Autonomous Process Chamber Requalification Using ML Sensor Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional substrate processing systems lack autonomy, relying on static recipes and scheduled maintenance, which can lead to inefficiencies and increased costs due to the inability to adapt to in-situ conditions.
Innovation Solution
The implementation of a substrate processing system with connected process chambers and computing devices that utilize trained machine learning models to receive sensor measurements, make decisions, and update their training based on actions taken, enabling real-time adjustments and autonomous operations such as determining when to stop processes, perform maintenance, and bring chambers back into service.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If static recipes and scheduled maintenance are used, then system simplicity is maintained, but productivity and adaptability deteriorate due to inability to respond to in-situ conditions
Solution Approach 1:
The system performs self-diagnosis and self-optimization through machine learning models that automatically analyze sensor data and adjust process parameters. The ML models enable the system to serve itself by making autonomous decisions about process adjustments and maintenance needs without external intervention, thereby improving adaptability while managing complexity through automation.
Solution Approach 2:
The system continuously collects sensor data during substrate processing and feeds this information back to the machine learning models. The models process this feedback in real-time to dynamically adjust process recipes and identify maintenance requirements, creating a closed-loop control system that adapts to changing conditions while maintaining manageable complexity through intelligent processing.
2Reliability
If static scheduled maintenance is performed, then maintenance planning is simplified, but reliability deteriorates due to inability to detect actual chamber condition
Solution Approach 1:
The system replaces traditional mechanical/time-based maintenance scheduling with an intelligent software-based maintenance management system. Machine learning models analyze sensor data to predict actual chamber condition and determine optimal maintenance timing, substituting automated intelligent decision-making for manual scheduled maintenance planning, thereby improving reliability while enhancing automation.
Solution Approach 2:
The machine learning models continuously monitor sensor data and predict potential chamber issues before they affect substrate processing quality. By performing preliminary analysis and predicting maintenance needs in advance, the system proactively schedules maintenance at optimal times, improving reliability through early detection while automating the maintenance decision-making process.
3Productivity
If real-time machine learning processing is implemented, then productivity improves through dynamic decision-making, but device complexity increases due to additional computing requirements
Solution Approach 1:
The computing system is segmented into distributed components, with machine learning models deployed across multiple computing devices including cloud-based training systems and edge-based inference systems. This segmentation allows real-time processing at the chamber level while complex model training occurs remotely, improving productivity through localized real-time decisions while managing computing complexity through distributed architecture.
Solution Approach 2:
The system separates model training and inference into different dimensional layers: complex model training occurs in the cloud dimension with full computational resources, while real-time inference executes at the edge dimension with optimized lightweight models. This dimensional separation enables real-time productivity improvements at the chamber level while managing computing complexity by offloading heavy training operations to a different operational dimension.
Data Source
AI summary
A substrate processing system comprises one or more transfer chambers; a plurality of process chambers connected to the one or more transfer chambers; and a computing device connected to each of the plurality of process chambers. The computing device is to receive first measurements generated by sensors of a first process chamber during or after a process is performed within the first process chamber; determine that the first process chamber is due for maintenance based on processing the first measurements using a first trained machine learning model; after maintenance has been performed on the first process chamber, receive second measurements generated by the sensors during or after a seasoning process is performed within the first process chamber; and determine that the first process chamber is ready to be brought back into service based on processing the second measurements using a second trained machine learning model.