Defect Cell Clustering Using Equipment Coherence
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional defect clustering methods fail to accurately group defects based on their causes, as they primarily rely on spatial proximity rather than the equipment involved, leading to inefficient analysis and stabilization in product and process optimization.
Innovation Solution
A method that generates a sample defect map by cell positions, selects high bad ratio cell positions, and groups them into clusters based on position coherence and suspected bad equipment information, allowing for non-adjacent defects to be clustered together, specifically tailored for small-sized display manufacturing processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If defects are clustered based on spatial proximity only, then the clustering process is simple and fast, but the clustering accuracy is low because defects from different causes are grouped together
Solution Approach 1:
The patent merges spatial proximity information with equipment information to form clusters. Defects are grouped based on both their physical location and the equipment that produced them, creating more accurate clusters that reflect true defect causes while maintaining computational feasibility through the integration of multiple data dimensions.
Solution Approach 2:
The patent changes the clustering parameters from purely spatial coordinates to a composite parameter set including both position and equipment identification. This parameter transformation enables the system to distinguish defects by their operational causes while preserving the simplicity of automated processing through standardized data structures.
2Reliability
If all defects are clustered regardless of equipment information, then the analysis covers all defects, but noise increases and meaningful patterns are obscured
Solution Approach 1:
The patent extracts and separates defects based on equipment information, isolating defect groups that share common equipment origins. This extraction process removes noise from the analysis by grouping defects with similar causes, thereby enhancing the signal-to-noise ratio and improving the reliability of subsequent analysis without losing important defect information.
3Adaptability or versatility
If conventional spatial-based clustering is used, then adjacent defects are grouped together, but non-adjacent defects with common causes are separated into different clusters
Solution Approach 1:
The patent adds equipment information as another dimension to the clustering process. Instead of clustering solely in spatial dimensions, the system incorporates equipment identification as an additional grouping criterion, enabling non-adjacent defects that share the same equipment origin to be clustered together while maintaining spatial awareness.
Data Source
AI summary
Provided are a method of clustering defects generated in bad samples shown on a defect map of bad samples including bad products, and an apparatus thereof. The defect cell clustering method includes generating a sample defect map showing a defect cell distribution by cell positions of bad samples comprised of products each including one or more defect cells among products each partitioned into a plurality of cells, selecting at least some cell positions having one or more defect cells as clustering targets from the sample defect map, selecting one or more suspected bad equipments for each of cell positions included in the clustering targets using pass equipment information for the product, and grouping the clustering targets into one or more clusters according to position coherence between a first cell position and a second cell position included in one cluster, the cell position and the second cell position each having at least one suspected bad equipment.