基于红外图像分析的TSV内部缺陷识别方法研究  

TSV Internal Defects Identification Methods Research Based on Infrared Image Analysis

作　　者：聂磊[1] 武丽丽 黄一凡 刘梦然 刘江林 NIE Lei;WU Li-li;HUANG Yi-fan;LIU Meng-ran;LIU Jiang-lin(School of Mechanical Engineering,Hubei University of Technology,Wuhan 430000,China)

机构地区：[1]湖北工业大学机械工程学院,湖北武汉430000

出　　处：《仪表技术与传感器》2023年第1期38-43,共6页Instrument Technique and Sensor

基　　金：国家自然科学基金项目(51975191)。

摘　　要：TSV三维封装内部缺陷难以用传统方法检测。然而其内部缺陷的存在会导致热阻发生变化,对系统温度分布产生影响,因此可以通过对红外图像的分析达到对缺陷进行识别及定位的目的。文中研究了缺陷对温度场的影响,分别通过理论分析、有限元仿真及实验方法对TSV三维封装系统进行了热-电耦合分析,得到了缺陷铜柱类型及位置不同时的温度分布数据集,搭建了卷积神经网络(CNN)模型对2组数据集单独进行分类预测。实验结果表明:利用仿真数据集与试验数据集分别对CNN模型进行特征训练,得到的缺陷识别与定位准确率为98.65%,98.36%。由上可知,缺陷类型及位置的不同会对温度场产生不同影响,利用CNN模型对TSV红外热图像进行特征训练可以有效识别与定位内部缺陷。Internal defects of TSV 3D packaging are difficult to detect by traditional methods.However,the existence of internal defects can lead to changes in thermal resistance and affect the temperature distribution of the system.Therefore,the purpose of identifying and locating defects can be achieved by analyzing infrared thermal images.On the basis of studying the influence of defects on the temperature field,the thermal-electric coupling analysis of the TSV three-dimensional packaging system was carried out through theoretical analysis,finite element simulation and experimental methods,and the temperature distribution data of different types and positions of defect copper pillars were obtained.A convolutional neural network(CNN)model was built to classify and predict the two sets of data sets separately.The experimental results show that using the simulation data set and the experimental data set to train the CNN model separately,the accuracy of defect identification and location were 98.65%and 98.36%.It can be seen from the above that the types and positions of defects can have different effects on the temperature field.Using the CNN model to perform feature training on TSV infrared thermal images can effectively identify and locate internal defects.

关 键 词：硅通孔 三维封装 内部缺陷 红外图像 卷积神经网络 识别与定位 

分 类 号：TN305.94[电子电信—物理电子学]