Semiconductor Defect Classification Tool with Iterative Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic defect classification systems for semiconductor devices are inefficient due to the time-consuming setup of training sets and subjective manual classification, which can lead to inaccurate defect classification and high technician fatigue.
Innovation Solution
The system automatically groups defects and presents a representative set to the user for classification, using techniques like natural grouping and the max/min algorithm to optimize manual classification efficiency, and iteratively refines classifications until all defects are accurately classified.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual classification of all defects is performed to create training set, then classification accuracy is improved, but time required and technician fatigue increase significantly
Solution Approach 1:
The patent segments the defect set into multiple groups based on visual characteristics (size, shape, color, texture) and presents only representative defects from each group to the user for classification. This segmentation allows the system to achieve accurate training set creation without requiring manual classification of all defects, thus reducing setup time while maintaining classification accuracy.
Solution Approach 2:
The patent creates a simplified representation (copy) of the defect data by selecting representative defects from each group. These representative defects serve as a copy of the essential classification information needed to train the system, eliminating the need to manually process every individual defect while preserving the accuracy required for effective defect classification.
2Reliability
If all defects are presented for manual classification, then complete training data is obtained, but ease of operation deteriorates due to overwhelming volume
Solution Approach 1:
The patent divides the large defect dataset into manageable groups based on visual characteristics and presents only representative defects from each group to the user. This segmentation transforms the overwhelming task of classifying all defects into a manageable process of reviewing representative samples, significantly improving ease of operation while maintaining training data completeness through the representative sampling approach.
Solution Approach 2:
The patent applies local quality by selecting representative defects that capture the essential characteristics of each defect group. Instead of presenting all defects uniformly, the system identifies and presents the most representative samples from each local group, making the classification process more efficient and easier to operate while ensuring comprehensive training data representation.
3Extent of automation
If conventional image processing techniques are used for automatic classification, then automation level increases, but classification accuracy decreases for complex defects
Solution Approach 1:
The patent implements a feedback mechanism where the user classifies representative defects and this feedback is used to train and refine the automatic classification system. The system continuously learns from user input, adjusting its classification algorithms to improve accuracy. This feedback loop enables high automation levels while maintaining or improving classification accuracy for complex defects through iterative refinement.
Solution Approach 2:
The patent enables the system to self-improve by using user classifications of representative defects to automatically train and update the classification model. The system serves itself by automatically generating training data from user input on representative samples, thereby achieving high automation in both the initial classification and the ongoing model improvement processes.
Data Source
AI summary
Disclosed are methods and apparatus for efficiently setting up and maintaining a defect classification system. In general terms, the setup procedure optionally includes automatically grouping a set of provided defects and presenting a representative set from each defect group to the user for classification. After the initial manual classification of the representative defects, the setup procedure includes an automatic procedure for classifying the non-reviewed or unclassified defects based on the manual class codes from the user-reviewed defects. After the automatic classification operation, the user may also be presented with defects from each class which may require re-classification. In particular embodiments, the user is iteratively presented with defects which have classifications that are suspect, which are near classification boundaries, or have classifications that have a low confidence level until each class is pure or contains a same type of defect classes as assigned by the user.