Fusion Model for Physical System Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in developing accurate models for physical systems due to incomplete understanding of physics, leading to complex and costly computations, and the need for expensive physical testing and validation.
Innovation Solution
A method and apparatus for generating a fusion model for a physical system by training models using low-fidelity and high-fidelity data, augmenting data with virtual sensors, and creating a fusion model that combines the predictions of both models to improve accuracy and reduce computational costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If models account for more physics to improve accuracy, then measurement precision improves, but device complexity and computational cost increase
Solution Approach 1:
The patent segments the modeling process into multiple fidelity levels (low-fidelity and high-fidelity models) that operate at different complexity tiers. The low-fidelity model handles general physics with simpler computations, while the high-fidelity model is invoked only when needed for specific accuracy requirements, thus dividing the computational workload and reducing overall system complexity.
Solution Approach 2:
The patent dynamically changes the model fidelity parameter based on operating conditions and accuracy requirements. The system transitions between low-fidelity and high-fidelity models by adjusting parameters such as computational resolution, physics inclusion, and data fidelity levels, allowing optimal balance between accuracy and computational cost for each scenario.
2Measurement precision
If high-fidelity models are used to improve accuracy, then measurement precision improves, but use of energy and computational cost increase
Solution Approach 1:
The patent applies partial action by using high-fidelity models only for specific portions of the operational domain where accuracy is critical, rather than consistently applying high-fidelity computations throughout. The system determines when high-fidelity modeling is necessary and applies it selectively, reducing overall computational cost while maintaining required accuracy levels.
Solution Approach 2:
The low-fidelity model serves as an intermediary that provides preliminary results and guides when high-fidelity modeling is needed. This intermediary approach allows the system to filter out cases where low-fidelity approximations are sufficient, thereby reducing unnecessary high-cost computational operations while preserving accuracy where required.
3Reliability
If physical testing is conducted to validate models, then reliability improves, but loss of time and cost increase
Solution Approach 1:
The patent creates virtual copies of physical systems through high-fidelity computational models that replicate physical behavior. These digital twins serve as substitutes for physical testing, allowing validation and exploration of system performance without requiring actual physical prototypes or tests, thereby reducing time and cost while maintaining reliability.
Solution Approach 2:
The system performs preliminary validation using low-fidelity models and synthetic data before conducting any physical testing or high-fidelity simulations. This preliminary action filters out obviously invalid scenarios and prepares validation protocols in advance, reducing the time and resources required for subsequent physical testing and verification.
Data Source
AI summary
A method and apparatus of a device for generating a fusion model for a physical system is described. In an exemplary embodiment, the device receives low-fidelity input data and low-fidelity output data that represent a low-fidelity measurement of a physical system. In addition, the device trains a first model to predict the low-fidelity output data using the low-fidelity input data. Furthermore, the device receives high-fidelity input data and high-fidelity output data that represent a high-fidelity measurement of the physical system. The device additionally invokes the first model with the high-fidelity input data to generate predicted low-fidelity data. The device further trains a second model to predict the high-fidelity output data using the high-fidelity input data augmented with the predicted low-fidelity data. In addition, the device creates a fusion model for the physical system based on the first model and the second model, the first model and the second model to receive input to the fusion model, the second model to receive output from the first model, and output of the fusion model corresponding to output of the second model.