as chromaticity shifts Δ<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">u</i>′<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">v</i>′, correlated color temperature (CCT), and chromaticity coordinates (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">u</i>′ and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">v</i>′) through spectral power distribution (SPD) prediction. The study also compares chromaticity coordinate projections with the TM-35-19 projection method and explores lifetime predictions using CCT binning, as outlined in ANSI C78.377, to assess long-term stability. Both analytical and machine learning (ML) models are employed for SPD prediction, with the support vector machine demonstrating superior performance. The results show that both models reliably estimate LED lifetime by predicting Δ<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">u</i>′<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">v</i>′ values, with errors remaining below 10%. The models offer high predictive accuracy, with minimal deviations in CCT and chromaticity coordinates, validating their effectiveness in reliability assessments with accepted methods widely. This work highlights the potential of ML-based approaches to enhance LED performance prediction through colorimetric parameters, and reduce testing time.