Skip to main content

Turbulent plane channel flow with smooth walls

Overview

Flow physics:

  • Internal flow
  • Moderate Reynolds number
  • Spanwise direction: Statistically homogeneous
  • Streamwise and channel-height directions: Statistically developing
  • Newtonian, single-phase, incompressible, non-reacting

Solver:

Tutorial case:

Physics and Numerics

Physical domain:

  • The case is a statistically-developing internal flow through parallel smooth walls which are two characteristic-length apart.
    • xx: Longitudinal direction (Mean-flow direction)
    • yy: Vertical direction (Wall-normal direction)
    • zz: Spanwise direction (Statistically homogeneous direction)
    • OO: Origin at the left-bottom corner of the numerical domain

Physical modelling:

  • Reynolds number based on friction velocity: Reuτ=Uτδ/νfluid=395\text{Re}_{u_\tau} = \vert \mathbf{U}_\tau \vert \delta / \nu_\text{fluid} = 395 [-]
    • (Estimated) Friction velocity: Uτ=(1.0,0.0,0.0) \mathbf{U}_\tau = (1.0, 0.0, 0.0), and Uτ=uτ=1.0\vert \mathbf{U}_\tau \vert = u_\tau = 1.0 [m⋅s<sup>-1</sup>]
    • Characteristic length (Channel half-height): δ=1.0\delta = 1.0 [m]
    • Kinematic viscosity of fluid: νfluid0.002532\nu_{\text{fluid}} \approx 0.002532
\[m<sup>2</sup>⋅s<sup>-1</sup>\]
  • Bulk velocity of flow: Ub=(17.55,0.00,0.00) \mathbf{U}_b = (17.55, 0.00, 0.00)
\[m⋅s<sup>-1</sup>\]
  • Turbulence model: Large eddy simulation with the Smagorinsky sub-filter scale model utilising the van Driest wall-damping function. The sub-filter scale model constants:
    • $$ C_k \approx 0.02655 $$
    • $$ C_e = 1.048 $$
    • $$ C_s \approx 0.065 $$
- \$$ C_s = (C_k {C_k/C_e}^{0.5} )^{0.5} $$

Numerical domain modelling:

  • Shape: Rectangular prism
  • Dimensions: (x,y,z)=(20.0π,2.0,π) (x, y, z) = (20.0\pi, 2.0, \pi) [m]
  • Sketch:

Numerical domain (not in scale)

Spatial domain discretisation:

  • Mesh type: Rectangular cuboid mesh
  • Mesher: blockMesh
  • Number of nodes, NN: (Nx,Ny,Nz)=(500,46,82) (N_x, N_y, N_z) = (500, 46, 82) [nodes]
  • Spatial resolution (Δ)(\Delta) distribution:
    • Uniform in (x,z)(x, z)-directions
    • Stretched in (y)(y)-direction; clustered nearby walls
  • Uniform mesh particulars:
    • Δx+=(Δxuτ)/νfluid49.6 \Delta_x^+ = (\Delta_x u_\tau )/\nu_{\text{fluid}} \approx 49.6 [-]
    • Δz+15.1 \Delta_z^+ \approx 15.1 [-]
  • Wall-normal mesh particulars:
    • Simple grading expansion ratio: 25.0 [-] (From top to bottom wall, the ratio is 0.04)
    • First wall-normal node height: Δy+1.1\Delta_y^+ \approx 1.1
    • Mesh details:
Mesh (Front part)
Mesh

Temporal domain discretisation:

  • Time-step size: Δt=0.004 \Delta_t = 0.004 [s]
  • Estimated Courant-Friedrichs-Lewy (CFL) number based on uxy+=392=20.133 { \overline{u_x} }_{y^+ = 392} = 20.133[m⋅s<sup>-1</sup>]: CFL 0.64\approx 0.64

Equation discretisation:

Spatial derivatives and variables:

Temporal derivatives and variables:

Numerical boundary conditions:

  • Velocity, U\mathbf{U}
PatchConditionValue [m⋅s<sup>-1</sup>]
InletturbulentDFSEMInlet(17.55, 0.00, 0.00)
OutletinletOutlet(0.0, 0.0, 0.0)
Sides (zz-dir)cyclic-
Walls (yy-dir)fixedValue(0.0, 0.0, 0.0)
  • Pressure, p
PatchConditionValue [m<sup>2</sup>⋅s<sup>-2</sup>]
InletzeroGradient-
OutletfixedValue0.0
Sides (zz-dir)cyclic-
Walls (yy-dir)zeroGradient-
  • Turbulent kinematic viscosity, nut (i.e. νt\nu_t)
PatchConditionValue [m<sup>2</sup>⋅s<sup>-1</sup>]
Inletcalculated-
Outletcalculated-
Sides (zz-dir)cyclic-
Walls (yy-dir)zeroGradient-

Solution algorithms and solvers:

  • Pressure-velocity: PISO
  • Parallel decomposition of spatial domain and fields: scotch
  • The bandwidth of the coefficient matrix is minimised by renumberMesh
  • Linear solvers:
FieldLinear SolverSmootherRelative Tolerance
UsmoothGaussSeidel0.0
pGAMGGaussSeidel0.0
nuTildasmoothGaussSeidel0.0

Initialisation and sampling:

  • Computation time for a single domain pass-through based on Uxy+=392=20.133 { \overline{U_x} }_{y^+ = 392} = 20.133) [m<sup>2</sup>⋅s<sup>-1</sup>] 3.121\approx 3.121 [s]
  • Initialisation pass-throughs = 2.7\approx 2.7 58
  • Sampling pass-throughs = 24.5\approx 24.5 58

Results

List of metrics:

  • Prescribed vs. reproduced Reynolds stress tensor components at inlet patch
  • uu\overline{u^\prime u^\prime} downstream development vs. x/δx/ \delta
  • vv\overline{v^\prime v^\prime} downstream development vs. x/δx/ \delta
  • uv\overline{u^\prime v^\prime} downstream development vs. x/δx/ \delta
  • Surface skin friction coefficient Cf\mathrm{C}_f vs. x/δx/ \delta
  • Streamwise mean flow speed and Reynolds stress tensor components at uniform-interval downstream profiles
  • Streamwise vorticity ωx\omega_x at x/δ=0.1x/ \delta = 0.1
  • Streamwise vorticity ωx\omega_x at x/δ=1.0x/ \delta = 1.0
  • Metrics are time and spanwise averaged
  • <>< \cdot > is the time-averaging operator
Prescribed vs. reproduced Reynolds stress tensor at inlet patch (Poletto et al., Fig. 4)
<u'u'>-component of Reynolds stress tensor - Downstream development (Poletto et al., Fig. 14)
<v'v'>-component of Reynolds stress tensor - Downstream development (Poletto et al., Fig. 15)
<u'v'>-component of Reynolds stress tensor - Downstream development (Poletto et al., Fig. 13)
Longitudinal skin friction coefficient at top patch - Downstream development (Poletto et al., Fig. 9)
Longitudinal skin friction coefficient at bottom patch - Downstream development (Poletto et al., Fig. 9)
Streamwise vorticity component at y/δ=0.05 (Poletto et al., Fig. 11)
Streamwise vorticity component at y/δ=1.0 (Poletto et al., Fig. 12)

Resources

Note: Links will take you to the <i>University of Texas at Austin</i> website

Datasets for verifications (plain text)

Reynolds stress tensor profiles:

Mean velocity profiles:

Two-point velocity correlations (i.e. Auto- and cross-correlation functions):