Ludovick Gagnon

Chih-Kang Huang, Ludovick Gagnon, Miha Založnik et al.

The multi-scale and non-linear nature of phase-field models of solidification requires fine spatial and temporal discretization, leading to long computation times. This could be overcome with artificial-intelligence approaches. Surrogate models based on neural operators could have a lower computational cost than conventional numerical discretization methods. We propose a new neural operator approach that bridges classical convex-concave splitting schemes with physics-informed learning to accelerate the simulation of phase-field models. It consists of a Deep Ritz method, where a neural operator is trained to approximate a variational formulation of the phase-field model. By training the neural operator with an energy-splitting variational formulation, we enforce the energy dissipation property of the underlying models. We further introduce a custom Reaction-Diffusion Neural Operator (RDNO) architecture, adapted to the operators of the model equations. We successfully apply the deep learning approach to the isotropic Allen-Cahn equation and to anisotropic dendritic growth simulation. We demonstrate that our physically-informed training provides better generalization in out-of-distribution evaluations than data-driven training, while achieving faster inference than traditional Fourier spectral methods.

Ludovick Gagnon, José Urquiza

For a Legendre-Galerkin semi-discretization of the 1-D homogeneous wave equation, the high frequency components of the numerical solution prevent us from obtaining the boundary observability (inequality), uniformly with regard to the discretization parameter. A classical Fourier filtering that filters out the high frequencies is sufficient to recover the uniform observability. Unfortunately, this remedy needs to compute all the frequencies of the underlying system. In this paper we present three cheaper alternative remedies, namely a spectral filtering, a mixed formulation and Nitsche's method to append Dirichlet type boundary conditions. Our numerical results show indeed that uniform boundary observability inequalities may be recovered. On another hand, surprisingly, none of them seems to provide a uniform direct (or trace) inequality, a property which is needed in some existing general convergence results for the adjoint boundary controllability problem. We finally show in a numerical example that, despite this fact, convergence of the control approximations holds whenever the uniform observability inequality is observed numerically.