Prognostic algorithms for L70 life prediction of solid state lighting

Abstract

The life of light-emitting diodes (LEDs) is difficult to measure by traditional testing methods as they are not likely to fail completely. The Illuminating Engineering Society of North America (IESNA) uses a standard regression approach based on short-term collected lumen data to predict the L70 lifetime of LEDs. In this paper, a model-based prognostics method is employed to determine the life of luminaires using LEDs. Unscented Kalman filter and particle filter algorithms are used for degradation model parameter estimation. An analytical approach based on three statistical models (Weibull, normal, lognormal) is employed and a best fit is determined by the Akaike information criterion. The resulting L70 is compared with L70 derived from the IESNA approach to accurately determine the best prognostic method.