NeuralFMU: Towards Structural Integration of FMUs into Neural Networks
This approach addresses the problem of modeling difficult physical effects in engineering and simulation domains by integrating FMUs and neural networks, though it appears incremental as it builds on existing FMU and neural network techniques.
The paper introduces NeuralFMU, a method for structurally integrating Functional Mock-up Units (FMUs) into neural networks, enabling the combination of industry-standard black-box models with data-driven machine learning to model complex physical effects that are hard to capture with first principles.
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.