arXiv Science⌕ Search

arXiv · 2611.16082

Gradient-Aware AI Agents for Autonomous Turbulence Model Discovery via Differentiable CFD

Abstract

Turbulence modeling relies on human choices of closure form, physical inputs, and tunable parameters. Training closures a posteriori is efficient only with gradients through long chaotic rollouts, which few solvers provide. Most a posteriori approaches therefore resort to gradient-free optimization, and nearly all fix the closure form in advance, learning only its parameters. In this work, we introduce a gradient-aware agentic framework that poses closure discovery as a bilevel problem: a multi-agent large language model (LLM) system searches over closure forms, while a GPU-native differentiable computational fluid dynamics (CFD) solver supplies gradients, making differentiable a posteriori learning an action of the autonomous AI agent. For each proposal, the agents construct a differentiable training program by embedding the analytical, neural, or hybrid closure in the solver, implementing a differentiable objective, and validating the gradient path. They then decide whether to evaluate the closure, compute its parameter sensitivities, or train it a posteriori, select the optimization settings, and revise subsequent proposals using the simulation outcomes and returned gradients. Truncated-horizon gradients make this training usable under chaotic dynamics for closures with tens of thousands of parameters. Demonstrations include dissipation-law and subgrid-scale closure discovery in two-dimensional turbulence and wall-model discovery for large-eddy simulation of three-dimensional channel flow. With gradient-based training, the agent recovers reference coefficients that forward evaluation alone does not reach within the same budget, and autonomously discovers and trains analytical and neural closures that improve agreement with the targeted reference statistics. These results demonstrate the feasibility of end-to-end autonomous turbulence model development in time-resolved flows.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Yi Liu, Xiantao Fan, Mingqi Xu, Jian-Xun Wang. 2026-11-11. Gradient-Aware AI Agents for Autonomous Turbulence Model Discovery via Differentiable CFD. https://arxiv.science/abs/2611.16082

Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite arXiv Science by Indiciae: BibTeX · RIS.

KEEP EXPLORING

Related papers

Statistical State Dynamics of Couette MHD Turbulence

The roll streak structure (RSS) is ubiquitous in shear flow turbulence and is fundamental to the dynamics of the self-sustaining process (SSP) maintaining the turbulent state. The formation and maintenance of the RSS in wall-bounded shear flow suggest the presence of an underlying instability that has recently been identified using statistical state dynamics (SSD). Due to the parallelism between the Navier-Stokes equation and the induction equation, it is reasonable to inquire whether the RSS in wall-bounded shear flow has a counterpart in the MHD equations formulated as an SSD. In this work we show that this is the case and that an analytic solution for the composite velocitymagnetic field RSS in the MHD SSD also arises from an instability, that this instability equilibrates to either a fixed point or to a turbulent state, that these turbulent statistical equilibria may be self sustaining, and that both the fixed point and the turbulent states may correspond to large scale coherent dynamos.

physics.flu-dyn↗

A physics-constrained neural solver for incompressible flows without labeled data

We present a physics-constrained neural solver for incompressible flow that requires neither labeled data nor supervised pretraining. The solver enforces incompressibility and boundary conditions (BCs) to strict tolerances. We attribute the limited Reynolds numbers of existing physics-informed approaches to insufficient enforcement of these constraints, on which conventional finite-difference and finite-volume solvers depend. We therefore formulate incompressible flow as a constrained optimization problem within the physics- and equality-constrained artificial neural network (PECANN) framework: the pressure Poisson residual is minimized subject to the momentum equations, continuity equation, and BCs imposed as equality constraints on the primitive variables, while a conditionally adaptive augmented Lagrangian method enforces each constraint. The pressure Poisson objective proves critical. Across all benchmarks, replacing it with the momentum residual fails to converge to accurate solutions under identical architecture and constraint enforcement, which we attribute to the global elliptic coupling of pressure and velocity that local momentum-residual minimization lacks. For advection-dominated flows, we introduce an adaptive vanishing entropy viscosity that stabilizes training while leaving the converged solution unchanged. The solver predicts lid-driven cavity flow up to Re = 7,500 with centerline velocity profiles matching benchmark data, and is further validated on flow over a backward-facing step at Re = 800, 3D unsteady Beltrami flow, and steady and unsteady flow past a circular cylinder under an inflow-outflow configuration, for which an ablation study identifies admissible outlet BCs for velocity and pressure. From a randomly initialized network and without external perturbations, the solver captures the spontaneous onset of periodic vortex shedding in unsteady cylinder flow.

physics.flu-dyn↗

Soft Mobility Theory

Predicting how a deformable body moves and deforms in a viscous flow underlies problems ranging from microorganism locomotion to soft microrobotics, yet existing frameworks are either problem-specific or ill-suited to inverse design. We propose a soft mobility theory. Applying the principle of virtual power to an assembly of hydrodynamically interacting spheres connected by springs and rods yields a saddle-point problem, whose explicit solution is a configuration-dependent ordinary differential equation for the generalized coordinates of the body. This soft mobility equation extends classical rigid-body mobility theory in that the elastic, body-force, and flow-coupling tensors all depend explicitly on the instantaneous deformation. It is not restricted to sphere assemblies: the same equation follows from the continuum problem of a hyperelastic body in a background Stokes flow, of which the assembly is a Galerkin restriction. We compute the mobility from the Rotne-Prager-Yamakawa approximation and validate the framework on canonical problems spanning rigid and flexible bodies in quiescent and shear flows. An open-source JAX implementation makes entire simulations end-to-end differentiable. This allows efficient gradient-based inverse design: as proofs of concept, we recover the asymptotic optimum of a three-sphere swimmer and design a soft gyrotactic "surfer" that exploits passive deformation to ascend faster than its rigid counterpart in a Taylor-Green flow.

physics.flu-dyn↗