Semiconductor Anomaly Detection via Dynamic Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection methods in semiconductor manufacturing fail to detect small anomalies due to wide normal value ranges and dead times, leading to missed detections during wafer processing.
Innovation Solution
A method where a semiconductor manufacturing apparatus calculates 3 standard deviations of process condition measurements at a predetermined interval and updates the upper and lower limits for anomaly detection in real time, allowing for precise anomaly detection without dead times, using high-resolution pressure measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a wide normal range is set for anomaly detection, then false detections are reduced, but small anomalies cannot be detected
Solution Approach 1:
The patent applies dynamics by making the upper and lower limits for anomaly detection variable rather than fixed. The limits are dynamically adjusted based on the actual process condition data collected during wafer processing, allowing the detection range to adapt to different process states and improving both reliability and sensitivity simultaneously
Solution Approach 2:
The patent changes the parameters used for anomaly detection from predetermined fixed values to dynamically calculated values based on standard deviations of actual process data. This parameter change enables the system to detect small anomalies while maintaining high reliability by adapting to actual process variations
2Reliability
If a dead time is set between steps in a recipe, then false detections during step transitions are reduced, but small anomalies during wafer processing may not be detected
Solution Approach 1:
The patent eliminates dead time by dynamically adjusting the detection sensitivity based on process conditions. Instead of stopping detection during step transitions, the system continuously monitors process data and adjusts the anomaly thresholds in real-time, maintaining both reliability and continuous detection coverage
Solution Approach 2:
The patent changes the detection parameters dynamically during step transitions rather than using fixed dead time periods. The upper and lower limits are recalculated based on real-time process data, allowing continuous anomaly detection without false detections during transitions
Data Source
AI summary
A method performed by a semiconductor manufacturing apparatus includes calculating, by a processor of the semiconductor manufacturing apparatus, 3 standard deviations of process condition measurements obtained at a predetermined interval from log information of processing of substrates that have been correctly processed, calculating at least one of an upper limit and a lower limit for anomaly detection based on the calculated 3 standard deviations, and detecting an anomaly in the processing of the substrates based on the at least one of the upper limit and the lower limit.