Wafer Fault Analysis via Multi-Parameter Cluster Signatures
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for automated defect detection on wafers are limited in identifying consistent anomalies across multiple wafers without requiring labels and are not adaptable to different manufacturing technologies or products, making it difficult to perform comprehensive fault analysis in real-time and historical context.
Innovation Solution
A method involving multiple iterations of cluster analysis using varying parameters on wafer maps, followed by dimensional reduction and clustering, to identify distinct signatures of anomalies, which can be used for fault analysis independently of manufacturing technology and product type, enabling real-time and historical issue detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple cluster analyses with different parameters are performed to identify consistent anomalies across wafers, then the accuracy and completeness of fault detection is improved, but the computational complexity and analysis time increase
Solution Approach 1:
The analysis process is segmented into multiple independent cluster analysis executions, each with different parameters. This allows the complex task of identifying consistent anomalies to be broken down into manageable segments that can be processed separately and then combined, improving both accuracy and computational efficiency
Solution Approach 2:
The method employs periodic execution of cluster analyses with varying parameters rather than a single continuous analysis. This periodic approach with different parameter sets enables comprehensive fault detection while allowing computational resources to be managed in discrete cycles, balancing accuracy with computational complexity
2Reliability
If multiple cluster analyses with different parameters are performed to identify consistent anomalies across wafers, then the completeness of signature identification is improved, but the processing time and computational resources increase
Solution Approach 1:
The method performs preliminary cluster analyses with different parameters to identify potential signatures before finalizing the fault analysis. This preliminary action with varied parameters ensures comprehensive signature identification while allowing the system to prioritize and filter results, reducing the time needed for complete analysis
Solution Approach 2:
The approach systematically changes parameters across multiple cluster analysis executions to identify consistent anomalies from different perspectives. By varying parameters and comparing results, the method achieves complete signature identification while the structured parameter changes allow for efficient resource allocation and time management
3Adaptability or versatility
If automated defect detection is performed without labels to find similar signatures, then the adaptability to different manufacturing technologies is improved, but the precision of defect classification decreases
Solution Approach 1:
The cluster analysis-based signature identification method is designed to be universal across different manufacturing technologies and product types. By using unsupervised clustering without technology-specific labels, the system achieves broad adaptability while the multiple parameter executions and signature consolidation mechanisms maintain sufficient classification precision for fault analysis across diverse manufacturing contexts
Data Source
AI summary
A method for fault analysis in wafers includes determining multiple wafer maps comprising indications of anomalies of the wafers, performing an evaluation based on the determined wafer maps, and performing the fault analysis based on the evaluation performed. Performing the evaluation includes multiple execution of a cluster analysis based on the determined wafer maps using different parameters, and identifying distinct clusters determined by the differently parameterized cluster analyses.