Sensitivity study on the precision of data acquisition for LED life prediction based on the degradation of luminous output

In recent years, LEDs are widely adopted as a new light source in various applications. Among the many advantages of LEDs over traditional light sources, one key advantage is its long lifespan which substantially reduces the overall maintenance cost. Nevertheless, such characteristic causes various difficulties in the prediction of lifespan and reliability assessment. The lifespan of LEDs can be predicted by a proper reliability test. One common method is to monitor the luminous flux maintenance of LEDs operating under different temperatures. The luminous flux is measured at different time intervals. A prediction model is then used to predict the lifespan of LEDs based on the data collected. In order to obtain an accurate prediction, both the luminous flux and the elapsed time are required to be recorded precisely during the reliability test. Since the luminous flux drop in the first few thousand hours is insignificant, it is common to record only a few percent drop throughout the entire test. This drop level is comparable to the measurement uncertainty. The accuracy of the lifespan prediction will be affected in particular when the error of the measurement system is significant. In this paper, a sensitive study of the data acquisition error on lifespan prediction was carried out. The confidence band was very sensitive to the luminous output measurement error, which usually comes from the periodic optical characterization during the reliability test. This error can be eliminated by in-situ monitoring of the illuminance of the samples in the chamber throughout the reliability test.