Please find enclosed the second of a series of three works on TANN that was recently published in Computer Methods in Applied Mechanics and Engineering (doi: 10.1016/j.cma.2022.115190, https://authors.elsevier.com/c/1fGS5AQEIzV-w).

In this paper we show how thermodynamics can be explicitly inserted into neural networks to guarantee thermodynamically consistent predictions of the behavior of complex materials with inelastic microstructure. We also show how internal state variables can be automatically discovered and give access to the salient micro-mechanisms that are related to the non-linear macroscopic behavior. Finally, based on a double-scale asymptotic homogenization scheme, we perform multiscale analyses achieving speed-ups of many orders of magnitude compared to micromechanical simulations.

Any comments will be highly appreciated!