Hello Tristan,
I see that you are having a Dymola model and planning to export it as an FMU for use in Simulink. It sounds like your goal is to train your model.
I assume you have the training data, if not, you can prepare the data from your FMU to train the neural network. This involves running simulations with different inputs and collecting the corresponding outputs to create a dataset. Use the 'From Workspace' and 'To Workspace' blocks to handle input and output data between Simulink and MATLAB.
After you have the training data ,you can try the following below steps:
- MATLAB functions such as trainnet along with trainingOptions can be used to define the training parameters and train the network.
- Once the neural network is trained, you can use the GENSIM command from the Neural Network Toolbox to export the network to Simulink and simulate it.
- After integrating the trained network, simulate your Simulink model to validate the performance of the neural network. Adjust the network and retrain as necessary based on the simulation results.
You can refer to the following MathWorks documentation links to know more about
- From Workspace-https://www.mathworks.com/help/simulink/slref/fromworkspace.html
- To Workspace-https://www.mathworks.com/help/simulink/slref/toworkspace.html
- trainnet-https://www.mathworks.com/help/deeplearning/ref/trainnet.html
- trainingOptions-https://www.mathworks.com/help/deeplearning/ref/trainingoptions.html
- GENSIM-https://www.mathworks.com/help/deeplearning/ref/network.gensim.html
Consider referring to these similar MATLAB answers which might help you
- https://www.mathworks.com/support/search.html/answers/101260
- https://www.mathworks.com/support/search.html/answers/1634875
Hope this helps you get started.
