18.

2 Release

Lecture 04: Natural Convection

ANSYS Fluent Heat Transfer Modeling

1 © 2017 ANSYS, Inc. December 28, 2017

 Outline

• Theory/Definition
• Phenomena
• Transition to turbulent flow
• Modeling tips
• Model setup in FLUENT
• Appendix

2 © 2017 ANSYS, Inc. December 28, 2017

 Outline

• Theory/Definition
• Phenomena
• Transition to turbulent flow
• Modeling tips
• Model setup in FLUENT
• Appendix

3 © 2017 ANSYS, Inc. December 28, 2017

 Phenomena

• In natural convection, fluid motion is P

P dx
generated due to density difference x
T  Tw
(buoyancy) in the fluid caused by
temperature gradients.  yx x
 yx 
 yx
dy
x

• Body forces y
T  Tf
• Typically gravitational P

• Centrifugal (rotating machinery)

• Coriolis (atmospheric and oceanic vortical motion)

• For this class of problems, flow and energy are

strongly coupled. Forces acting on a fluid particle in
natural convection.

4 © 2017 ANSYS, Inc. December 28, 2017

 Laminar to Turbulent Transition

• In natural convection, the Reynolds number no longer characterizes the

flow.

• With an appropriate reference velocity, it is possible to determine a

critical value of the Rayleigh number (RaL).
• Experiments show that the critical Rayleigh number, Rac, is around 109.
• The transition zone is quite large as Ra varies between 106 and 1010.

 g L3 T 
Ra L  GrL Pr  where Pr 
 

5 © 2017 ANSYS, Inc. December 28, 2017

 Boundary Layers

• Impact on numerical modeling for turbulent flow

Vertical
• Energy and momentum equations are strongly coupled. Wall
Thermal
• For high fidelity simulations intended to predict the local heat flux Boundary
distribution on surfaces with a high degree of accuracy, it is
T Layer
w
Free
recommended to construct the mesh such that y+ ≤ 1 in order to Stream
correctly resolve both the momentum AND thermal viscous sublayers.
L
– This is the same as forced convection T
• This is straightforward for Pr ~ 1 or Pr < 1. x
• When Pr > 1, the thermal sublayer is much thinner than the momentum
viscous sublayer.
• It is suggested that, provided that the momentum boundary layer
structure is explicitly resolved by the mesh (y+ ≤ 1 for the first cell layer
and at least 10 cells between 1 < y+ < 30), in many cases the thermal wall
treatment can predict heat flux and temperature profile without
explicitly resolving the thermal sublayer with the mesh (e.g. T+ ≤ 1)
– Probably a good idea to test this with a grid refinement study when starting
with a new problem
6 © 2017 ANSYS, Inc. December 28, 2017
 Boundary Layer Grid Generation

• Start with a 2D test case – This is a good way to confirm what are the characteristic
integration time steps and mesh size required for the desired physics.
• Also, sometimes it is not possible beforehand to estimate the physical distance corresponding to a given value
of y+ and if there is a reasonable 2D representation of the real geometry, this can save considerable time
before running the 3D case

Velocity field Temperature field

Initial temperature (K) Expected flow pattern during cool down in pipe
cross section due to buoyancy forces
7 © 2017 ANSYS, Inc. December 28, 2017
 Outline

• Theory/Definition
• Phenomena
• Transition to turbulent flow
• Modeling tips
• Model setup in FLUENT
• Appendix

8 © 2017 ANSYS, Inc. December 28, 2017

 Numerical Discretization

• Pressure interpolation scheme

• Recommended to use either Body Force Weighted or PRESTO!
• Standard pressure discretization can give rise to incorrect velocity near the wall.

Standard PRESTO! or Body Force Weighted

(Non-physical velocity near wall) (Correct near-wall velocity)

9 © 2017 ANSYS, Inc. December 28, 2017

 Unsteady Simulations

• Estimating the time step size for unsteady simulations:

• Estimate the time constant from
L L2 L
  
U  Ra Pr  g L T
• Use a time step of Δt = τ / 4.

