基于时序分解退化轨道的LED产品可靠性置信统计评估  

A Statistical Assessment of Reliability Confidence Evaluation Method Using Degradation Orbit of Time Series Decomposition

作　　者：缪思巧 周子韩 MIAO Siqiao;ZHOU Zihan(Department of Statistics,School of Mathematics,Southwest Jiaotong University,Chengdu 611756,China;National Engineering Laboratory of Integrated Transportation Big Data Application Technology,Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]西南交通大学数学学院统计系,四川成都611756 [2]西南交通大学综合交通大数据应用技术国家工程实验室,四川成都611756

出　　处：《照明工程学报》2023年第3期71-78,共8页China Illuminating Engineering Journal

基　　金：西南交通大学新时代“大思政”育人工作项目(DSZ2019-ZLTS-19)。

摘　　要：针对加速退化试验下LED产品寿命预测与可靠性评估问题,充分考虑性能退化量的自相关性,用时序分解法对LED性能退化轨道进行描述;分别用多项式样条方法和最小二乘法估计参数拟合趋势项与残差项并进行预测,通过距离加权将寿命预测区间转换为完全样本。介绍了对数正态分布型产品小样本数据的可靠性评估方法,利用极大似然法估计分布参数值,再结合加速模型外推正常情况下产品可靠度,最后利用置信推断方法给出区间估计。以某T5型LED为例,预测得到其寿命为58688.56 h,验证了模型的可行性和准确性。Aiming at the problems of life prediction and reliability evaluation of LED products under accelerated degradation test,the autocorrelation of performance degradation was fully considered,and the LED performance degradation trajectory was described by time series decomposition method.Multinomial spline method and least square method were used to estimate the parameters to fit the trend term and residual term and make predictions respectively.The life prediction interval was converted into complete samples by distance weighting.The reliability evaluation method of lognormal distribution product small sample data is introduced.The maximum likelihood method is used to estimate the distribution parameter values,and the product reliability under normal condition is extrapolated by the accelerated model.Finally,the interval estimation is given by the confidence inference method.Taking a T5 LED as an example,its life is predicted to be 58688.56 h,which verifies the feasibility and accuracy of the model.

关 键 词：LED 加速退化 时序分解 退化轨道 多项式样条 可靠性评估 置信推断 

分 类 号：O212.3[理学—概率论与数理统计] TB114.3[理学—数学]