基于贝叶斯概率统计的发光二极管瞬态热阻分析  

Analysis of Transient Thermal Resistance of LED Based on Bayesian Probability and Statistics

作　　者：陈伟[1,2] 阙秀福 冯伟[1,2] 张建华[1] 杨连乔[1] 

机构地区：[1]上海大学新型显示技术及应用集成教育部重点实验室,上海200072 [2]上海大学材料科学与工程学院,上海200072

出　　处：《上海交通大学学报》2016年第4期636-640,共5页Journal of Shanghai Jiaotong University

基　　金：国家重点基础研究发展规划(973)项目(2011CB013103)资助

摘　　要：应用贝叶斯概率统计的方法来反卷积进行网络识别,分析发光二极管(LED)封装器件的瞬态热阻,有效避免了计算机中频域法解卷积出现的病态问题以及贝叶斯迭代法中迭代次数选择的问题.讨论了该方法中优化参数与时间常数谱的半波宽以及纹波的关系,选择适当的优化参数进行反卷积,再由Foster网络模型转换到Cauer网络模型,获得LED器件的瞬态热阻.同时,讨论了该方法对含噪信号的反卷积的可靠性,并分析了采用此种反卷积方法得到的LED封装器件的热阻的正确性.Based on the theoretical basis of transient thermal test,back-stepping algorithm from the result of transient thermal measurement to structure function generally include data fitting,smoothing,derivation,deconvolution and network structure conversion.The Bayesian probability and statistics method was presented for the network identification by deconvolution（NID）in analyzing transient thermal resistance of light emitting diode（LED）,which effectively avoided the ill posed problems in the frequency domain method of the computer and the choice of the iteration number in the Bayesian iterative method.The relationship between the optimization parameterεand half-wave width,and its relationship with ripple of time constant spectrum were discussed.The proposed method successfully extracted the transient thermal resistance in the transformation from Foster network model to Cauer network model with an optimizedε.Meanwhile,the reliability of this method in deconvolution of noisy signal and the accuracy in obtaining the thermal resistance of LED package devices were discussed.

关 键 词：贝叶斯 反卷积网络识别 发光二极管 瞬态热阻 时间常数谱 

分 类 号：O212.8[理学—概率论与数理统计] TN312[理学—数学]