Searcharxiv⌕ Search

arXiv subjects

Ismail Oubarka

Publications and source records attributed to Ismail Oubarka.

2 recordsLinked to original sources

Well-Balanced WE-PINNs for the Shallow Water Equations over Smooth and Discontinuous Bathymetries

Physics-informed neural networks (PINNs) solve partial differential equations without a mesh, but they do not in general preserve the steady states that well-balanced schemes reproduce exactly. We present a well-balanced extension of the weak and entropy PINN (WE-PINN) framework to the shallow water equations over smooth and discontinuous bathymetry. The residual is a weak space--time control-volume identity enforced on randomly placed control volumes that may overlap, so no mesh is needed. In free-surface variables, the geometric source is integrated by parts onto the bed, so that the network is never differentiated, and a cell-local bed-slope correction makes the quadrature reproduce the exact bed increment. Bed discontinuities, located from the bathymetry alone, separate sub-domains carrying independent networks, coupled through conservation of discharge and total head. We prove that the fully discrete residual vanishes identically on the lake at rest for any control-volume family, quadrature order and nominal bed slope, so that the equilibrium is an exact global minimiser of the training objective; that the continuous residual characterises weak solutions; that the slope correction is consistent; and that no approximation with a continuous free surface can satisfy the head condition at a step with non-zero discharge, so that the decomposition is necessary. Numerically, the residual vanishes to round-off on four lake-at-rest configurations over smooth, discontinuous, compound and tabulated beds, and the trained network reaches the equilibrium to round-off under an extended training budget. On seven Riemann problems at a bottom step, covering sub-, super- and critical transitions, the network attains relative $L^1$ free-surface errors between $6.9\times10^{-4}$ and $3.5\times10^{-3}$, and reaches critical states without prescribing them.

math.NA↗

Weak and entropy physics-informed neural networks for conservation laws

We propose Weak and Entropy PINNs (WE-PINNs) for the approximation of entropy solutions to nonlinear hyperbolic conservation laws. Standard physics-informed neural networks enforce governing equations in strong differential form, an approach that becomes structurally inconsistent in the presence of discontinuities due to the divergence of strong-form residuals near shocks. The proposed method replaces pointwise residual minimization with a space--time weak formulation derived from the divergence theorem. Conservation is enforced through boundary flux integrals over dynamically sampled space--time control volumes, yielding a mesh-free control-volume framework that remains well-defined for discontinuous solutions. Entropy admissibility is incorporated in integral form to ensure uniqueness and physical consistency of the weak solution. The resulting loss functional combines space--time flux balance and entropy inequalities, without resorting to dual-norm saddle-point formulations, auxiliary potential networks, or fixed discretization meshes. This makes the proposed method remarkably easy to implement, requiring only a simple standard neural network architecture. We establish a rigorous convergence analysis linking the network's loss function to the $L^1$ error towards the entropy solution, providing the first explicit $L^1$ convergence rate for a mesh-free control-volume PINN formulation via the Bouchut-Perthame framework for scalar conservation laws. Numerical experiments on the Burgers equation, the shallow water equations, and the compressible Euler equations demonstrate accurate shock resolution and robust performance in both smooth and shock-dominated regimes.

math.NA↗