Wafer Defect Analysis Using AI Similarity Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for classifying wafer defects in semiconductor manufacturing are time-consuming and lack quick and accurate feedback, particularly when new defect types emerge, as they rely on supervised learning and unsupervised image clustering with re-training.
Innovation Solution
A wafer defect analysis device utilizing an artificial intelligence model to extract feature vectors from image data, perform similarity analysis with a database of reference vectors, and analyze defect types based on manufacturing process data, enabling rapid identification of defects before testing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If supervised learning or unsupervised image clustering methods are used for defect classification, then defect types can be classified based on image labeling, but it takes a lot of time to learn each type of defect and quick feedback cannot be provided
Solution Approach 1:
The system performs preliminary actions by generating reference wafer image data from manufacturing process data before actual defect analysis. Feature vectors are extracted from these reference images and stored in a database in advance, enabling rapid comparison and classification during actual defect detection without requiring time-consuming retraining of classification models
Solution Approach 2:
The system creates copies of reference wafer image data from manufacturing process parameters. Instead of using actual defect images for comparison, the system generates synthetic reference images based on process data, extracts feature vectors from these copies, and stores them for rapid matching against target wafer images, significantly reducing the time required for defect type identification
2Loss of information
If defect types are classified after wafer testing, then comprehensive defect information can be obtained, but quick and accurate feedback on wafer defects cannot be provided
Solution Approach 1:
The system performs preliminary defect analysis using manufacturing process data before wafer testing is completed. By generating reference image data from process parameters and comparing target wafer images against these references in advance, the system provides early defect type identification and feedback without waiting for the completion of time-consuming wafer testing procedures
Solution Approach 2:
The system introduces manufacturing process data as an intermediary between defect detection and classification. Instead of directly comparing target images with reference defect images after testing, the system uses process data to generate intermediate reference images and feature vectors, enabling earlier and faster defect type classification while maintaining accuracy
Data Source
AI summary
A wafer defect analysis device may obtain manufacturing process data related to a manufacturing process of a reference wafer, generate reference wafer image data based on reference information according to a type of a wafer and the manufacturing process data, update a database based on a plurality of reference feature vectors extracted from the reference wafer image data using an artificial intelligence model, perform similarity analysis between a target feature vector extracted from target wafer image data of a target wafer and the plurality of reference feature vectors included in the updated database, and analyze a defect type of the target wafer based on a result of the similarity analysis.