Cross-Device Process Estimation for Reproducible Material Manufacturing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of reproducibly manufacturing new materials across different manufacturing devices, such as small-scale research equipment and large-scale production facilities, is hindered by equipment scale, usage conditions, environmental differences, and individual device variations, leading to inefficiencies and material waste.
Innovation Solution
A computer-based process estimation method and device that uses machine learning to create regression models correlating process data from one manufacturing device with structure data, enabling the estimation of process data for a different device to achieve consistent material production.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If machine learning regression models are used to estimate process data across different manufacturing devices, then manufacturing precision and reproducibility are improved, but device complexity and data processing requirements increase
Solution Approach 1:
The patent introduces machine learning regression models as an intermediary between process data from different manufacturing devices. These models learn the relationships between process parameters and material structure data, enabling accurate estimation and translation of process conditions across devices with different scales and characteristics, thus achieving reproducibility without direct device modification
Solution Approach 2:
The patent replaces direct physical replication of manufacturing processes with a data-driven computational system. Instead of physically adjusting each device to match identical operating conditions, the system uses machine learning models to substitute the complex mechanical and physical adjustments with algorithmic process estimation and translation
2Adaptability or versatility
If multiple regression models are created to account for device differences, then adaptability across devices is improved, but loss of time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions by training regression models in advance using historical process data from multiple manufacturing devices. The models learn and store the relationships between different device operating conditions and material outcomes beforehand, enabling rapid process estimation and translation when needed without performing extensive real-time analysis
Solution Approach 2:
The patent utilizes parameter changes in the regression models to adapt to different device characteristics. By adjusting model parameters during training based on specific device data, the system achieves high adaptability across devices with different scales, heating methods, and operating conditions while maintaining efficient estimation performance
Data Source
AI summary
A computer-performed process estimation method and a process estimation method using the computer are provided for estimating, based on a first process data including a process information of a predetermined target step performed in a first manufacturing device that manufactures a material through at least one step including the target step, a second process data including a process information of the target step performed in a second manufacturing device that is a different device from the first manufacturing device and manufactures the material through at least one step including the target step. This method includes machine-learning a relationship between the first process data and a first structure data obtained from a sample after the target step in the first manufacturing device, and creating a first regression model representing a correlation between the first process data and the first structure data, machine-learning a relationship between the second process data and a second structure data obtained from a sample after the target step in the second manufacturing device, and creating a second regression model representing a correlation between the second process data and the second structure data, creating a third regression model representing a correlation between the first process data and the second process data based on the first regression model and the second regression model, and by using the third regression model, estimating an estimated second process data that includes the second process data corresponding to an estimation source-first process data including the first process data that is an arbitrary estimation source.