Semiconductor Process Prediction Model Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current semiconductor manufacturing process prediction methods face challenges in accurately predicting the yield and electrical function of final products due to the complexity of the processes involved, often resulting in a high number of defective products.
Innovation Solution
A two-stage semiconductor manufacturing process prediction method that uses a machine learning model to obtain prediction confidence and yield, with the ability to modify the model if confidence is low and adjust yield based on physical defect data, enhancing prediction accuracy and flexibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a machine learning model is used for prediction, then prediction speed is improved, but prediction accuracy deteriorates when process complexity increases
Solution Approach 1:
The prediction system is divided into multiple machine learning models, each specialized for predicting specific defect types or process stages. This segmentation allows each model to focus on specific patterns, improving accuracy for complex processes while maintaining overall prediction speed through parallel processing of multiple specialized models.
Solution Approach 2:
The system dynamically selects and adjusts which machine learning models to use based on the specific process data being analyzed. When process complexity increases, the system can activate more specialized models or adjust model parameters dynamically, allowing prediction accuracy to adapt to process complexity while maintaining efficient processing.
2Measurement precision
If more process data is collected to improve prediction accuracy, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The system extracts only the most relevant features and data points from the process data that have the highest impact on prediction accuracy. By identifying and extracting key predictive features while discarding redundant information, the system maintains high prediction accuracy without requiring complex processing of all available data, thus reducing system complexity.
Solution Approach 2:
The machine learning models are designed to handle multiple types of process data and predict multiple defect types using a unified framework. This multi-functionality allows the system to process diverse process data through standardized models, improving prediction accuracy across different scenarios without proportionally increasing system complexity.
3Adaptability or versatility
If the machine learning model is modified frequently to adapt to process changes, then adaptability is improved, but loss of time increases due to retraining
Solution Approach 1:
The system performs preliminary actions by continuously collecting and preprocessing process data in real-time, preparing it in advance for model retraining. When process changes are detected, the pre-processed data is already ready, allowing rapid model updates without the time-consuming steps of data collection and preprocessing, thus improving adaptability while minimizing retraining time.
Solution Approach 2:
The system implements continuous feedback mechanisms that monitor prediction accuracy and process changes in real-time. When deviations are detected, the feedback loop automatically triggers selective retraining of only the affected model components using recent data, rather than complete retraining. This feedback-driven approach improves adaptability to process changes while significantly reducing the time loss associated with model updates.
Data Source
AI summary
A semiconductor manufacturing process prediction method and a semiconductor manufacturing process prediction device are provided. The semiconductor manufacturing process prediction method includes the following steps. A plurality of process data are obtained. According to the process data, a machine learning model is used to execute prediction and obtain a prediction confidence and a prediction yield. Whether the prediction confidence is lower than a predetermined level is determined. If the prediction confidence is lower than the predetermined level, the machine learning model is modified. According to the process data, the prediction yield is adjusted.