晶圆缺陷分析模型的训练方法及晶圆缺陷检测方法
By training a multimodal large model that combines overall wafer information, the problems of low accuracy and efficiency in wafer defect detection and analysis are solved, achieving more efficient wafer production quality control.
CN118967600BActive Publication Date: 2026-07-17SHENZHEN ZHIXIAN FUTURE IND SOFTWARE CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN ZHIXIAN FUTURE IND SOFTWARE CO LTD
- Filing Date
- 2024-07-24
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing wafer defect detection and analysis methods cannot accurately classify and analyze their causes, resulting in low production efficiency and quality.
Method used
By acquiring a training dataset, including wafer defect images, structured data, and descriptive text, a multimodal large model is trained to generate a wafer defect analysis model, which is then combined with overall wafer information to analyze defect types and causes.
Benefits of technology
It improves the accuracy of wafer defect type detection, reduces analysis costs, and enhances production efficiency and quality.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN118967600B_ABST
Abstract
本说明书提供一种晶圆缺陷分析模型的训练方法及晶圆缺陷检测方法，获取训练数据集，训练数据集包括晶圆局部的晶圆缺陷图像、用于表征晶圆缺陷和晶圆整体之间关系的结构化数据，用于描述晶圆缺陷图像中的信息的描述文本，以及用于表征图像中晶圆缺陷的类型的标签。通过上述训练数据集，训练多模态大模型，得到晶圆缺陷分析模型。并通过该晶圆缺陷分析模型，进行晶圆缺陷检测。通过上述方法进行晶圆缺陷检测，输出了描述文本，通过描述文本可以获知当前生产过程中存在的缺陷，可以提高半导体的生产质量和效率。通过输入结构化数据，避免晶圆缺陷分析模型仅依靠晶圆缺陷图像在多个晶圆缺陷类型之间摇摆不定，可以提升晶圆缺陷类型检测的准确率。
Need to check novelty before this filing date? Find Prior Art