Water swirling into a vortex eventually stops being smooth: foam and spray appear. Mathematicians describe such flows with equations, but there's a puzzle — sometimes the solution 'explodes,' becoming infinite, and no one knows why. Stochastic modeling doesn't help here; we need a precise tool.
The authors took a small neural network with an unusual design: instead of straight lines, it uses waves. Such a network can't draw sharp corners, so when a rupture appears in the flow, it starts making errors, and the error concentrates exactly where the liquid tears. It's like trying to draw a crack on glass with a soft pencil: the pencil slips and leaves a mark right at the fracture line.
The method was tested on a classic example — the Taylor–Green vortex, where fluid swirls in a cube. At normal viscosity, the network's error was weak, but when the fluid became nearly weightless, the error grew 13-fold and gathered at a single point — where theory predicts a rupture. The network also found a boundary: if viscosity exceeds a certain value, ruptures disappear and the flow stays smooth.
This approach is like a sensor on an airplane wing that senses in advance where the air is about to break into a vortex. For now, it's a mathematical tool, but it could help engineers predict dangerous swirls in turbines or around hulls.
as written by the authors arXiv:2603.18061
We introduce a method for detecting regularity loss in solutions to the three-dimensional Navier-Stokes equations using the approximation error of Sinusoidal Representation Networks (SIRENs). SIRENs use sin() activations, producing C-infinity outputs that cannot represent non-smooth features. By classical spectral approximation theory, the SIREN error is bounded by O(N^{-s}) where s is the local Sobolev regularity. At a singularity (s to 0), the error is O(1) and localizes via the Gibbs phenomenon. We decompose the velocity field into a cheap analytical baseline (advection-diffusion) and a learned residual (pressure correction), training a compact SIREN (4,867 parameters). We validate on the 3D Taylor-Green vortex, where error concentration increases from 4.9x to 13.6x as viscosity decreases from 0.01 to 0.0001, localizing to the stagnation point -- the geometry matching the singularity proven by Chen and Hou (2025) for 3D Euler. On axisymmetric equations, we reproduce blowup signatures (T* converging across resolutions) and identify a critical viscosity nu_c = 0.00582 for the regularization transition.
The classic Taylor–Green vortex, on which the method was tested, was first described in 1937 to study turbulence, and since then it has remained a benchmark for testing new ideas in fluid dynamics.
In Stanislaw Lem's novel 'Solaris,' the ocean behaves like a fluid with unpredictable ruptures — perhaps similar methods could help understand its nature.