Semiconductor Sensor Failure Detection Using Regression Slope
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current failure detection methods in semiconductor manufacturing apparatuses are inadequate in accurately identifying sensor failures, particularly when sensor values exhibit gradual or rapid changes, leading to potential misinterpretation and reduced detection accuracy.
Innovation Solution
A failure detection system that generates time-series data, calculates a regression line, and determines sensor failure based on the slope and change amount of the regression line, using control values set by users to differentiate between normal and abnormal sensor behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional failure detection methods (SPC charts, correlation charts) are used to monitor sensor values, then the system can track sensor readings over time, but the system cannot accurately distinguish between normal gradual changes and actual sensor failures, leading to reduced detection accuracy
Solution Approach 1:
The patent transforms the detection approach by changing from monitoring absolute sensor values to monitoring the rate of change (derivative) of sensor values. By calculating the time derivative of the regression line slope, the system identifies sensor failures through abnormal changes in the rate of change, rather than relying on fixed threshold comparisons of raw sensor data. This parameter transformation enables accurate distinction between normal gradual drift and actual sensor failures.
2Ease of operation
If the system monitors sensor values using fixed threshold methods, then the detection process is simple, but the system produces erroneous detections when sensor values exhibit gradual changes over time
Solution Approach 1:
The patent introduces dynamic analysis by continuously calculating the regression line and its time derivative rather than using static threshold comparisons. The system adapts to gradual sensor changes by modeling the normal rate of change through regression analysis, then detects failures when the actual rate of change deviates significantly from this dynamic model. This dynamic approach maintains operational simplicity while eliminating erroneous detections caused by gradual sensor drift.
3Measurement precision
If the system uses complex analysis methods to improve detection accuracy, then false detections are reduced, but the device complexity increases
Solution Approach 1:
The patent introduces regression line analysis as an intermediary layer between raw sensor data and failure detection. Instead of directly comparing sensor values to thresholds, the system first fits a regression line to the time-series data, then analyzes the time derivative of this regression line's slope. This intermediary approach simplifies the complexity by providing a smooth, continuous model of sensor behavior that is easier to analyze than raw fluctuating data, while maintaining high detection accuracy.
Data Source
AI summary
A failure detection system detects a failure of a sensor that detects a state of a semiconductor manufacturing apparatus. The failure detection system includes a generation unit configured to generate times-series data of information on a detection value of the sensor during a determination period, a calculation unit configured to calculate a regression line of the times-series data, and a failure determination unit configured to determine whether the sensor has failed based on a slope of the regression line.