Synthesizing Measurement and Simulation Data for Semiconductor Process Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The semiconductor field faces challenges in generating a machine learning model with high consistency and generalization due to the lack of adequate and diverse measurement data, which limits the availability of sufficient training data for accurate semiconductor process modeling.
Innovation Solution
A method is developed to generate simulation data that reflects the uncertainty of measurement data by obtaining simulation input and output datasets, extracting noise information, and synthesizing these with measurement data to create a noise simulation dataset, ultimately forming a synthesized dataset for training data augmentation in semiconductor process modeling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If measurement data is used for training machine learning models, then the models reflect real-world conditions, but the lack of data quantity and diversity limits model performance
Solution Approach 1:
The patent combines measurement data with simulation data to create a hybrid training dataset. Measurement data provides real-world accuracy while simulation data expands quantity and diversity, resolving the contradiction between data quality and quantity for machine learning model training in semiconductor manufacturing.
Solution Approach 2:
The patent uses simulation to create virtual copies of measurement data through detailed process modeling. These simulated datasets replicate real manufacturing conditions and variations, providing additional training examples without requiring physical measurements, thus increasing data quantity while maintaining realism.
2Adaptability or versatility
If simulation data is generated to augment measurement data, then data diversity increases, but the complexity of the simulation process increases
Solution Approach 1:
The patent segments the data generation process into distinct components: measurement data extraction, simulation model execution, and data synthesis. This modular approach manages complexity by breaking down the overall process into manageable stages, each handling specific tasks in the data augmentation workflow.
Solution Approach 2:
The patent introduces a structured data synthesis framework that acts as an intermediary between simulation models and machine learning training requirements. This framework standardizes the integration process, managing complexity through formalized procedures for combining simulation outputs with measurement data.
3Reliability
If noise information is extracted and synthesized, then the training data reflects real-world uncertainty, but the processing complexity increases
Solution Approach 1:
The patent extracts noise and uncertainty characteristics from measurement data through statistical analysis, separating these properties from the core measurement values. This extraction enables the noise information to be independently applied to simulation data, enhancing realism while managing processing complexity through modular operations.
Data Source
AI summary
A method of augmenting training data for a semiconductor process modeling includes obtaining a simulation input data set and a measurement data set, obtaining a simulation output data set generated based on performing simulation based on the simulation input data set, extracting reference noise information associated with the measurement data set from the measurement data set, extracting distribution information associated with each simulation case included in the simulation output data set based on synthesizing the reference noise information and the simulation output data set, generating a noise simulation data set based on sampling data based on the distribution information, and generating a synthesized data set based on synthesizing the simulation input data set, the noise simulation data set, and the measurement data set.