Lifetime prediction of a multi-chip high-power LED light source based on artificial neural networks

Abstract

A high power light-emitting diodes (LED) light source, which is assembled by many LED chips and heat sink, has a not uniform temperature distribution at working condition. Analysis and prediction of a high-power LED system lifetime are complicated and time-consuming. In this study, two artificial neural networks (ANN) are employed to simplify the LED lighting system’s lifetime prediction and increase the precision of such analysis. The temperature distribution of the high power light source is calculated by Finite element method (FEM) with LED chip photo-electro-thermal (PET) ANN. With the precise LED temperature distribution, the lifetime for multi-chip high-power LED light source can be predicted by lifetime ANN. The PET ANN is used to simplify the multi-physics coupling in LED PET analyzing. The lifetime ANN can predict the lifetime of a LED chip across a wide range of temperatures and not merely corresponding to several testing and interpolation points. This work aims to rapidly analyze multi-chip high-power lighting systems characteristics. The repeated tests and calculations of PET and lifetime are avoided with this method. This approach presents a new and effective way to assess the reliability of high-power LED light source.