Substrate State Determination Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing substrate processing technologies face challenges in accurately determining the state of substrates during rotation, particularly in identifying when a substrate is warped or not properly seated, which can lead to inefficient processing and potential apparatus failures.
Innovation Solution
A substrate state determining apparatus that uses machine learning to generate a model from training data, where images of the substrate are input to determine its state, including warpage or proper seating, and decides whether the rotary table can be rotated, incorporating cameras to capture images and an arithmetic processing unit for real-time determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is used to determine substrate state from images, then measurement precision is improved, but device complexity increases
Solution Approach 1:
A camera serves as an intermediary to capture images of the substrate, which are then processed by a machine learning model. This separates the detection function from the determination function, allowing high-precision image capture while using software-based machine learning for complex analysis, thereby improving measurement precision without proportionally increasing overall device complexity
Solution Approach 2:
The patent replaces traditional mechanical or optical measurement systems with a machine learning-based image analysis system. Instead of using complex mechanical sensors or optical interferometry to detect substrate state, the system uses standard imaging combined with machine learning algorithms, achieving high precision while reducing mechanical complexity
2Productivity
If real-time substrate state determination is implemented, then productivity is improved, but use of energy increases
Solution Approach 1:
The system performs substrate state determination at periodic intervals during the substrate processing cycle, rather than continuously. The camera captures images and the machine learning model processes them at key moments (before and after rotation), enabling real-time monitoring and quick response while reducing energy consumption compared to continuous operation
Solution Approach 2:
The machine learning model is trained offline using predetermined training data, so the actual real-time determination during substrate processing requires minimal computational resources. The model serves itself by making rapid inferences from captured images without requiring complex real-time computation, thus improving productivity while maintaining low energy consumption during operation
Data Source
AI summary
According to an aspect of the present disclosure, a substrate state determining apparatus includes: an image capturing unit that captures an image of a substrate placed on a stage; a learning unit that executes a machine learning using training data in which information indicating a state of the substrate is attached to the image of the substrate, so as to generate a substrate state determination model in which the image of the substrate is taken as an input and a value related to the state of the substrate corresponding to the image of the substrate is taken as an output; and a determination unit that determines the state of the substrate corresponding to the image of the substrate captured by the image capturing unit, using the substrate state determination model generated by the learning unit.