摘要
提出了一种新型的基于遗传算法(GA)优化的误差反向传播(BP)神经网络的寿命预测模型。选取不同公司生产的LED,以LED光源光通量维持率测量方法 (LM-80-08)测试报告中的电流、结温、初始光通量和初始色坐标作为神经网络的输入,LED在网络输入的应力条件下的寿命为输出,可以预测LED在任意电流和结温下的寿命。研究结果表明,该GA-BP模型相比于LED光源长期流明维持率的预测方法 (TM-21-11)更具灵活性,预测误差较传统BP神经网络降低了65.5%,平均相对误差达到1.47%,优于Adaboost模型的54%和3.16%,训练样本相关系数达到99.4%,GA-BP模型预测LED寿命误差更小,普适性更高,在LED的寿命预测中具有实际意义。
A new lifetime prediction model based on the error back propagation( BP) neural network optimized by genetic algorithm( GA) was presented. A variety of LEDs produced by different companies were selected for test. The parameters which were obtained by test reports( LM-80-08) of light flux maintenance rate of LEDs light sources were served as the inputs of the neural network,such as electric current,junction temperature,initial luminous flux and initial chromaticity coordinates,while the lifetime of LEDs under the stress of network's input was served as the output of the neural network. The lifetime of LEDs could be predicted under any current and junction temperature. The research results show that this GA-BP model is more flexible than the prediction method( TM-21-11) of long term lumen maintenance rate of LED light sources. Compared with the traditional BP neural network,the prediction error of GA-BP model is reduced by 65. 5%,the average relative error is 1. 47%,which is better than the Adaboost model of 54% and 3. 16%. The correlation coefficient of training samples reaches 99. 4%. The GA-BP model is more accurate and adaptable in the prediction lifetime of LEDs,which is of practical significance in the prediction lifetime of LEDs.
作者
吴志杰
孔凡敏
李康
Wu Zhijie;Kong Fanmin;Li Kang(School of Information Science and Engineering, Shandong University, Jinan 250100, China)
出处
《半导体技术》
CAS
CSCD
北大核心
2018年第5期375-380,共6页
Semiconductor Technology
基金
国家高技术研究发展计划(863计划)资助项目(2015AA03A102)
国家自然科学基金资助项目(61475084)