Shisha PM2.5 DPM Case Study -- Setup, Configuration, and Validation

#+CATEGORY: cfd

#+TAGS: DPM shisha PM2.5 case-study validation

#+MATURITY: growing

Shisha PM2.5 DPM Case Study :: Setup, Configuration, and Validation

This page is the working reference for a shisha (hookah) simulation using DPM to track PM2.5 smoke particles. The configuration is locked to the current project parameters below. All setup decisions trace back to the theory in DPM Theory and the setup procedure in DPM Setup Activation.

Locked Project Parameters

Why DPM Is the Correct Model

The shisha simulation has these properties: 1. Smoke particles are dilute. Volume fraction is $\sim 8 \times 10^{-6}$ (well below the 10% DPM limit). 2. Particles track the flow closely. The Stokes number for PM2.5 in air at 0.5 m/s is $\text{Stk} \ll 1$. 3. The particles are solid/aerosol (not evaporating liquid droplets). 4. Lagrangian tracking gives per-particle trajectories, which are needed for concentration mapping and deposition analysis.

DPM satisfies all four requirements. An Eulerian model would overpredict computational cost. A VOF model would be wrong (no sharp interface between smoke and air).

Particle Properties -- PM2.5 Smoke

Material definition (in Fluent Materials > Solid):

Size distribution:

PM2.5 is defined as particles with $d_p \le 2.5$ \mu m. Shisha smoke typically follows a log-normal distribution:

Match to experimental data: The user's PM2.5 measurement data should define the actual size distribution. Use histogram fitting to extract the geometric mean and standard deviation. Replace the values above with data-driven parameters.

Particle injection properties:

Physical Models -- Enabled Features

Brownian force UDF:

Fluent does not include a built-in Brownian force. Implement via DEFINE_DPM_BODY_FORCE:

#include "udf.h"
#include "dpm.h"

#define BOLTZMANN 1.3806488e-23  /* J/K */

DEFINE_DPM_BODY_FORCE(brownian_force, p, i)
{
  real T = C_T(P_CELL(p), P_CELL_THREAD(p));
  real mu = C_MU_EFF(P_CELL(p), P_CELL_THREAD(p));
  real dp = P_DIAM(p);
  real dt = CURRENT_TIMESTEP;
  
  real sigma = sqrt(2.0 * BOLTZMANN * T * mu * PI * dp / dt);
  real lambda = fabs(RAND_NUMBER);
  
  /* Return random Brownian force component in direction i */
  return sigma * lambda / (P_MASS(p) * PI / 6.0 * dp * dp * dp * P_RHO(p));
}

Compile and load the UDF before initializing.

Boundary Conditions

Steady-state note: With trap at walls, Fluent counts trapped particles. This gives the deposition map directly. Use Report > Discrete Phase > Summary to get global escape/trap counts.

Mesh Requirements

Stokes number check: For PM2.5 ($d_p = 2.5$ \mu m, $\rho_p = 1200$ kg/m^3) in air at 0.5 m/s:

• Particle relaxation time: $\tau_p \approx 2 \times 10^{-5}$ s
• Flow time scale: $\tau_f \approx L/U \approx 10^{-1}$ s
• Stokes number: $\text{Stk} = \tau_p / \tau_f \approx 2 \times 10^{-4} \ll 1$

Particles are tightly coupled to the flow. They follow streamlines closely. Fine mesh near walls is the key accuracy constraint, not the bulk mesh density.

Validation Strategy

The user has PM2.5 experimental data. The validation plan is:

Step 1: Grid convergence

1. Run on coarse (500K), medium (1M), fine (2M) meshes. 2. Compare DPM concentration at monitoring points. 3. Report percentage change between successive grids. 4. Select mesh where change is < 5%.

Step 2: Compare simulation vs experiment

Step 3: Sensitivity analysis

Post-Processing Plan

What to visualize and how:

See DPM Post-Processing Guide for full TUI and GUI commands.

Run Checklist

• [ ] Fluent version confirmed: 2024 R1
• [ ] 3D mesh imported and quality checked
• [ ] Material pm25-smoke created (rho=1200, cp=1000, k=0.1)
• [ ] Turbulence model selected (k-omega SST recommended)
• [ ] Energy equation enabled (if heated bowl)
• [ ] DPM model activated (steady tracking)
• [ ] Stochastic tracking enabled (10 tries)
• [ ] Brownian force UDF compiled and loaded
• [ ] Thermophoretic force enabled
• [ ] Injection created (surface, inert, pm25-smoke, log-normal)
• [ ] Boundary conditions set (walls=trap, outlet=escape)
• [ ] Continuous phase solved to convergence (500+ iterations)
• [ ] DPM iterated
• [ ] DPM summary reviewed (escape/trap counts)
• [ ] Particle tracks generated
• [ ] Concentration contours created
• [ ] Sample file written on breathing zone plane
• [ ] Results compared against experimental PM2.5 data
• [ ] Grid convergence verified (3 meshes)

See Also

• DPM Theory -- Governing Equations
• DPM Setup and Activation in Fluent
• DPM Post-Processing Guide
• DPM UDF Programming
• DPM Brownian Motion for Fine Particles
• DPM Physical Models -- Fluent Panel
• ANSYS Fluent Multiphase Modeling