Adaptive Model Training for Semiconductor Process Drift Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Semiconductor manufacturing equipment faces challenges in detecting when a machine learning model used for process control becomes out of specification due to process drift, making it difficult to accurately predict ex situ measurements, and retraining such models is computationally intensive.
Innovation Solution
A system and method for adaptive model training that receives ex situ and in situ data from multiple process chambers, calculates error metrics, and updates the machine learning model based on these metrics, ensuring that a new model is generated and deployed only when it meets performance criteria, thereby maintaining accurate process control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the machine learning model is continuously retrained to maintain accuracy, then prediction accuracy is improved, but computational cost increases
Solution Approach 1:
The system implements a feedback mechanism where prediction errors are continuously monitored and fed back to trigger model retraining only when necessary. The error metric calculated from ex situ measurements is compared against thresholds to determine whether retraining should be initiated, creating a closed-loop control system that balances accuracy with computational efficiency.
Solution Approach 2:
The system changes the parameter of model update frequency from continuous to conditional based on error metrics. Instead of retraining at fixed intervals or continuously, the model parameters are updated only when the calculated error metric exceeds predefined thresholds, adapting the retraining schedule to actual performance degradation.
2Reliability
If the machine learning model is updated frequently to adapt to process drift, then reliability is improved, but productivity decreases due to computational overhead
Solution Approach 1:
A feedback-based monitoring system tracks prediction errors using ex situ measurements and triggers model updates only when reliability thresholds are breached. This feedback mechanism ensures the model remains reliable through targeted updates while avoiding unnecessary computational interruptions that would reduce manufacturing productivity.
Solution Approach 2:
The system implements periodic evaluation of model performance using calculated error metrics, but executes model updates only periodically when performance degradation is detected. This periodic action pattern maintains reliability through systematic monitoring while minimizing computational overhead by updating only when necessary.
3Measurement precision
If ex situ measurements are used to validate model predictions, then measurement precision is improved, but device complexity increases
Solution Approach 1:
Ex situ measurements serve as an intermediary validation mechanism between the machine learning model and the actual process state. These independent measurements act as a mediator to objectively assess prediction accuracy and trigger model updates, improving validation precision while adding minimal system complexity through existing measurement infrastructure.
Data Source
AI summary
Various embodiments herein relate to systems and methods for adaptive model training. In some embodiments, a computer program product for adaptive model training is provided, the computer program product comprising a non-transitory computer readable medium on which is provided computer-executable instructions for: receiving, from a plurality of process chambers, ex situ data associated with wafers fabricated using the process chambers and in situ measurements, wherein a first machine learning model is used to predict the ex situ data using the in situ measurements; calculating a metric indicating an error associated with the first machine learning model; determining whether to update the first machine learning model; and generating a second machine learning model using the ex situ data and the in situ measurements.