Semiconductor Exhaust Pressure Estimation for Abnormality Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing monitoring methods for semiconductor manufacturing equipment have low accuracy in detecting abnormalities in the exhaust system, primarily due to reliance on gas temperature measurements.
Innovation Solution
A program that acquires process data including chamber pressure, valve opening degree, and gas flow rate, inputs this data into a learning model to estimate pressure values, and determines abnormal states by comparing estimated and measured pressure values, creating time-series data to predict when differences exceed predetermined thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If gas temperature measurement is used to determine abnormality, then the monitoring method is simple, but the accuracy of determining exhaust system abnormality is low
Solution Approach 1:
The patent changes the monitored parameter from gas temperature to chamber pressure, which provides more direct and accurate information about exhaust system abnormalities. Pressure changes in the chamber directly reflect issues with the exhaust system, enabling more precise abnormality detection compared to temperature-based methods.
Solution Approach 2:
The patent introduces a learning model as an intermediary that processes multiple process parameters (chamber pressure, valve opening degree, gas flow rate) to estimate the pressure at the exhaust system inlet. This intermediary system synthesizes information from multiple sources to achieve high accuracy in abnormality detection while managing system complexity through intelligent processing.
2Measurement precision
If multiple process parameters are monitored, then the accuracy of abnormality detection is improved, but the complexity of data processing increases
Solution Approach 1:
The learning model serves as an intermediary that automatically processes multiple process parameters (chamber pressure, valve opening degree, gas flow rate) to estimate exhaust system inlet pressure. This intermediary handles the complexity of multi-parameter analysis, providing accurate abnormality detection without requiring complex manual data processing systems.
Solution Approach 2:
The system continuously monitors the difference between estimated and actual chamber pressure, creating a feedback mechanism that improves detection accuracy over time. This feedback loop allows the system to adapt and refine its abnormality detection capabilities while managing data processing complexity through iterative learning.
Data Source
AI summary
A program according to one embodiment of the present disclosure causes a computer to execute processing of: acquiring process data including at least two of chamber pressure, a valve opening degree of an automatic pressure control device and a kind and a flow rate of gas supplied to a semiconductor manufacturing equipment; inputting the process data to a learning model to output an estimated pressure value, the learning model being trained so as to output an estimated pressure value within the semiconductor manufacturing equipment if process data is input; and determining whether or not a state is an abnormal state based on the estimated pressure value.