1.Hubei University of Technology;2.Hubei Taihe Electric Co., Ltd.
The National Natural Science Foundation of China (General Program, Key Program, Major Research Plan)
随着三维集成技术的不断发展，硅通孔(Through Silicon Via，TSV)三维封装因其独特工艺而受到广泛关注，其缺陷的检测问题也不容忽视。为对TSV三维封装内部缺陷进行分类与量化，提出一种针对TSV三维封装内部缺陷的激光主动激励分类与量化方法。以激光为主动热源，激发TSV内部缺陷，通过理论与仿真分析，掌握缺陷特征在主动激励下的温度分布表现规律；构建卷积神经网络对缺陷样本信息进行训练，实现内部缺陷的分类识别与量化。试验表明，该方法能在不损坏样品的前提下有效对内部缺陷进行识别分类及量化，准确率可达95.56%。
With the continuous development of three-dimensional integration technology, through-silicon via three-dimensional packaging has attracted widespread attention due to its unique process. However, the detection of defects in TSV three-dimensional packaging cannot be ignored. In order to classify and quantify internal defects in TSV three-dimensional packaging, a laser-induced active excitation classification and quantification method for TSV internal defects is proposed. By using a laser as an active heat source, the internal defects of TSV are excited, and through theoretical and simulation analysis, the temperature distribution characteristics of defects under active excitation are understood. A convolutional neural network is then established to train the defect sample information, achieving the classification and quantification of internal defects. The experiments demonstrate that this method can effectively identify, classify, and quantify internal defects without damaging the samples, with an accuracy rate of up to 95.56%.