Life Prediction Modeling and Accuracy Verification of Photoelectric Products: A Combination of PSO-SVM and Accelerated Luminance Attenuation

Intelligence algorithms used to predict the conventional life of photoelectric products significantly reduce the time and cost of product life test, which is a popular life evaluation method in the photoelectric industry. However, the accuracy of prediction results remains a challenge for most existing models. In this study, three innovative high-precision life prediction models for photoelectric products were first proposed based on the particle swarm optimization-support vector machine (PSO-SVM) algorithm combining ALA. Then for the sake of validating the accuracy of these models, luminance attenuation data of a certain plant lighting LED lamp were collected through our self-developed accelerated life test system, and input into PSO-SVM as training data after being processed according to a self-designed data preprocessing scheme. Eventually, three models were respectively applied to predict the conventional life of the plant lighting LED lamp. The results demonstrate that the PSO-SVM life prediction models proposed in this work exhibit a high accuracy in predicting conventional life of photoelectric products, with the minimum relative error of only 0.74% compared with the referred conventional life value. The above-mentioned achievements can provide a new approach for evaluating modern photoelectric product life.