Automatic Defect Classification Knowledge Base Learning

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Manual defect classification in semiconductor manufacturing is time-consuming and inefficient, hindering the quality and precision of defect recognition in the production process.

Innovation Solution

A method for learning a knowledge-based database for automatic defect classification, where user-input parameters and data are used to optimize light intensity and detection parameters, allowing for automatic grouping and classification of defects, reducing the need for manual classification and improving the quality of the knowledge base.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual classification of defects is used, then the knowledge base can be created with detailed defect information, but the time required to create the ADC protocol becomes extremely long

Engineering Contradiction:

Improvedefect classification precisionVSAvoidtime to create ADC protocol

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs automatic defect classification using an ADC run that automatically approaches defects, captures images, extracts descriptors, and groups defects into classes without requiring manual intervention. The knowledge base is self-generated through automated processing of wafer images and defect data.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The manual mechanical process of defect classification is replaced with an automated electronic system that uses image processing, descriptor extraction, and algorithmic grouping to classify defects. The ADC system substitutes human operators with automated software that processes defect images and creates classification categories.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Loss of information

If manual classification steps are performed in sequence, then comprehensive defect data can be collected, but the number of steps and complexity of the process increases

Engineering Contradiction:

Improvedefect data completenessVSAvoidprocess steps complexity

Core Design Contradiction:

Loss of informationVSDevice complexity

Solution Approach 1:

Multiple separate manual operations (approaching defects, capturing images, extracting features, classifying defects) are merged into a single integrated ADC run that executes automatically. The system combines image acquisition, descriptor extraction, and defect grouping into one unified automated process that reduces procedural complexity.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The system performs preliminary automatic classification by running ADC on sample defects to generate initial defect groups and descriptors before final knowledge base creation. This preliminary automated processing prepares the data structure and classification framework, reducing the complexity of subsequent finalization steps.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS7623698B2Method of learning a knowledge-based database used in automatic defect classification

Publication Date: 2009.11.24 KLA TENCOR MIE

AI summary

The invention relates to a method of learning a knowledge-based database used in automatic defect classification. According to this method, the user is spared a series of entries as the system carries out an automatic learn mode, which requires a reduced number of user entries.