摘要
在对超辐射发光二极管(SLD)实施加速退化试验(ADT)并评估其寿命及可靠性的过程中,其性能参数的温度漂移特性影响了评估的精度,为了剔除这种影响,进而提高评估的精度,需要建立相应的温度与性能参数的关系模型来对试验数据进行处理。对SLD的温度建模进行了研究,为了确定一种适用于SLD及ADT数据处理的温度建模方法。通过调研分析讨论了回归方法、人工神经网络、支持向量机等3种温度建模方法;从建模精度、稳定性等方面对上述方法进行了对比分析,并总结了各种温度建模方法的特点;最终确定支持向量机的温度建模方法更适用于开展SLD及ADT的温度建模工作,从而为SLD及相关产品的温度建模和ADT的评估提供了理论依据。
Temperature drift is a characteristic of super-luminescent diode（SLD）, and it is negative to the evaluation accuracy in accelerated degradation testing （ADT）. To eliminate the temperature drift and improve the evaluation accuracy, a temperature model should be established for processing the experimental data. In order to find a feasible temperature modeling method applied to SLD and data processing of ADT, regression method, BP neural network and support vector machine （SVM） were studied. With the comparison of modeling accuracy, stability and so on, the SVM is finally determined to be the most suitable temperature modeling method for SLD and ADT, and this conclusion is also suitable for other electronic products.
出处
《红外与激光工程》
EI
CSCD
北大核心
2011年第10期1904-1909,共6页
Infrared and Laser Engineering
基金
国防预研项目(51319030301)
关键词
超辐射发光二极管
加速退化试验
温度建模
BP神经网络
支持向量机
super-luminescent diode
accelerated degradation testing
temperature drift modeling
BP neural network
support vector machine
