Predicting Luminance Decay of a Micro-LED display via Machine Learnings on Temperature Distribution and LED Degradation

Abstract A new method for predicting the luminance decay of Micro Light Emitting Diode (Micro-LED) displays by machine learning models is proposed herein with experiments of temperature distribution and degradation established. Although Micro-LEDs can be used as a direct light source for large outdoor advertising billboards, harsh outdoor conditions may lead to the degradation of Micro-LED displays. As a result, a temperature model is first built to predict the temperature distribution for the surface of a Micro-LED display based on illuminated patterns and the temperature sensors installed on the back of the display, followed by the establishment of degradation models for predicting luminance decay of the display based on Micro-LED enclosure temperature, input current, and illumination time. Based on the different degradation characteristics observed for red, green, and blue light in the experiments, their degradation models are established separately. In addition, exponential curve-fitting and interpolation are conducted based on TM-21 for the high accuracy when predicting for a much longer aging period. The temperature model built exhibits a prediction error of less than 1.1°C, while an average error is kept below 1.05% (roughly 9 nits). Moreover, the predicted period using the proposed degradation model with high accuracy can reach up to tens of thousands or even hundreds of thousands of hours. It is evident that the predicted results within this long period meets the requirement of the exponential curve defined by TM-21.