Fault Pattern Recognition via Angle Spectrum Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional fault monitoring methods in semiconductor manufacturing equipment face challenges in accurately defining fault patterns, leading to incorrect fault identification and potential damage to devices, due to reliance on statistical distance and difficulty in determining data inclusion within predefined fault regions.
Innovation Solution
The method involves performing angle spectrum analysis to re-classify fault points on a plane with component axes, calculating angles for each fault point, and forming reference fault patterns to determine matches with standard information, allowing for accurate fault pattern recognition and transmission of treatment information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If statistical distance method is used to determine fault patterns, then fault regions can be predefined and stored in database, but accurate identification of fault patterns becomes difficult when data falls within boundaries or near regions
Solution Approach 1:
The patent segments the fault space by calculating angles between fault data vectors and reference vectors, dividing the continuous statistical space into discrete angular sectors. Each sector corresponds to a specific fault type, enabling clear classification without boundary ambiguity. This segmentation transforms the continuous statistical distance problem into discrete angular region classification.
Solution Approach 2:
The patent introduces an angular dimension to the traditional statistical distance approach. Instead of relying solely on Euclidean distance in the original feature space, the method projects fault data onto angular coordinates relative to reference vectors, adding a directional dimension that resolves boundary ambiguity and improves classification accuracy.
2Productivity
If new fault patterns are determined based on statistical distance, then fault data can be classified into existing regions, but transition time is required for determining new fault patterns and equipment must be taken offline
Solution Approach 1:
The patent performs preliminary angle spectrum analysis during equipment operation to identify emerging fault patterns in real-time. By continuously monitoring angular distributions and comparing them against the database, the system can detect new fault types without stopping equipment, eliminating the transition time required in conventional methods.
Solution Approach 2:
The system implements feedback by continuously comparing real-time fault data angular spectra with the database, automatically updating fault pattern classifications. This closed-loop feedback mechanism enables dynamic adaptation to new fault patterns while equipment remains operational, eliminating offline transition periods.
3Device complexity
If fault patterns are defined using statistical distance and proximity, then fault regions can be established on a plane, but incorrect fault identification may occur leading to wrong corrections and potential device damage
Solution Approach 1:
The patent changes the classification parameter from statistical distance to angular spectrum distribution. By measuring the angle between fault data vectors and reference vectors instead of Euclidean distance, the method achieves more reliable fault identification. This parameter change maintains simplicity in region definition while dramatically improving identification accuracy and reliability.
Data Source
AI summary
Provided is a method of forming reference information for defining a fault pattern of equipment, and monitoring equipment. One example embodiment method may include performing an angle spectrum analysis by re-classifying fault points distributed on a plane, the plane including a first component axis and a second component axis, and the re-classifying fault points including calculating an angle for each of the fault points with reference to any one of the first component axis and the second component axis of the plane, and forming a reference fault pattern for defining a fault pattern of the re-classified fault points.