• For conjugate heat transfer problems where you are only interested in
the steady solution, the density and heat capacity should be reduced (by
factor of 1000 for instance) for the solid material to neglect the thermal
inertia of the unsteady term in the energy equation.
• This applies in cases where the temperature in a solid is steady but the flow field is unsteady
in some parts of the domain

10 © 2017 ANSYS, Inc. December 28, 2017

 Operating Density

• Momentum equation along the

direction of gravity (z in this case)
W  P
    U W     2W  g
t z

• In FLUENT, a variable change is

done for the pressure field since
gravity is enabled.
P  P  0 g z

• Hydrostatic head is removed from pressure field

• Momentum equation becomes
W  P
    U W     2W     0  g
t z
where P' is the static pressure used by FLUENT for
boundary conditions and post-processing. This avoids
round off error and simplifies the setup of pressure
boundary conditions.

11 © 2017 ANSYS, Inc. December 28, 2017

 Operating Density

• Example – Pipe flow with water

Case 1

Static Pressure Velocity Magnitude

g The static pressure (P' = P – ogz) is

nearly uniform at the outlet.
Water density = 1000 kg/m³ Pressure outlet with constant gauge
Operating density = 1000 kg/m3 static pressure of 0 is correct.

Case 2

Velocity Magnitude
With o=0, P' = P, so the hydrostatic
g pressure variation is clearly visible in the
contours.
Water density = 1000 kg/m³ Pressure outlet with constant gauge static Incorrect pressure specification (because
pressure of 0 is no longer correct – here o = 0) leads to unphysical velocity
Operating density = 0 kg/m3 near outlet
remember the value entered in the
boundary conditions panel is treated as P'
12 © 2017 ANSYS, Inc. December 28, 2017
 Reference Density

• Example – Pipe flow with water

Case 3

Static Pressure Velocity Magnitude

g A hydrostatic pressure profile is

applied at the outlet
Water density = 1000 kg/m³
Same (correct) velocity field as in Case 1
Operating density = 0 kg/m3

A custom field function can be used to quickly define this

profile. More details are available in the appendix.
The purpose of this demo is not to suggest that the
operating density should be set to zero, but to illustrate
that if for some reason a value different from the density
of the fluid is used, care must be taken in setting pressure
boundaries. For this reason it is almost always better to
use the density of the fluid, as in Case 1 on the previous
slide.
13 © 2017 ANSYS, Inc. December 28, 2017
 Setting the Operating Density

• Example – Door and roof vents on a building with heated wall

• The roof static pressure is set to 0 while the door static pressure must be given a hydrostatic
head profile based on the height of the building.

Roof Outlet
Pressure outlet
So, the correct pressure BCs are:
Ps = 0 Pbuoy = ρo g H Pstop  0 g H
Psbot  0 g y  amb g H  y 
Or, equivalently,
g
H
y Pstop  0
Heated
wall
Psbot  amb  0  g H  y 
Note: In this case, if you can set the reference
Door Inlet
density equal to the external ambient density
Pressure inlet
then the hydrostatic component can be
Pbuoy = ρo g y
ignored:
Ps = ρamb g (h - y)

14 © 2017 ANSYS, Inc. December 28, 2017

 Example: Operating Density With Pressure Boundaries and
Natural Convection

wall, q"=0
Ambient at 20 °C (pressure outlet boundary condition type)

Air, incompressible ideal gas, ( = 1.204169 kg/m³ at 20 °C)

g Heated wall, T = 50 °C (all 4 sides)

wall, q"=0

15 © 2017 ANSYS, Inc. December 28, 2017

 Results at Two Different Operating Densities

Operating density corresponds to air at pressure outlet Wrong operating density (for pressure outlet
temperature (20 °C). Flow goes up at hot plate, entraining temperature) results in unphysical strong
some air from the sides. upward flow on cold pressure outlets.
16 © 2017 ANSYS, Inc. December 28, 2017
 Density Methods

• Fluid density can be set up

as a function of temperature
using a number of different
methods.
• Ideal gas
• Incompressible ideal gas
• Boussinesq
• Piecewise linear
• Piecewise polynomial
• Polynomial
• User-defined
• In FLUENT, the body force is
always calculated as
  0  g

• If density is constant, this term vanishes.

