A comparison study of the prognostics approaches to Light Emitting Diodes under accelerated aging

Light Emitting Diode (LED) lighting systems are being implemented as a future light source in many sectors. They have advantages such as power efficiency, higher reliability, small in size, faster switching speed, etc. Previous research has shown that same types of LEDs, from same manufacturer may have significantly different characteristics and behaviour under similar operating condition. These findings indicate the difficulties in assessing and maintaining the LED lighting systems in the field after their deployment, particularly in the case of safety critical, emergency and harsh environment applications. This paper demonstrates two different prognostics and health management (PHM) approach namely data driven and model driven approach to assess the reliability and predict the remaining useful lifetime (RUL) of LED lighting systems in the field. Focus of this paper is to compare the performance of these two different modelling approaches under thermal and electrical overstress conditions. Results indicate the predictions made by model driven and data driven approach are within the reasonable limit and hence they can be used to predict the catastrophic failures caused by the thermal and electrical overstress in the field. Both techniques gain accuracy as time progresses and make better prediction closer to failure. This paper will also propose a fusion based approach to increase the accuracy of the prediction at the early stage of use. This will provide benefits in terms of planning and maintenance for large scale LED deployment especially in the safety critical applications.