NeuralFMU: Towards Structural Integration of FMUs into Neural Networks

TL;DR

This paper introduces FMI.jl and FMIFlux.jl, open-source tools for integrating FMUs into Julia and neural networks, enabling hybrid modeling of physical systems with data-driven techniques.

Contribution

It presents a novel library extension that combines industry-standard FMUs with neural networks for improved physical modeling.

Findings

Enables structural integration of FMUs into neural networks.

Facilitates hybrid modeling combining physics-based and data-driven approaches.

Provides open-source tools for the Julia programming environment.

Abstract

This paper covers two major subjects: First, the presentation of a new open-source library called FMI.jl for integrating FMI into the Julia programming environment by providing the possibility to load, parameterize and simulate FMUs. Further, an extension to this library called FMIFlux.jl is introduced, that allows the integration of FMUs into a neural network topology to obtain a NeuralFMU. This structural combination of an industry typical black-box model and a data-driven machine learning model combines the different advantages of both modeling approaches in one single development environment. This allows for the usage of advanced data driven modeling techniques for physical effects that are difficult to model based on first principles.