Prognostics and reliability assessment of light emitting diode packaging

Compared to the traditional light sources, applications of LED lights are continuously increasing as they have many advantages including high reliability, greater energy efficiency, long life time, small in size, and faster switching speed. Even though LEDs have high reliability and long life time, manufacturers and lighting systems designers still need to assess the reliability of LED lighting systems and the failures in the LED. This is very important with respect to the maintainability of the LED lighting systems after the deployment. In particular assessing the reliability of the LED lighting systems which are used for safety critical and emergency applications is a requirement to ensure the light output meets the standards all the time. This paper investigates distance measure techniques for a real-time prognostics system which can assess the reliability, predict the failure in advance and estimate remaining useful life time of a LED packaging. In particular Euclidean and Mahalanobis distance measure techniques are used to measure the degradation in the light output of the LEDs and used to predict the failure in advance. LEDs are tested under accelerated voltage conditions to collect the training and test data sets. Results show that both distance measure techniques predict the degradation in the light output. Simple linear state extrapolation technique is used to estimate the remaining useful life time of the LEDs, based on these two distance measure techniques.