Model Order Reduction

Finite Element models consists in solving complex PDEs on often huge meshes. Simplifying those models (for fast design space exploration purposes) is not an easy task as we will quickly be lacking accuracy in prediction. For example, it's easy to understand that a Neural Network model trained on an airfoil lift coefficient will be less effective than computing this same lift coefficient based on the CFD simulated pressure field around the airfoil.

Model Order Reduction (Proper Orthogonal Decomposition here) is one way to make machine learning with FEM data.

Run fast design space exploration

Scilab Model Order Reduction is providing you with the mandatory utilities to reduce a ...

  • 2D/3D
  • steady/transient
  • multiparametric
  • Finite Element/Volume

... model in a non-intrusive manner.

Leveraging Proper Orthogonal Decomposition, model is computed based on simulation results. You can build your model importing data from any software as text file using either the provided user interface or the command line mode.

Deploy among simulation experts

Thanks to Model Order Reduction, you can now deploy Finite Element/Volume models at different stage of the design process. Coupling Scilab Model Order Reduction with Xcos, you will be able to design multi-scale and multi-physics systems which could later be used for real-time control (Model Predictive Control).


Deploy beyond simulation experts

Thanks to Model Order Reduction, you are now able to access in-depth knowledge in a matter of seconds. Use the provided user interface to share this knowledge through parametric visualization. With this tool you can validate the good behaviour of your model but also share your model understanding with on-site technicians, or non-technical colleagues.


Introduction to model order reduction

Airfo il shape optimization

Wind farm optimization