Processing Space Condition Evaluation Using Time-Series Relationships
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In manufacturing processes, particularly in semiconductor manufacturing, it is challenging to accurately evaluate and adjust the conditions of processing spaces to maintain product quality due to the complex interplay of various time series data sets, which existing methods fail to effectively capture and utilize.
Innovation Solution
An analysis device employing machine learning to analyze time series data groups, calculating values indicating relationships between measurement items and evaluating unknown conditions of processing spaces based on known data, allowing for real-time adjustments to maintain optimal processing conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is used to analyze time series data groups to evaluate processing space conditions, then measurement precision and evaluation accuracy are improved, but device complexity and computational requirements increase
Solution Approach 1:
The system segments the complex analysis task into distinct functional units: a learning unit that performs machine learning to extract relationship values from time series data, and an evaluation unit that uses these values to evaluate processing space conditions. This segmentation allows the complex ML functionality to be integrated modularly, improving measurement precision while managing device complexity through structured organization.
2Loss of information
If relationship values between time series data are calculated using machine learning, then information completeness about processing conditions is improved, but loss of time for data processing increases
Solution Approach 1:
The learning unit performs machine learning in advance to extract and store relationship values between different time series data items. These pre-calculated relationship values are then reused by the evaluation unit to quickly assess processing space conditions without重新 performing complex ML computations, thereby reducing data processing time while maintaining information completeness.
Solution Approach 2:
The system creates simplified copies of complex relationships by extracting relationship values that represent the essential connections between time series data items. These copied relationship representations allow the evaluation unit to assess conditions efficiently without processing the full complexity of the original time series data, reducing processing time while preserving critical information.
3Measurement precision
If multiple types of time series data are analyzed simultaneously, then measurement precision of processing conditions is improved, but device complexity increases
Solution Approach 1:
The learning unit is designed with multi-functionality to handle various types of time series data (process data, sensor data, inspection data) through a unified machine learning approach. This universal processing capability allows the system to analyze multiple data types simultaneously with a single integrated unit, improving condition evaluation accuracy while avoiding the need for separate specialized systems for each data type.
Data Source
AI summary
An analysis device includes learning circuitry configured to perform machine learning using a time series data group measured in association with a processing of an object in a processing space and to calculate a value indicating a relationship of time series data in a corresponding time range between respective measurement items and evaluation circuitry configured to evaluate an unknown condition of the processing space based on the value indicating the relationship calculated by performing machine learning by the learning circuitry using a time series data group measured in association with a processing of the object under a known condition of the processing space.