On the Prandtl-Kolmogorov 1-equation model of turbulence

Kiera Kean, William J. Layton, Michael SchneierarXiv 2021 · arXiv

Abstract

Turbulence modeling in Computational Fluid Dynamics (CFD) often struggles with the phenomenon of over-dissipation, where numerical models artificially dampen the kinetic energy of a flow, leading to physically inaccurate results. This paper presents a refinement of the classic Prandtl-Kolmogorov 1-equation model by introducing a novel hybrid turbulence length scale. By adjusting the near-wall asymptotics of the length scale, the authors demonstrate that the model can maintain physical consistency with the Navier-Stokes equations, specifically achieving the expected $O(U^3/L)$ energy dissipation scaling. The proposed model is validated through both rigorous mathematical proofs and numerical simulations of Taylor-Couette flow.

Introduction

Predicting turbulent flow is a fundamental challenge in fluid mechanics, typically addressed through the Unsteady Reynolds Averaged Navier-Stokes (URANS) equations. These equations rely on the concept of "eddy viscosity" to represent the influence of small-scale turbulent fluctuations on the mean flow. However, a persistent issue in many existing URANS implementations is their tendency to be overly viscous. This over-dissipation causes simulations to behave as if they possess a lower Reynolds number than intended, effectively smoothing out the chaotic, fine-grained structures that characterize real-world turbulence.

A primary factor contributing to this error is the specification of the turbulence length scale ($l$). In traditional models, the length scale near solid boundaries is often defined as $l = 0.41d$, where $d$ is the distance to the wall. While standard, this approach fails to capture the correct near-wall physics, leading to energy dissipation rates that vanish too slowly as the boundary is approached.

Methodology: The Hybrid Length Scale

This research revisits the 1-equation model originally derived by Prandtl (1945) and Kolmogorov (1942). The authors propose a new, two-part "hybrid" definition for the length scale $l$ to correct the asymptotic behavior of the model:

  1. Interior Flow (Away from walls): The length scale is defined as $l = \sqrt{2k} \tau$, where $k$ represents the turbulent kinetic energy and $\tau$ is a selected time scale. This formulation aligns with Prandtl's early intuition that the length scale should relate to the distance a fluid mass travels before losing its identity within the surrounding flow.
  2. Boundary Flow (Near walls): To correct the near-wall behavior, the traditional formula is modified to $l = 0.41d \sqrt{d/L}$, where $L$ is a global length scale.

This specific adjustment ensures that the eddy viscosity behaves like $O(d^2)$ near the wall. This matches the true physical asymptotics of turbulent flow, preventing the model from becoming "too thick" in the boundary layer. The study utilizes the "background flow" method to establish mathematical upper bounds on energy dissipation and employs the FEniCS software suite for numerical validation.

Theoretical Insights and Scaling

The paper identifies the ratio of near-wall average viscosity to effective global viscosity as the critical parameter determining a model's propensity for over-dissipation. By refining the length scale, the authors provide a theoretical guarantee (Theorem 4.1) that the long-time average energy dissipation $\langle \epsilon \rangle_\infty$ is bounded by a constant times $U^3/L$.

This result is significant because it ensures the model's energy budget remains physically consistent with the Navier-Stokes equations. Without the $d \sqrt{d/L}$ correction, the model would over-predict dissipation near the wall, leading to the aforementioned smoothing of turbulent structures.

Numerical Validation: Taylor-Couette Flow

To test the model, the authors simulated Taylor-Couette flow—the motion of a fluid between two concentric, rotating cylinders. This classic shear flow problem is an ideal benchmark for testing boundary layer physics and vortex formation.

The numerical results closely mirrored the theoretical predictions. The dissipation rate followed the scaling law: $$\langle \epsilon \rangle \simeq (0.5 + 4Re^{-1}) \frac{U^3}{L}$$

Furthermore, visual analysis using the Q-criterion confirmed that the simulations captured persistent, non-axisymmetric vortices at high Reynolds numbers. This demonstrates that the refined 1-equation model maintains the "liveliness" of the flow, preserving complex turbulent structures that are often lost in less sophisticated URANS models.

Conclusion and Future Directions

The introduction of a hybrid length scale provides a robust solution to the problem of over-dissipation in the Prandtl-Kolmogorov 1-equation model. By ensuring correct near-wall asymptotics, the model achieves universal scaling consistent with physical reality.

Future work remains in several areas. The constants used in the theoretical proofs (such as 5/2 and 8) are currently upper bounds and could be further optimized. Additionally, while the model performs well in practice, proving the existence and guaranteed positivity of the turbulent kinetic energy ($k$) for this specific length scale remains an open challenge in mathematical fluid dynamics. Finally, the role of the diffusion term ($-\nu \Delta k$) in the $k$-equation continues to be a point of interest, as its inclusion appears intrinsically linked to the accurate near-wall behavior identified in this study.

Summary updated

BibTeX
@misc{kean2021prandtl,
  author = {Kiera Kean and William J. Layton and Michael Schneier},
  title = {{On the Prandtl-Kolmogorov 1-equation model of turbulence}},
  year = {2021},
  eprint = {2106.09855},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2106.09855},
}
Figure