| 注册
首页|期刊导航|安徽大学学报(自然科学版)|资源受限环境下TSV芯片内部缺陷检测技术研究

资源受限环境下TSV芯片内部缺陷检测技术研究

叶方平 欧阳泉 张鸣 于晨睿 黄一凡 宋梓贤 聂磊

安徽大学学报(自然科学版)2026,Vol.50Issue(5):32-39,8.
✕
安徽大学学报(自然科学版)2026,Vol.50Issue(5):32-39,8.DOI:10.3969/j.issn.1000-2162.2026.05.005

资源受限环境下TSV芯片内部缺陷检测技术研究

Research on an internal defect detection technique for TSV chips in resource-constrained environments

叶方平 1欧阳泉 1张鸣 1于晨睿 1黄一凡 2宋梓贤 2聂磊1

作者信息

  • 1. 湖北工业大学机械工程学院,湖北武汉 430068
  • 2. 华中科技大学机械科学与工程学院,湖北武汉 430074
  • 折叠

摘要

Abstract

Through-silicon via(TSV)is an important interconnection technology in 3D packaging,which achieves high space utilization and improves packaging density through vertical stacking of chips.However,the TSV 3D packaging process is relatively complex,which may generate internal defects that are difficult to detect directly.To address this issue,a method is proposed to detect internal defects in packaging by applying external excitation with infrared laser and identifying the characteristics of external temperature distribution field of internal defects through a neural network model.At the same time,considering detection efficiency and integration with the production line,the model is optimized for lightweight to meet the detection requirements in environments with limited computing resources.Experiments show that the lightweight model proposed in this paper achieves a classification accuracy of 95.06%,a localization accuracy of 97.52%,a classification speed of 1.12 ms per frame,and a localization speed of 1.58 ms per frame for typical internal defects in TSV 3D packaging.The method proposed in this paper offers a valuable new perspective for the practical application of 3D packaging defect detection technology in the production line.

关键词

硅通孔/红外热成像/轻量化模型/资源受限环境/内部缺陷检测

Key words

through-silicon via/infrared thermography/lightweight model/resource-constrained environments/internal defect detection

分类

信息技术与安全科学

引用本文复制引用

叶方平,欧阳泉,张鸣,于晨睿,黄一凡,宋梓贤,聂磊..资源受限环境下TSV芯片内部缺陷检测技术研究[J].安徽大学学报(自然科学版),2026,50(5):32-39,8.

基金项目

湖北省国际科技合作项目(2024EHA007) (2024EHA007)

安徽大学学报(自然科学版)

1000-2162

访问量0
|
下载量0
段落导航相关论文