Digital Twin Generation Using Pre-Trained Asset Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for creating digital twins are manual, time-consuming, and expensive, requiring significant human effort, especially when not built during the design phase of the physical asset.
Innovation Solution
A computer-implemented method using a large digital twin model and generative transformer to automatically derive or retrieve digital twins based on virtual asset representations and metadata, leveraging pre-trained models and industrial Metaverse applications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If digital twins are built manually by humans, then the digital twin model can be created with high precision and customization, but the process becomes time-consuming and expensive
Solution Approach 1:
The patent uses pre-trained digital twin models as templates that can be copied and adapted for new physical assets. Instead of building digital twins from scratch, the system copies existing model structures and populates them with asset-specific data, dramatically reducing creation time while maintaining model quality through the copying process
Solution Approach 2:
The patent implements pre-training of digital twin models on large datasets before actual digital twin creation. This preliminary action prepares the models with general knowledge and patterns, so that when a specific digital twin is needed, the model is already equipped to generate accurate results quickly without requiring manual construction
2Adaptability or versatility
If digital twins are built manually by humans, then the digital twin can be customized to specific requirements, but the engineering effort and cost increase significantly
Solution Approach 1:
The patent creates universal digital twin models that can serve multiple purposes and be applied to different asset types. The pre-trained models possess general capabilities that can be adapted to various specific requirements through configuration rather than reconstruction, reducing engineering effort while maintaining versatility
Solution Approach 2:
The patent enables customization of digital twins by changing parameters and configuration settings of pre-trained models rather than modifying the underlying model structure. This allows high adaptability to specific requirements while keeping the engineering effort low, as parameter adjustment is simpler than model reconstruction
3Adaptability or versatility
If digital twins are created after the design phase, then flexibility is maintained, but the cost and time consumption increase
Solution Approach 1:
The patent enables rapid copying of pre-trained digital twin models to new assets regardless of when the process occurs. This copying mechanism maintains flexibility to create digital twins at any stage while dramatically improving efficiency compared to manual building, as the same template can be reused across multiple assets and time points
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
A method and system for providing a digital twin (DT) of a physical asset comprising an asset representation loader (ARL) adapted to load a virtual asset representation of the physical asset; an asset compare and lookup unit (ACL) adapted to process the loaded virtual asset representation of the physical asset to provide an asset class of the respective physical asset and/or to extract metadata related to identified physical characteristics of the respective physical asset; and a large digital twin model (LDTM) comprising a generative transformer pre-trained on digital twins (DTs) adapted to derive a digital twin (DT) of the physical asset based on the loaded virtual asset representation of the physical asset and the asset class and/or the extracted metadata of the physical asset provided by the asset compare and lookup unit (ACL).