FMU-Based Surrogate Modeling for License-Free Parallel Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Proprietary modeling tools are computationally expensive and limited in their ability to create surrogate models for real-time applications, requiring costly licenses and restricting the number of tests that can be run, while also being restricted to models developed within their specific tools.
Innovation Solution
A method to convert proprietary models into tool-agnostic surrogate models using the Functional Mockup Interface (FMI) standard, allowing for automated dataset generation and creation of input/output datasets, which are then used to generate a surrogate model that is free from licensing requirements, enabling license-free real-time simulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If proprietary modeling tools are used to create high-fidelity models, then model accuracy is improved, but licensing costs increase and the number of tests is limited
Solution Approach 1:
The patent creates surrogate models that are simplified copies of high-fidelity proprietary models. These surrogate models replicate the essential behavior and accuracy of the original models but can be executed without proprietary licensing, enabling numerous parallel test runs. The copying principle allows the system to maintain model accuracy while eliminating licensing constraints on the number of tests.
Solution Approach 2:
The patent generates multiple disposable surrogate models that can be created and executed freely without the high cost of proprietary licenses. These surrogate models serve as temporary, cost-effective alternatives for specific testing purposes, allowing extensive parallel testing without the financial burden of maintaining multiple proprietary licenses.
2Reliability
If proprietary modeling tools are used for real-time simulations, then model fidelity is improved, but computational cost and licensing fees increase
Solution Approach 1:
The patent extracts the essential behavioral characteristics and transfer functions from high-fidelity proprietary models and embeds them into simplified surrogate models. This extraction process removes the computationally intensive components while retaining the critical fidelity needed for real-time simulations, thereby reducing computational cost while maintaining model reliability.
Solution Approach 2:
The patent transforms the complex parameters and equations of proprietary models into simplified parameter representations suitable for real-time execution. By changing the mathematical representation from detailed physics-based models to simplified transfer functions with fitted parameters, the system achieves real-time performance with reduced computational cost while preserving essential model fidelity.
3Ease of manufacture
If proprietary modeling tools are used, then model development capability is improved, but adaptability to different tools is reduced
Solution Approach 1:
The patent creates surrogate models with universal interfaces that can be deployed across multiple different simulation tools and platforms. The standardized output format and generic model structure allow the same surrogate model to be used in various environments without requiring proprietary tool-specific formats, thereby achieving multi-tool adaptability while maintaining the benefit of high-fidelity model development.
4Measurement precision
If high-fidelity models are used for batch simulations, then accuracy is improved, but execution time increases
Solution Approach 1:
The patent creates simplified copy models (surrogate models) that replicate the accuracy of high-fidelity models but execute much faster. These copied models are specifically optimized for batch simulation scenarios where numerous runs are needed, providing the same accuracy level with significantly reduced execution time by eliminating computationally intensive calculations.
Solution Approach 2:
The patent performs preliminary model analysis and parameter fitting to create pre-optimized surrogate models before batch simulations are executed. This preliminary action of creating simplified models with pre-fitted parameters allows the subsequent batch simulations to run quickly while maintaining accuracy, as the heavy computational work has already been done in the model creation phase.
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
A method, node, and computer-readable medium are provided to convert a proprietary model to a tool-agnostic surrogate model using a functional mockup interface (FMI) standard. A proprietary model is received as a functional-mockup unit (FMU) An automated dataset generation is performed on the FMU to create input/output datasets based on design of experiments and input requirements. Steady-state operational-points are determined. The tool-agnostic surrogate model is generated based on the input/output datasets and the steady-state operational-points. The tool-agnostic surrogate model is output as an output FMU model that is free of licensing requirements of a license for the proprietary model. The tool-agnostic surrogate model may be a steady-state surrogate model, a dynamic surrogate model, or a combination thereof.