Machine-Learning Substrate State Prediction in Finishing Processes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
It is challenging to accurately monitor and predict the state of a substrate during or after the finishing process in substrate processing, particularly due to the complex interactions of operating states in substrate processing apparatuses.
Innovation Solution
An information processing apparatus that acquires finishing conditions, including substrate-holder state information and finishing-fluid-supply-structure state information, and uses a machine-learning model to predict substrate state information, thereby enabling appropriate prediction of the substrate state during or after processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a sensor is directly attached to each substrate to detect the state of the substrate, then the measurement precision of substrate state is improved, but the device complexity and cost increase significantly
Solution Approach 1:
The patent creates a virtual model (copy) of the substrate state through machine learning inference instead of using physical sensors on each substrate. The learning model replicates the substrate state prediction capability without requiring physical sensor attachment, thus achieving measurement precision while avoiding device complexity increase.
Solution Approach 2:
The patent introduces an intermediary machine learning model that infers substrate state from apparatus operating states. This intermediary system bridges the gap between measurable apparatus parameters and the desired substrate state information without requiring direct substrate sensing.
2Adaptability or versatility
If the substrate processing apparatus components operate with complex and varying states, then the adaptability of the processing system is improved, but the difficulty of analyzing the effects of respective operating states on substrate state increases
Solution Approach 1:
The patent transforms the complex multi-parameter substrate state analysis problem into a parameter change detection problem. The machine learning model learns how changes in apparatus operating parameters correlate with substrate state changes, making the analysis of complex interactions tractable through pattern recognition rather than direct causal analysis.
Solution Approach 2:
The patent implements a feedback mechanism where the learning model continuously infers substrate state from apparatus operating states and uses this information to improve predictions. The system learns from the relationship between operating states and substrate outcomes, creating a closed-loop understanding that simplifies complex causal analysis.
3Productivity
If machine learning is used to predict substrate state from finishing conditions, then the productivity of quality monitoring is improved, but the loss of information about direct substrate measurement increases
Solution Approach 1:
The patent performs preliminary action by training the machine learning model in advance using historical data that includes both apparatus operating states and actual substrate state measurements. This pre-training ensures that when the model is deployed for productivity improvement, it has already learned the relationships needed to make accurate predictions without losing critical information.
Solution Approach 2:
The patent enables the system to self-service by having the learning model automatically infer substrate state from operating conditions without requiring external direct measurement. The model serves itself by learning from historical data and then autonomously predicting substrate state, maintaining information accuracy while improving monitoring productivity.
Data Source
AI summary
An information processing apparatus includes: an information acquisition section (500) configured to acquire finishing conditions including substrate-holder state information indicating a state of a substrate holder and a finishing-fluid-supply-structure state information indicating a state of a finishing-fluid supply structure in a finishing process of a substrate performed by a substrate processing apparatus including the substrate holder configured to hold the substrate and the finishing-fluid supply structure configured to supply a substrate finishing fluid onto the substrate; and a state prediction section (501) configured to predict substrate state information for the substrate on which the finishing process is performed under the finishing conditions by inputting the finishing conditions acquired by the information acquisition section (500) to a learning model (10A, 10B) that has been generated by machine learning that causes the learning model to learn a correlation between the finishing conditions and substrate state information indicating a state of the substrate on which the finishing process is performed under the finishing conditions.