Comparative Analysis of Conventional and Machine Learning Models for LED Lifetime Estimation

The LED buyers are more concerned about reliability even though LEDs come in a variety of configurations in market. The manufacturers do not provide the lifetime of an LED under real operating profiles. Currently, the LED manufacturers use conventional (exponential) model to define lifetime at each temperature and current independently. Therefore, the conventional model cannot be used to estimate lifetime of LED under different electrical and temperature stress levels. To address this, the study proposes machine learning (ML) based model to define LED lifespan under any thermal and electrical profiles. The study compares the results of conventional model with proposed model. Temperature and current are incorporated as input elements with hours in the ML models, which is absent in the conventional model, and which just takes an hour as input. According to the results, Support Vector Machine (SVM) model get superior outcomes with APE smaller than 10%. This demonstrates that the suggested ML model can forecast the lifespan of an LED at any temperature and current stress level. The goal of this study is to acknowledge the LM80 report supplied by LED device makers. The research will also assist consumers in determining the lifespan of LEDs for any operational profiles before they utilise them.