Probabilistic Approach for Temperature Driven Fatigue Lifetime Data Analysis to Improve Prognostics and Health Management of LED Packages

Prognostics and health management (PHM) is crucial in the lifecycle monitoring of LED packages, especially for applications in a harsh environment. In this context, optimal interpretation of PHM results means translating PHM information into maintenance policies and decisions that minimize life cycle costs, or maximize availability. The fundamental reliability analysis has to take into account all contributors of system failure and also uncertainty in the system behaviour. In this work, system knowledge and measurement variabilities in the form of probability distributions were analysed and their impact on temperature driven prediction models accuracy using the temperature sensitive electrical parameter measurements (TSEP) of an LED package are investigated and discussed. In addition, a Bayesian regression modelling approach was used to quantify the (epistemic) parameter uncertainty in temperature driven model estimations. The investigated uncertainty quantification considers the TSEP measurement, the system boundary conditions and variabilities in temperature calibration on the one hand, and the (epistemic) uncertainty in model parameters on the other hand.