.png){kind=icon}
.png){kind=icon}
Return
A Neural-Network Framework to Learn History-Dependent Constitutive Laws and Identifiability of Internal Variables
.png){kind=icon}
PRE
AI
M
L
A
K
DOI：10.1016/j.jmps.2026.106807.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A thermodynamically consistent polyconvex neural network is trained on the space of symmetric deformation gradient tensors and is extended to the space of non-symmetric deformation gradients in a manner consistent with objectivity. • The framework is demonstrated by fitting the Taylor-averaged response of a polycrystalline magnesium microstructure, achieving a relative error of 2%, and resulting in a substantial speed-up over the original model without any compromise on accurate prediction. • Empirically, it is shown that the internal variables are identifiable up to a linear transform; theoretical results supporting this observation are provided.
Keywords:
Polyconvex Neural Network
Crystal Plasticity
Internal Variables
Numerical Homogenisation
Journal
IF:
6
Papers:
5.1K
Citations:
3.0W
Organization
No organization information available
Cited Papers