2025 AIChE Annual Meeting

(218a) Stabilizing Yield-Stress, High-Viscosity CFD Simulations Using Implicit Large Eddy Simulation

Authors

Brian DeVincentis, M-Star Simulations
A generalized numerical framework for simulating yield-stress, high-viscosity fluids in laminar and transitional flow regimes is presented. The framework employs the Implicit Large Eddy Simulation (ILES) paradigm, where numerical dissipation introduced by the discretization scheme serves as an unstructured surrogate for subgrid-scale turbulence modeling. Although originally developed for high Reynolds number flows, we demonstrate that ILES provides robust stabilization even in low-Reynolds number systems where traditional turbulence models are inapplicable and explicit stabilization terms are poorly defined.

The fluid rheology is modeled using a Bingham plastic constitutive law, with the governing Navier–Stokes equations solved via the lattice Boltzmann method on a uniform grid. Simulations are executed in fully transient form without time-averaging or simplification of the constitutive physics. Across all cases, the ILES approach damps non-physical oscillations near yield surfaces and suppresses pressure instabilities typically observed in plug-dominated flows. No additional artificial viscosity, filtering, or empirical closure terms are required.

As an illustrative application, we apply this framework to tank geometries and material properties representative of those encountered in nuclear waste treatment processes. In these systems, fluids present measurable yield stresses, strong shear-thinning behavior, and transient off-gassing. We find that ILES enables stable simulation of unsteady yield-surface motion and interface deformation, with accuracy consistent across changes in viscosity, yield stress, and driving force.

The results support the broader utility of ILES as a general-purpose stabilization strategy for complex rheological systems—particularly where plug formation, shear localization, and low-speed recirculation zones render conventional methods unstable or overly parameterized.