Neural Network Control for Lithography Thermal Drift
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional control systems for production systems, such as lithographic apparatuses, face challenges in maintaining optimal operating conditions due to thermal and mechanical drift, and require calibration for each specific system, limiting their efficiency and accuracy.
Innovation Solution
A computer-implemented method using neural network modules to process observation data and update history information, generating control actions that adapt to the current state and history of the system, enabling precise thermal control of optical elements and improving performance over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional control systems use mathematical models for each production system, then measurement precision is improved, but device complexity increases and ease of operation deteriorates
Solution Approach 1:
The patent applies universality by training a single machine learning model on aggregated data from multiple production systems, enabling the model to generalize across different systems without requiring separate calibration for each one. This universal model replaces the need for system-specific mathematical models, reducing complexity while maintaining prediction accuracy.
Solution Approach 2:
The patent uses copying by creating a virtual digital twin of the production system through machine learning. This virtual model replicates the behavior of multiple physical systems and can be used for predictions without requiring physical access to or calibration for each individual system, simplifying the control architecture.
2Ease of operation
If simplified mathematical models are used, then ease of operation is improved, but measurement precision deteriorates during transitions
Solution Approach 1:
The patent applies dynamics by using a machine learning model that can adapt and update its predictions in real-time as the production system transitions between different settings. Unlike static mathematical models, the ML model dynamically adjusts to changing conditions, maintaining accuracy during transitions while keeping the system easy to operate.
Solution Approach 2:
The patent implements feedback by continuously monitoring actual production data and using it to refine and update the machine learning model's predictions. This feedback mechanism allows the system to maintain high measurement precision during transitions by learning from actual system behavior, while the automated nature of the feedback keeps ease of operation high.
3Measurement precision
If detailed simulation models are calibrated for each production system, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent applies universality by creating a single machine learning model that serves multiple production systems simultaneously. Instead of calibrating separate detailed simulation models for each system, the universal model is trained on aggregated data from multiple systems and can be deployed once to control all of them, dramatically improving deployment efficiency while maintaining prediction accuracy.
4Manufacturing precision
If conventional control systems are used, then manufacturing precision is maintained, but adaptability to new conditions deteriorates
Solution Approach 1:
The patent applies dynamics by using a machine learning model that can adapt to new production settings and conditions in real-time. When new settings or conditions are introduced, the ML model can learn from the new data and adjust its predictions accordingly, maintaining manufacturing precision while providing the adaptability that static conventional control systems lack.
Data Source
AI summary
A method of generating control actions for controlling a production system, such as by transmitting the control actions to a control system of the production system. The method includes receiving, by a memory unit, a set of observation data characterizing a current state of the production system; processing, by a first neural network module of the memory unit, an input based on at least part of the observation data to generate encoded observation data; updating, by a second neural network module of the memory unit, history information stored in an internal memory of the second module using an input based on at least part of the observation data; obtaining, based on the encoded observation data and the updated history information, state data; and generating, based on the state data, one or more control actions.