Extrapolations of the trained network to a wide range of problems are also validated through numerical experiments, including

  • Linear elasticity with high contrast of phase properties

  • Nonlinear history-dependent plasticity

  • Finite-strain hyperelasticity under large deformations.

The deep material network is capable of accurately capturing nonlinear material and geometric responses with much less DOFs, which is important for material design and concurrent multiscale simulation.

 

A unique feature of the deep material network is that the computational time is proportional to the number of DOFs in the system. ​

Comparison of DNS and data-driven model for elasto-plastic RVE

Comparison of DNS and data-driven model for hyperelastic RVE under large deformations