The Role of CFD in Modeling Multiphase Flows
Computational Fluid Dynamics (CFD) is one of the most powerful tools available to modern engineers, utilized in applications ranging from optimizing combustion chambers in jet engines to predicting weather patterns, and from analyzing aircraft wing icing to modeling drug delivery in human blood vessels. Systems where solid particles or liquid droplets are dispersed within a continuous fluid are known as "multiphase flows."
When setting up a CFD simulation for a multiphase flow, the most critical decision an analyst must make is how the software will track and compute the particles. Should the software track millions of individual particles along their unique trajectories (the Lagrangian approach), or should it treat the particles as a secondary, interpenetrating fluid (the Eulerian approach)? The mathematical compass guiding this crucial decision is the Stokes number ($Stk$). Before configuring your simulation models, you can quickly analyze your system parameters using our Stokes Number Calculator.
How Stokes Number Dictates CFD Method Selection
The Stokes number is defined as the ratio of the particle's relaxation time to a characteristic timescale of the flow. In CFD simulations, this characteristic flow time is usually defined either by a macroscopic dimension (like an obstacle size) or by the lifespan of turbulent eddies (the Kolmogorov timescale).
The appropriate mathematical modeling approach is strictly governed by the magnitude of the Stokes number:
- $Stk \ll 1$ (Very Low Stokes Number): Particles respond instantaneously to every fluctuation in the fluid flow. Under these conditions, Eulerian-Eulerian (or Mixture) models are highly recommended. Because the particles lack significant inertia independent of the fluid, there is no need to track them individually. They can be modeled as a dense gas mixed into the primary fluid, which saves an enormous amount of computational (CPU) time.
- $Stk \approx 1$ (Intermediate Stokes Number): Particles exhibit complex trajectories, partially following the fluid but frequently deviating due to their own inertia. Here, the Euler-Lagrange approach, commonly implemented as the Discrete Phase Model (DPM), is mandatory. The continuous fluid is solved on an Eulerian grid, while individual particles (or representative "parcels") are tracked using a Lagrangian method, calculating the forces and trajectory for each one.
- $Stk \gg 1$ (High Stokes Number): The particles possess massive inertia. Their motion is dictated largely by their initial kinetic energy, gravity, and collisions (particle-wall and particle-particle interactions) rather than aerodynamic drag. DPM is still used, but the solver settings must heavily prioritize contact physics and rebound models over fluid drag equations.
Turbulence Modulation and Particle Coupling in CFD
The Stokes number dictates not only the trajectory of the particles but also how those particles influence the fluid flow itself, a concept known as "coupling":
- One-way Coupling: When the mass loading of particles is very low, the fluid drags the particles, but the particles do not exert enough force to meaningfully alter the fluid's velocity field.
- Two-way Coupling (Turbulence Modulation): As particle concentration increases (especially when $Stk \sim 1$), the particles begin to affect the fluid turbulence. Large particles with high $Stk$ values generate wakes as they plow through the fluid, creating new eddies and increasing overall turbulence. Conversely, fine particles with low $Stk$ values absorb kinetic energy from the fluid eddies, acting as a viscous damper and suppressing turbulence. For CFD solvers to accurately predict this phenomenon, calculating the particle Stokes number relative to the turbulent eddy size is absolutely essential.
Realistic Example: Soot Simulation in an Exhaust Pipe
Scenario: You are tasked with simulating the flow of soot particles in a diesel engine's exhaust pipe using ANSYS Fluent or OpenFOAM. The exhaust gas velocity is high, and the soot particles are extremely fine. You need to decide which multiphase model to employ (DPM or a Mixture/Eulerian model) to balance accuracy with computational cost.
Data:
- Soot particle diameter ($d_p$): $2 \ \mu m \ (2 \times 10^{-6} \ m)$
- Soot density ($\rho_p$): $1800 \ kg/m^3$
- Exhaust gas velocity ($U$): $25 \ m/s$
- Exhaust pipe diameter ($L$ - characteristic length): $0.1 \ m$
- Exhaust gas viscosity ($\mu$): $3 \times 10^{-5} \ Pa\cdot s$ (higher than standard air due to high temperature)
- Cunningham correction ($C_c$): $\approx 1$
Calculation Steps:
- Particle Relaxation Time ($\tau_p$):
$\tau_p = \frac{\rho_p \cdot d_p^2 \cdot C_c}{18 \cdot \mu} = \frac{1800 \cdot (2 \times 10^{-6})^2 \cdot 1}{18 \cdot 3 \times 10^{-5}} = \frac{7.2 \times 10^{-9}}{5.4 \times 10^{-4}} \approx 1.33 \times 10^{-5} \ s$ - Flow Timescale ($\tau_f$):
$\tau_f = \frac{L}{U} = \frac{0.1}{25} = 0.004 \ s$ - Stokes Number ($Stk$):
$Stk = \frac{\tau_p}{\tau_f} = \frac{1.33 \times 10^{-5}}{0.004} \approx 0.0033$
Interpretation: The calculated Stokes number ($Stk = 0.0033$) is far below $0.1$. These soot particles have virtually no independent inertia; they are perfectly locked to the gas flow.
CFD Strategy: This result gives the CFD analyst a clear directive: "Do not waste computational resources tracking millions of individual particles with a Lagrangian DPM model." Because the soot behaves like a tracer gas, using an Eulerian scalar transport equation (or Mixture model) will provide an accurate solution. This decision can reduce simulation runtimes from weeks to mere hours. To explore at what particle size the $Stk$ approaches the critical threshold of $1$, you can experiment with our Stokes Number Calculator.
Important Warnings and Edge Cases in CFD
When interpreting the Stokes number for CFD setup, analysts must be wary of several pitfalls:
- Ambiguity in Timescale Selection: In the example above, we used the macroscopic pipe diameter ($0.1 \ m$). However, if the goal is to analyze "particle-turbulence interaction," using the macroscopic velocity is incorrect. One must calculate the "Turbulent Stokes Number ($St_{\eta}$)" using the Kolmogorov length and velocity scales. It is entirely possible for a system to have a macroscopic $Stk \ll 1$, but a microscopic (turbulent) $Stk > 1$.
- Mesh Resolution Limits: In Eulerian-Lagrangian (DPM) simulations, a fundamental assumption is that the particle (or parcel) volume is much smaller than the CFD grid cell volume. If the Stokes number is low, particles tend to cluster in specific flow regions (like recirculation zones). This artificial accumulation can lead to cell volume fractions exceeding limits, causing severe numerical instability (divergence) in the solver.
- Phase Change Dynamics: When modeling evaporating droplets (e.g., fuel sprays) or sublimating particles, the particle's mass ($\rho_p$) and diameter ($d_p$) shrink with every timestep. Consequently, the Stokes number is not a static property; it changes dynamically. The CFD solver must continuously re-evaluate the local Stokes number to update the drag laws appropriately.
