An Adaptive GA-SVM Life Prediction Model for Solid-State Light-Emitting Devices Under the Failure Scenario of Accelerated Life Test: A Case Study of an OLED Product

The practice of establishing a life prediction model from accelerated life test (ALT) data to assess products' reliability is becoming increasingly popular in industries producing long-life products such as solid-state light-emitting devices (solid-state LEDs). Nevertheless, if ALT fails and results in data missing, retesting will cost a lot of time and affect the production cycle. To address this issue, a life prediction model based on support vector machine (SVM) improved by adaptive genetic algorithm (GA) was proposed in this study. First, adaptive GA-SVM was used for the restoration of failure ALT data, and then the conventional life of the product could be predicted according to the restored data combined with probabilistic statistical methods. Furthermore, in order to verify the prediction performance of the model, the failure ALT of an organic light-emitting diode (OLED) product was taken as an example. The results demonstrate that the data restoration effect of the model is satisfactory, and the relative error between the life predicted according to the restored data and the actual life of the product is merely 2.88%, which proves the effectiveness of the model proposed. Meanwhile, the achievement provides a good remedy for the special situation of failed ALT.