Semiconductor Tool Abnormality Detection via Statistical Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor manufacturing, identifying abnormal equipment contributing to decreased yield rates is challenging due to the vast amount of manufacturing history data, which is typically analyzed manually without systematic analysis of association between manufacturing history and inspection data.
Innovation Solution
A method involving statistical and correlation analysis of manufacturing history and measurement data to determine abnormal equipment, considering tool stack run count, and ranking tools based on performance indicators such as bad and good ratios and correlation coefficients to accurately identify problematic tools.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual analysis of manufacturing history data is used to identify problematic tools, then the analysis process is simple to implement, but the efficiency and accuracy of identifying yield-decreasing tools deteriorates due to the huge amount of data
Solution Approach 1:
The patent replaces manual mechanical analysis with an automated computer-based system that performs statistical and correlation analysis on manufacturing history data. The system automatically processes huge datasets, calculates tool performance parameters, and identifies problematic tools, thereby maintaining ease of implementation while dramatically improving identification efficiency and accuracy.
Solution Approach 2:
The patent introduces an intermediary computer-based analysis system that acts as a mediator between the raw manufacturing history data and the identification of problematic tools. This intermediary system performs systematic statistical analysis, calculates correlation coefficients, and generates ranked lists of tools, bridging the gap between data collection and actionable insights.
2Ease of operation
If manual determination of problematic tools is performed directly based on production data, then the method is straightforward, but the accuracy of identifying the actual cause of yield rate decrease deteriorates when data is not analyzed effectively
Solution Approach 1:
The patent replaces manual determination with an automated computer-based system that performs systematic statistical and correlation analysis. The system calculates multiple parameters including bad ratios, good ratios, and correlation coefficients for each tool, then ranks tools based on these parameters to accurately identify the actual cause of yield rate decreases, significantly improving identification accuracy while maintaining operational simplicity.
Solution Approach 2:
The patent implements a feedback mechanism where the system continuously analyzes manufacturing history data, calculates tool performance parameters, and provides ranked lists of problematic tools. This feedback loop enables systematic refinement of identification accuracy by incorporating multiple statistical parameters and correlation analysis, allowing operators to make informed decisions based on objective data-driven insights.
3Device complexity
If traditional analysis methods are used without considering tool stack run count, then the analysis process is simpler, but the ability to detect cumulative performance degradation deteriorates
Solution Approach 1:
The patent segments the analysis by dividing tools into different stacks and further segmenting each stack by run count. This segmentation allows the system to analyze performance degradation at different stages of tool usage, identifying cumulative effects that traditional methods would miss. The segmented approach maintains manageable analysis complexity while significantly improving detection accuracy for progressive tool degradation.
Solution Approach 2:
The patent introduces dynamic analysis by incorporating tool stack run count into the evaluation framework. Instead of treating all tool usage equally, the system dynamically adjusts analysis based on the number of times each tool has been used, allowing detection of performance degradation patterns that evolve over time. This dynamic approach improves reliability in detecting cumulative degradation while keeping analysis complexity manageable through systematic parameter calculation.
Data Source
AI summary
A method for determining abnormal equipment in semiconductor manufacturing system includes processing wafers. A measurement data relating to wafers at respective processing steps and at each tool stack run count for respective tools is provided. The method also includes performing statistical and correlation analysis on the production history data and the measurement data to determine multiple parameters including bad ratio (Rb) and good ratio (Rg) for each tool. A first bad-to-good probability ratio (R1) for each tool is obtained by dividing Rb by Rg at the tool stack run count. A second bad-to-good probability ratio (R2) of each tool is an overall probability ratio of Rb to Rg of each tool. A first correlation coefficient (C1) is provided for the measurement data corresponding to the tool stack run count. A second correlation coefficient (C2) is provided for the first bad-to-good probability ratio (R1) corresponding to the tool stack run count.