IJNME Article On Robust Inverse Materials Design using Voigt-Reuss-NN is now online

February 25, 2026 /

In our recent article we combine physics-constrained machine learning and inverse materials design

Our approach for robust inverse materials design is centered on a physics-enforcing spectral normalization scheme used within our "Voigt-Reuss-Network": A feed-forward neural network that rigorously respects the upper Voigt and the lower Reuss bound. We apply the method to 2D and 3D problems of linear steady-state heat conduction and linear elasticity.

The new paper titled

"Robust inverse material design with physical guarantees using the Voigt-Reuss Net" by Sanath Keshav, Felix Fritzen

addresses technical aspects as well as challenging applications, comparing the VRNet against a variety of other models.

The article is now online at International Journal for Numerical Methods in Engineering.

 

Link to paper

To the top of the page