GRA-Based Wall Model

• The paper presents a predictive closure using the generalized Reynolds analogy to link mean velocity and enthalpy for accurate wall shear and heat flux predictions.
• It decomposes the flow into a roughness sublayer and an outer layer, applying compressibility transformations to match DNS data over prism-shaped rough surfaces.
• Validation shows high accuracy with drag errors under 10% and heat flux errors below 6% for high-speed compressible flows.

A GRA-based wall model is a predictive closure for heat transfer and friction in compressible turbulent boundary layers over rough surfaces, anchored in the generalized Reynolds analogy (GRA). It provides a mechanistic link between mean velocity and enthalpy (or temperature) profiles above complex roughness, enabling physically interpretable models for wall shear stress and heat flux in high-speed flows. This framework is applicable to zero-pressure-gradient, compressible turbulent boundary layers (CTBLs) above regular, prism-shaped wall roughness and validated against direct numerical simulations at Mach numbers 2 and 4 (Cogo et al., 9 Jan 2026).

1. Theoretical Foundations: Generalized Reynolds Analogy

The GRA asserts a quasi-linear relationship between the mean defect of generalized enthalpy $\bar H_g$ and mean velocity $\bar u$ above the roughness sublayer, of the form:

$\bar H_g - \bar H_w = U_w\,\bar u$

where $U_w = -Pr\,q_w/\tau_w$ is a local velocity scale coupling wall heat flux $q_w$ and wall shear $\tau_w$, and $Pr$ is the Prandtl number. $\bar H_g = c_p T + (r_g/2)\,u^2$ is the generalized total enthalpy, with $c_p$ the specific heat, $T$ mean temperature, $u$ mean velocity, and $r_g$ an effective recovery factor. The analogy is exact for smooth-wall pipe, channel, and is approximately correct for boundary layers, now extended to arbitrary wall temperature conditions and independent of Prandtl number.

In rough-wall flows, GRA is strictly invalid within the roughness sublayer $0$k$), but becomes asymptotically valid in the outer layer $y>k$; this split is essential for roughness-resolving wall modeling (Cogo et al., 9 Jan 2026).

2. Multi-layer Decomposition and Roughness Parameterization

The boundary layer is decomposed as follows:

• Roughness sublayer ($0 Dominated by form drag and enhanced thermal mixing, causing breakdown of velocity-enthalpy similarity. Both mean velocity and enthalpy fields are discontinuous at $y=0$ and $y=k$, and cannot be predicted by GRA alone.
• Outer (smooth-like) layer ($y>k$): Exhibits similarity to smooth-wall flows after compressibility transformations (e.g., van Driest). GRA holds, with mean velocity profile given by a log law with modified origin and roughness length.

This distinction enables the wall model to match to a known or resolved velocity (and temperature) at $y=y_m>k$ and predict wall quantities through analytic or semi-analytic expressions, incorporating roughness via effective parameters (Cogo et al., 9 Jan 2026).

3. Governing Equations and Coupling Mechanisms

The core equations are:

• Wall shear and friction velocity:

$\tau_w = \rho_w u_\tau^2$

$C_f = 2\,\tau_w/(\rho_\infty u_\infty^2)$

• Stanton number:

$St = q_w / (\rho_\infty u_\infty c_p [T_r - T_w])$

• Generalized Reynolds analogy (GRA) coupling:

$\bar H_g - \bar H_w = U_w \,\bar u$

$U_w = -Pr\,q_w/\tau_w$

• Compressibility transformation (van Driest):

$$u_{VD}^+(y^+) = \int_0^{\bar u(y)/u_\tau} \sqrt{\frac{\bar\rho(\tilde y)}{\rho_w}}\, d(\frac{u}{u_\tau})$$

with the outer-layer log-law:

$$u_{VD}^+ = \frac{1}{\kappa}\ln\left(\frac{y-d}{z_0}\right) + B$$

where $\kappa$ is von Kármán’s constant, $d$ the virtual origin, and $z_0$ the roughness length.

