Mesh Generation and Mesh Independence Verification

In Computational Fluid Dynamics (CFD), the mesh serves as the foundational bridge connecting continuous physical domains

In Computational Fluid Dynamics (CFD), the mesh serves as the foundational bridge connecting continuous physical domains to discrete computational systems. By dividing the fluid volume into a finite number of control cells, meshing enables the transformation of governing partial differential equations into solvable algebraic systems. The quality of this discretization directly dictates numerical accuracy, convergence behavior, and overall computational expenditure.

Core tasks in mesh generation encompass:

  • Geometric Discretization: Deconstructing complex physical boundaries into manageable sub-domains.
  • Cell Distribution: Populating these sub-domains with structured or unstructured elements.
  • Quality Assurance: Enforcing strict criteria on element shape, sizing, and orthogonality to satisfy numerical solver requirements.
    Methodology Primary Application Key Characteristics
    Structured Grid Simple geometries, regular boundaries Ordered node indexing, high memory efficiency, fast solution speeds.
    Unstructured Grid Complex geometries, local refinement needs Utilizes triangles (2D) or tetrahedra (3D), offering superior geometric flexibility.
    Hybrid Grid Large-scale engineering, high-gradient zones Combines structured grids in far-fields with unstructured cells in critical local regions.
    Adaptive Mesh Refinement (AMR) Transient flows with rapidly shifting gradients Dynamically refines or coarsens mesh density based on real-time error estimation.

Structured Grid Workflow (Rectangular Domain Example)

  1. Coordinate Initialization: Define node counts $N_x$ and $N_y$ along the respective Cartesian axes.
  2. Node Generation:
    [
    x_i = x_{\min} + (i-1)\Delta x,\quad \Delta x = \frac{x_{\max}-x_{\min}}{N_x-1}
    ]
    Repeat identically for $y_j$.
  3. Cell Construction: Connect every adjacent group of four nodes to form rectangular control volumes.

Unstructured Grid Workflow (Triangular Mesh Example)

  1. Point Cloud Sampling: Scatter nodes along boundaries uniformly or proportionally to local curvature density.
  2. Delaunay Triangulation: Connect points while maximizing the minimum internal angles of each triangle.
  3. Mesh Smoothing: Apply Laplacian or optimization-based smoothing algorithms to enhance element aspect ratios.
  4. Local Refinement: Manually or automatically condense elements near walls, stagnation points, and separation zones.

Evaluating Mesh Quality

Reliable numerical solutions demand rigorous monitoring of specific geometric metrics:

  • Orthogonality: The angle formed by the face normal vector and the line connecting adjacent cell centroids. Values approaching 90° are optimal.
  • Aspect Ratio: The ratio of an element's longest side to its shortest side, recommended to remain $\le 5$ in 2D and $\le 10$ in 3D.
  • Minimum Internal Angle: For triangular and tetrahedral elements, angles should typically remain $\ge 20^\circ$ to prevent singular numerical behavior.
  • Smoothness: The spatial variation rate of adjacent cell dimensions ($\frac{\Delta h}{h}$), ideally kept below 0.2.

Tip: Modern CFD pre-processors (such as ANSYS Meshing or ICEM CFD) feature automated diagnostic tools to export comprehensive statistical distributions for these parameters.

Principles of Mesh Independence Verification

Mesh independence ensures that physical quantities of interest—such as drag coefficients ($C_D$), lift coefficients ($C_L$), or peak temperatures—reach an asymptotic plateau where further refinement yields negligible change. The standard verification protocol follows these steps:

  1. Establish a Series of Meshes: Generate 3 to 4 progressively finer iterations of the grid.
  2. Freeze Numerical Settings: Keep physical models, solver algorithms, discretization schemes, and convergence criteria strictly constant.
  3. Monitor Target Outputs: Record the variables of interest and plot them against characteristic mesh sizing metrics ($h$).
  4. Evaluate Convergence: Calculate relative percentage errors between successive grids. If the deviation falls below a predefined threshold ($\epsilon$, typically 1%), the solution is deemed grid-independent.

Quantitative convergence is evaluated using:
[
\text{Relative Error} = \frac{|Q_{i+1} - Q_i|}{Q_i} \times 100%
]
where $Q_i$ represents the output metric for the $i$-th mesh level.

Implementation Case Study: 2D Airfoil Drag Evaluation

Step 1: Geometry Setup
    - Import the NACA0012 airfoil profile with a chord length of L = 1 m.

Step 2: Discretization
    - Mesh A: Structured C-grid, first wall-normal layer height Δy⁺ ≈ 1, total elements ≈ 5×10⁴.
    - Mesh B: Same topology, total elements ≈ 1×10⁵ (doubled resolution).
    - Mesh C: Same topology, total elements ≈ 2×10⁵ (quadrupled resolution).

Step 3: Numerical Setup
    - Solver: Steady-state RANS, SST-k-ω turbulence model.
    - Boundary Conditions: Inlet velocity 30 m/s, pressure outlet, no-slip walls.
    - Convergence Criteria: Residuals below 1e-5, monitor CD until stabilized at 0.001.

Step 4: Execution & Data Collection
    - Mesh A: CD = 0.0254
    - Mesh B: CD = 0.0249
    - Mesh C: CD = 0.0247

Step 5: Error Analysis
    - A→B Relative Error = (0.0254 - 0.0249) / 0.0249 ≈ 2.0%
    - B→C Relative Error = (0.0249 - 0.0247) / 0.0247 ≈ 0.8%

Step 6: Final Verdict
    - Because the B→C variation drops beneath the 1% threshold, Mesh B (~1×10⁵ elements) satisfies grid independence criteria.

Practical Considerations and Common Pitfalls

  • Near-Wall Resolution: For high-Reynolds-number turbulence models, the height of the first boundary layer cell must strictly satisfy target $y^+$ constraints to prevent wall-function failure.
  • Targeted Refinement: Restrict high-density meshing to localized high-gradient regions (such as shockwaves, wakes, or separation zones) to avoid unnecessary computational bloat.
  • Numerical Diffusion: Abrupt volume transitions between adjacent cells induce truncation errors; smoothing algorithms are vital to preserve accuracy.
  • Physics vs. Discretization: Mesh independence only eliminates numerical discretization errors; foundational errors stemming from improper turbulence modeling or poor boundary definitions will persist.
  • Automation: Utilizing batch scripting tools within frameworks like OpenFOAM (blockMesh, snappyHexMesh paired with Bash scripts) significantly streamlines multi-mesh generation workflows.

Summary

Mesh generation forms the bedrock of reliable CFD workflows. Meticulous spatial discretization paired with rigorous mesh independence verification ensures that numerical predictions are both accurate and computationally efficient. By systematically balancing element quality, sizing, and resolution verification, engineers can generate high-fidelity simulations that provide a solid analytical foundation for downstream engineering design and optimization.