Surrogate Model Language Conversion for Simulation Interoperability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing scientific computing languages face inefficiencies when models from different languages are coupled for simulation, with the Functional Mockup Interface (FMI) standard being non-compatible with all languages, reducing performance and stability, and requiring resource-intensive computations.
Innovation Solution
A method and system that use machine learning to transform a model in one scientific computing language into a surrogate in another language, allowing direct simulation without coupling, using techniques like Continuous Time Echo State Networks (CTESN) and interpolating functions to generate surrogates that can be trained across various inputs, and integrated with Julia Computing's ModelingToolkit.jl for enhanced performance and automation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If models from different scientific computing languages are coupled together using existing interfaces, then interoperability between languages is achieved, but simulation performance deteriorates and computational resources are consumed excessively
Solution Approach 1:
The patent creates a surrogate model that copies the input-output behavior of the original model from a different scientific computing language. Instead of coupling the original models directly, the surrogate model replicates its functionality in the target language's native format, enabling direct usage without interface overhead while maintaining interoperability.
Solution Approach 2:
The patent replaces the mechanical coupling interface (FMI or other standards) with a data-driven surrogate model. The explicit interface coupling mechanism is substituted by a trained neural network surrogate that directly computes the model behavior in the target language, eliminating the performance penalty of interface overhead.
2Adaptability or versatility
If the Functional Mockup Interface standard is used to couple models from different languages, then model interchangeability is enabled, but numerical stability deteriorates due to interface constraints
Solution Approach 1:
The surrogate model copies the numerical behavior of the original model without being constrained by FMI interface specifications. By training the surrogate on input-output data pairs, it learns to replicate the stable numerical characteristics of the original model in the target language's native computational environment.
3Adaptability or versatility
If models are translated to the FMI standard for coupling, then interoperability between different modeling languages is achieved, but computational performance is reduced
Solution Approach 1:
The patent creates a surrogate model that copies the functional behavior of the FMI-translated model but executes it natively in the target scientific computing language. This eliminates the performance overhead of FMI interface processing while maintaining the model's computational capabilities.
Solution Approach 2:
The surrogate model is trained in advance on comprehensive input-output data from the original model. This preliminary training phase captures the model's behavior patterns, allowing the surrogate to execute predictions rapidly without requiring real-time interface translation or coupling during simulation.
4Adaptability or versatility
If coupling interfaces are used to integrate models from different languages, then multi-language simulation capability is achieved, but device complexity increases
Solution Approach 1:
The patent creates a surrogate model that copies the original model's behavior in the target language's native format. This eliminates the need for coupling interfaces, model translation layers, and synchronization mechanisms, thereby reducing system complexity while maintaining multi-language simulation capability through the surrogate.
Data Source
AI summary
Methods and computer systems for transforming a model of a first scientific computing language to a model (e.g., surrogate) of a second scientific computing language such that the surrogate is trained across a plurality of possible inputs is disclosed. Model inputs in a first scientific computing language are received. A surrogate is generated based on the received model input. The surrogate may be trained across a plurality of possible inputs by selecting an input function representation for the model input, selecting a parameter space, sampling the parameter space to generate a training set of time series for each parameter set, simulating a reservoir, computing projections from the simulated reservoir, and fitting an interpolating function between the projections to establish an approximate projection for unknown input functions and parameter values. The surrogate is then deployed, either trained or untrained depending on the embodiment.