In many modelling areas, pure physical models are either not available or not accurate enough. Here, surrogate modelling and machine learning can help. One solution is the training of neuronal networks with measurement data. A well-known and accepted platform for this purpose – TensorFlow – can deliver suitable model descriptions. But how to integrate these neuronal networks into Modelica models?

Here is our solution for you: SMArtIInt - Simple Modelica Artificial Intelligence Interface Library. It is open-source and available for free under GitHub .