Hybrid Surrogate Physics Model Balancing Accuracy and Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning-based surrogate models, while reducing computational load, may be less accurate during extreme events due to sparse training data, and existing methods lack effective strategies for balancing their deployment with physics-based models in complex simulations like weather or financial forecasts.
Innovation Solution
A system that runs a deep learning-based surrogate model and partially runs a physics-based model to check reliability, using a comparison module to determine accuracy and switch between models, with online training of the surrogate model using high-fidelity data from the physics-based model to maintain performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a deep learning based surrogate model is used to reduce computational load, then productivity is improved, but reliability deteriorates under extreme events due to sparse training data
Solution Approach 1:
The patent combines a deep learning based surrogate model with a physics based mathematical model into a hybrid system. The surrogate model handles normal conditions efficiently, while the physics based model ensures accuracy during extreme events. The system dynamically switches between or blends outputs from both models, merging their complementary strengths to achieve both computational efficiency and reliability.
Solution Approach 2:
The patent introduces a switching mechanism or gating network as an intermediary that determines when to use the surrogate model and when to rely on the physics based model. This intermediary component monitors system state or uncertainty levels and routes queries to the appropriate model, ensuring reliability during extreme events while maintaining computational efficiency during normal operations.
2Reliability
If a physics based mathematical model is used to ensure accuracy, then reliability is improved, but productivity deteriorates due to high computational requirements
Solution Approach 1:
The patent applies partial action by using the computationally expensive physics based model only partially - specifically during extreme events or when the surrogate model's predictions fall outside its reliable range. For normal conditions, the system relies on the efficient surrogate model, thus avoiding unnecessary computational overhead while maintaining accuracy when needed.
Solution Approach 2:
The patent segments the simulation domain or time periods into regions where the surrogate model is sufficient and regions where the physics based model is required. This segmentation allows the system to apply computational resources selectively, using the high-fidelity physics based model only where and when it is necessary for accuracy.
3Measurement precision
If the physics based model is run fully to verify surrogate model reliability, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The patent uses partial running of the physics based model for verification purposes - rather than always executing the full physics based model, it runs the model partially or selectively to verify surrogate model predictions only when necessary, such as during extreme events or at periodic intervals, thus reducing time loss while maintaining verification precision.
Data Source
AI summary
A computer-implemented method, a computer program product, and a computer system for optimally balancing deployment of a deep learning based surrogate model and a physics based mathematical model in simulating a complex problem. One or more computing devices or servers compare results of running the deep learning based surrogate model with results of partially running the physics based mathematical model or with observations. One or more computing devices or severs output the results of running the deep learning based surrogate model as system outputs of simulating the complex problem, in response to determining that the deep learning based surrogate model is reliable. One or more computing devices or servers output results of running the physics based mathematical model as the system outputs of simulating the complex problem, in response to determining that the deep learning based surrogate model is not reliable.