Semiconductor Process Search Using Transfer Learning Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the control parameters of semiconductor processing face challenges in accurately predicting processing conditions due to the selection of inappropriate reference processing data, leading to increased process development periods and costs.
Innovation Solution
A search apparatus and method that utilize transfer learning to generate and regenerate learning models based on first and second data, automatically adding new reference processing data to improve prediction accuracy when the generated model does not meet a predetermined criterion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If transfer learning is performed using existing reference processing data, then the initial model generation is accelerated, but the prediction accuracy deteriorates when the reference data is inappropriate
Solution Approach 1:
The system implements automatic evaluation of the learning model using verification data, and when the accuracy does not meet the predetermined criterion, it triggers feedback to automatically acquire new reference processing data and regenerate the model, creating a closed-loop system that continuously improves prediction accuracy
Solution Approach 2:
The system performs preliminary action by automatically acquiring new reference processing data and regenerating the learning model before final deployment, ensuring that the model meets the required accuracy threshold through pre-validation with verification data
2Measurement precision
If the learning model is regenerated with added reference processing data, then the prediction accuracy is improved, but the process development time increases
Solution Approach 1:
The system performs self-service by automatically evaluating the learning model, determining when accuracy criteria are not met, acquiring new reference processing data, and regenerating the model without requiring manual intervention, thereby reducing overall process development time despite multiple regeneration cycles
Solution Approach 2:
The system dynamically adjusts the reference processing database by automatically adding new data only when the evaluation indicates insufficient accuracy, making the data acquisition process adaptive rather than static, thus optimizing the balance between accuracy improvement and time consumption
3Measurement precision
If manual verification experiments are performed to validate prediction processing conditions, then the accuracy of target processing condition determination is improved, but the process development cost increases
Solution Approach 1:
The system uses verification data to automatically evaluate the learning model and provides feedback on prediction accuracy, replacing costly manual verification experiments with automated computational evaluation that maintains accuracy validation while reducing development costs
Solution Approach 2:
The system substitutes manual mechanical verification experiments with automated computational evaluation using verification data and predetermined accuracy criteria, replacing human labor and physical experimentation with automated information processing
Data Source
AI summary
In order to enable a user to utilize optimum reference processing data for searching for a target processing condition from a large number of stored reference processing data without requiring special knowledge of machine learning, a search apparatus that searches for a manufacturing condition corresponding to a desired processing result of a semiconductor manufacturing apparatus by predicting the manufacturing condition corresponding to the desired processing result using a learning model is configured to generate a learning model by transfer learning using first data and second data, and regenerate, when the generated learning model does not satisfy a predetermined determination criterion, a learning model by transfer learning using the first data and the added second data.