Convex signed singular value neural network (CSSV-NN) for isotropic hyperelasticity: learns a strain-energy density from stress data alone, with objectivity, isotropy, polyconvexity and thermodynamic consistency guaranteed by construction rather than by penalty terms. Reference implementation for Geuken et al., JMPS 203 (2025).
machine-learning neural-network tensorflow scientific-computing neural-networks computational-mechanics continuum-mechanics hyperelasticity material-modeling constitutive-modeling polyconvexity
-
Updated
Aug 10, 2026 - Python