• Roughness velocity shift:

$$\Delta U^+ = \frac{1}{\kappa}\ln\left(\frac{k-d}{z_0}\right)$$

$\Delta H^+ = - Pr\,\Delta U^+$

• Friction temperature at crest:

$$T_\tau = \left[\bar T_k - T_w + r_g\,\frac{\bar u_k^2}{2c_p}\right]\;\frac{\bar u_k}{u_\tau Pr}$$

These relations constitute a coupled system for $\tau_w$, $q_w$ for arbitrary wall temperature, outer boundary condition, Mach and Reynolds numbers, and roughness geometry (Cogo et al., 9 Jan 2026).

4. Algorithmic Implementation

The model advances in several steps:

1. Thermodynamic field initialization: Construct mean density $\bar\rho(y)$ via the equation of state assuming zero pressure gradient, from known or guessed values at outer boundary and wall.
2. Compute recovery factor $r_g$: Use boundary and edge layer values as described in (Cogo et al., 9 Jan 2026).
3. Set roughness shift: Prescribe crest velocity $\bar u_k/u_\tau$ via roughness parameters ($a, d, z_0$) from the drag model of Yang et al. (2016).
4. Initial guess and integration: Guess $u_\tau$ and integrate the van Driest log-law from $y=k$ to $y_m$, matching to resolved $\bar u_m$. Iterate until convergence.
5. Friction temperature and heat flux update: Calculate $T_\tau$ at $y=k$, update $q_w = \rho_w c_p u_\tau T_\tau$, and if necessary, recompute $U_w$ and $r_g$ to self-consistency.
6. Output wall fluxes: Deliver $\tau_w$ and $q_w$ to the outer solver (e.g., RANS or LES code).

Boundary conditions require no-slip/adhesion at the rough wall and prescribed $T_w$. Inputs are $M_\infty$, $Re_\tau$, $k$, and wall-to-recovery temperature ratio $\Theta$ (Cogo et al., 9 Jan 2026).

5. Model Validation and Predictive Performance

Validation against DNS data was performed for Mach 2 and 4, both adiabatic and cold walls, with prism roughness ($k^+\approx 55-57$, $Re_\tau\approx 1580$):

Mach/Case	$\epsilon_{\tau_w}$	$\epsilon_{q_w}$
M2A	+1.8%	0%
M2I	-0.7%	+5.5%
M4A	-10.8%	0%
M4I	-20.9%	-0.4%

Here, $$\epsilon_{\tau_w} = 100(\tau_{w,DNS}-\tau_{w,model})/\tau_{w,DNS}$$ and $\epsilon_{q_w}$ is the analogous heat-flux error. Heat-flux errors do not exceed 6% for isothermal walls. For Mach 4 cases, larger errors in drag are attributed to non-idealities in the van Driest transformation, but these remain an order of magnitude lower than for traditional correlations (Hill et al. 1980, $>200$% error). (Cogo et al., 9 Jan 2026)

6. Applicability, Limitations, and Extensions

Applicability:

• Predicts both wall shear and heat flux over high-speed, prismatically roughened walls in zero-pressure-gradient CTBLs.
• Enables coupling with high-fidelity outer-layer solvers (RANS/LES).

Limitations:

• Assumes van Driest compressibility transformation holds in the log layer above roughness, which may break down at strong wall cooling, extreme Mach number, or in the presence of strong nonequilibrium effects.
• Requires robust roughness drag model fit to geometry; extension beyond prism-shaped, regularly spaced elements would require recalibration.
• Assumes turbulent Prandtl number near unity; highly variable $Pr_e$ or nonequilibrium aerothermochemistry may not be captured.

Potential extensions include adaptation to other wall geometries, incorporation of pressure gradient effects, or coupling to advanced compressible wall laws (Cogo et al., 9 Jan 2026).

7. Context and Comparison with Alternative Wall Models

GRA-based wall models fundamentally differ from conventional finite-volume ODE and integral wall models by exploiting a physically motivated, outer-layer analogy and relating enthalpy and velocity defects directly via wall stress and heat flux, without relying on empirical correlation for the Nusselt or Stanton numbers. Compared to grid-free algebraic spectral wall models (Hayat et al., 2021), the GRA-based approach is specialized for compressible, rough-wall flows and explicitly incorporates surface geometry and thermal boundary conditions in a closed form. The core advantage is improved predictive capability, with accuracy at the few-percent level in tested regimes. For equilibrium, attached flows over rough surfaces and at high Reynolds and Mach, GRA-based wall models represent a rigorous framework bridging turbulence theory and heat transfer engineering (Cogo et al., 9 Jan 2026).