LED Reliability Assessment Using a Novel Monte Carlo-Based Algorithm

Abstract

Application of Monte Carlo (MC) simulations in the statistical analysis of LED lumen maintenance is presented in this paper. Lumen maintenance data is acquired using experimental tests accomplished in the electro-optics laboratory of the Mazinoor lighting industry, which is an accredited laboratory by Iranian National Standards organization. The sampling rate and the duration of the experiments are consistent with LM-80-15 standard introduced by the Illumination Engineering Society of North America. In some cases, due to the existence of nonlinear dynamics in real trends of light flux, particularly in the first 1,000 hours, features are not completely captured using traditional reliability assessment techniques such as TM-21. In this study, a two-phase model is applied to cover features in lumen maintenance data. Furthermore, to estimate the parameters of the dedicated model in mild and severe operating conditions, a nonlinear Kalman filter-based method known as the iterated extended Kalman filter (IEKF) is used. A set of MC simulations are run to construct the probability density functions (PDFs) for the estimated parameters. Each simulation uses different values of the parameters chosen from the corresponding distribution. Finally, lifetime PDFs are constructed to extract reliability indices. All of the simulations are conducted in MATLAB and the results are compared with the conventional and well-known TM-21 approach.