Manufacturing Digital Twin Control With Reduced-Order Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing manufacturing processes face challenges in efficiently utilizing high-fidelity simulation models due to computational complexity and data entry complexity, limiting their effectiveness in real-world applications.
Innovation Solution
Implementing reduced order models and neural network-based deep learning models trained on design of experiments (DoE) data, deployed on edge computers for real-time data processing using Unified Name Space (UNS) and MQTT protocol, to create digital twins of products and processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-fidelity simulation models are used for manufacturing processes, then manufacturing precision and reliability are improved, but computational complexity and data entry complexity increase
Solution Approach 1:
The patent creates a digital twin model that is a virtual copy of the physical manufacturing process. This digital twin replicates the behavior and characteristics of the actual manufacturing system, allowing high-fidelity simulation without requiring repeated physical trials. The digital twin model captures the essential dynamics and relationships of the manufacturing process, enabling accurate predictions and optimizations while reducing the need for complex computational resources compared to traditional high-fidelity simulations.
2Manufacturing precision
If high-fidelity simulation models are used for manufacturing processes, then manufacturing precision is improved, but data entry complexity increases
Solution Approach 1:
The digital twin model automatically acquires and updates its data from the actual manufacturing process through sensors and data collection systems. Instead of requiring manual data entry, the system self-updates by continuously monitoring process parameters, machine states, and product characteristics. This automated data flow reduces the burden of data entry while maintaining high precision through continuous real-time data synchronization between the physical system and its digital counterpart.
3Productivity
If reduced order models and neural network-based deep learning models are used, then computational complexity is reduced and productivity is improved, but manufacturing precision may be compromised
Solution Approach 1:
The patent employs a dynamic approach where the digital twin model adapts and evolves over time. Neural network-based deep learning models are trained on historical manufacturing data and continuously refined as new data becomes available. This dynamic learning process allows the simplified models to progressively improve their accuracy and precision while maintaining computational efficiency. The system balances model complexity with performance requirements, adjusting the level of detail based on the specific application needs.
Data Source
Figure 1A~1B
Figure 2A~2C
Figure 3
AI summary
A method of controlling a manufacturing process includes: creating a digital twin model representing one of the manufacturing process or a manufactured article that is formed or modified by the manufacturing process; revising the digital twin model with real-time data regarding a physical instance of the one of the manufacturing process or the manufactured article; making a decision, based on the digital twin model, regarding the physical instance of the one of the manufacturing process or the manufactured article; and causing or modifying an action, based on the decision regarding the physical instance of the one of the manufacturing process or the manufactured article.