Fault Detection Matching Using Sigma and Mean Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semiconductor manufacturing, fault detection and classification between different processing tools relies heavily on user knowledge and experience, lacking a standardized criterion to identify abnormal situations and processing deviations effectively.
Innovation Solution
A fault detection and classification system that utilizes a host device to analyze raw data from chambers, calculating sigma values and mean values to identify target chambers with fault operations by executing sigma matching and mean matching processes, and implementing noise-filtering to increase accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user knowledge and experience are used to determine abnormal situations on FDC charts, then fault detection can be performed, but the detection accuracy and consistency are insufficient due to lack of standardized criteria
Solution Approach 1:
The patent transforms fault detection from subjective user judgment to objective parameter-based analysis by calculating statistical parameters (mean, standard deviation, sigma values) from process data. This parameter transformation enables standardized comparison across different tools and identifies abnormal situations through quantitative thresholds rather than user experience alone.
Solution Approach 2:
The patent replaces the mechanical system of human expert judgment with an automated computational system that calculates sigma values and performs statistical analysis. This substitution eliminates variability in human judgment while maintaining the ability to detect faults through systematic mathematical processing of process parameters.
2Reliability
If standardized criteria are implemented for fault detection, then detection accuracy improves, but the system complexity increases due to multiple matching processes
Solution Approach 1:
The patent segments the fault detection process into distinct modular steps: data collection, sigma value calculation, mean matching process, and sigma matching process. Each module performs a specific function and can be independently implemented or adjusted, reducing overall system complexity while maintaining high reliability through systematic progression through each detection stage.
Solution Approach 2:
The patent introduces sigma values as an intermediary parameter that bridges raw process data and fault detection decisions. These sigma values serve as a standardized mediator that enables comparison across different tools and conditions, facilitating reliable fault identification without requiring direct complex comparisons of raw multi-dimensional process data.
3Measurement precision
If noise-filtering is implemented to increase detection accuracy, then false positives are reduced, but processing time increases
Solution Approach 1:
The patent applies noise-filtering selectively based on the calculated sigma ratio, rather than applying it uniformly to all data. When the sigma ratio indicates low noise levels, the filtering process is skipped or reduced, maintaining fast processing. Filtering is applied more intensively only when the sigma ratio indicates high noise levels that would compromise detection accuracy, thus balancing precision and processing time.
Data Source
AI summary
Raw data from chambers is received. Based on received raw data, if a fault exists in operations of the chambers is detected. The detecting includes at least one of operations outlined below. Sigma values respectively corresponding to the chambers are generated based on the raw data of the chambers. A determination is made to determine whether a sigma ratio corresponding to the sigma values is smaller than a threshold value. Mean outlier indexes respectively corresponding to the chambers is generated by executing a mean matching process for the chambers in a condition that the sigma ratio is smaller than the threshold value. One of the chambers, which has a worst first mean outlier index of the first mean outlier indexes, is identified as a target chamber having fault operation.