基于混合极限学习机的TSV缺陷检测技术  

TSV Defect Detection Technology Based on the Hybrid Extreme Learning Machine

作　　者：陈寿宏[1] 康怀强 侯杏娜[1] 尚玉玲[1] Chen Shouhong;Kang Huaiqiang;Hou Xingna;Shang Yuling(School of Electronic Engineering and Automation,Guilin University of Elctronic Technology,Guilin 541004,China)

机构地区：[1]桂林电子科技大学电子工程与自动化学院,广西桂林541004

出　　处：《半导体技术》2020年第7期557-563,共7页Semiconductor Technology

基　　金：国家自然科学基金资助项目(61661013);广西自然科学基金资助项目(2018GXNSFAA281327)。

摘　　要：随着三维集成技术的飞速发展,硅通孔(TSV)缺陷的检测问题不容忽视。提出了一种新型无损TSV缺陷检测方法,该方法采用混合极限学习机模型对TSV缺陷的S参数进行训练分类,用来预测TSV发生空洞缺陷的大小及高度、发生针孔缺陷的大小及高度及发生微衬底未对齐缺陷的偏移量。仿真结果表明,所提出的方法在TSV缺陷检测过程中可以避免对被测样品的损坏,且与原极限学习机相比,其缺陷定位准确率提高了11.51%,达到94.61%。基于混合极限学习机的TSV缺陷检测方法,既可以对不同类型的TSV缺陷进行分类,也能针对具体缺陷类型进行定位。With the rapid development of 3 D integration technology, the problem of through silicon via(TSV) defect detection cannot be ignored. A new TSV defect non-destructive detection method was proposed. A hybrid extreme learning machine model was used to train and classify the S parameters of TSV defects to predict the size and height of the void defect and pinhole defect, and the offset of the micro substrate misalignment. The simulation results show that the proposed method can avoid the damage to the tested samples during the TSV defect detection process. Compared with the original extreme learning machine, the accuracy of the defect location is increased by 11.51% to 94.61%.The TSV defect detection method based on the hybrid extreme learning machine can classify different types of TSV defects, and locate specific defect types.

关 键 词：硅通孔(TSV)缺陷检测 机器学习 S参数 极限学习机(ELM) 粒子群优化(PSO)算法 

分 类 号：TN407[电子电信—微电子学与固体电子学]