• In natural convection cases, this is the driving mechanism for fluid motion.
17 © 2017 ANSYS, Inc. December 28, 2017
 The Boussinesq Approximation

• Boussinesq model assumes the fluid density is constant in all terms of

the momentum equation except the body force term.
Constant (operating) density
0 W  P
 
   0 U W    W 2
   0  g
t z
  0  g  0  T  T0  g Variable (local) density

• In the body force term, the fluid density is linearized.

• For many natural convection problems, this treatment provides faster convergence than other
temperature-dependent density descriptions.
• The assumption of constant density in the other terms reduces nonlinear nature of the
governing equations.

• The Boussinesq approximation is valid when density variations are small.

• Cannot be used with species transport or reacting flows.
18 © 2017 ANSYS, Inc. December 28, 2017
 Boussinesq Setup

• Define density model

• In the Materials panel, select Boussinesq
as the density method and assign a
constant value, ρ0.
• Set the Thermal Expansion Coefficient, β.

• Set Operating Temperature, T0

Ideally the density and the thermal expansion coefficient
entered in the Materials panel are the values at T = T0 entered in
Operating Conditions
19 © 2017 ANSYS, Inc. December 28, 2017
 Boussinesq Density Model Example

• Hot cylinder in cool water:

• Open systems: T0  T and 0  
• Thermal expansion coefficient, β:
• For water at 8 °C, β = 0.0002 K-1 Velocity vectors on
temperature contour
• Density is around 1000 kg/m3
Pressure Outlet
g P' = 0 P  P0  0 g z
T = 8 °C
P  0 g H Hot wall, 25°C

Reference Temperature
Tref = 8°C
Ground (Adiabatic wall)
If you choose a different reference temperature, a pressure
profile needs to be specified at the boundaries or spurious
buoyancy induced flow will occur in the numerical solution
20 © 2017 ANSYS, Inc. December 28, 2017
 Incompressible Ideal Gas Setup

• Define density model

• In Materials panel, specify
incompressible-ideal-gas.
• Specify Operating Pressure
• Set the Operating density (if
desired). If not specified, FLUENT
will calculate ρ0 from a cell average
(default, every iteration).
P M  R = Universal gas constant (8314 J/kmol· K)
 0   op  M = Molecular weight
 R T average Pop = Operating pressure

21 © 2017 ANSYS, Inc. December 28, 2017

 Boussinesq vs. Incompressible Ideal Gas

• When the fluid is a gas, either the Boussinesq or the incompressible-

ideal-gas option can be selected.
• Boussinesq is not appropriate when the in-situ density variation implies a strong velocity
variation (to conserve momentum).
• Recommendation is to use Boussinesq when there is no more than 20% change in fluid density
throughout the domain.
• Example where Boussinesq is not valid – Hot wire at T = 540 K in ambient air at 270 K.

Incompressible ideal gas Boussinesq

22 © 2017 ANSYS, Inc. December 28, 2017

 Natural Convection in a Closed Domain

• For a closed domain, mass conservation depends on density treatment.

• This is automatically done when using the Boussinesq approximation (density is assumed to be
constant).
• Density only depends on temperature with incompressible ideal gas law since the operating
pressure is constant (mass is not conserved).
(  0 ) g  0  (T  T0 ) g

• In reality, pressure in the domain changes in such a way that mass is

conserved. Numerically,
• Density is a function of local pressure (ideal gas or real gas laws).

23 © 2017 ANSYS, Inc. December 28, 2017

 Natural Convection in a Closed Domain

• For steady solver

• Initial condition is merely a guess for the final solution.
• Final solution does not necessarily correspond to any particular initial mass.
• No EOS is imposed in order to conserve total mass
• Boussinesq model must be used to ensure mass conservation
• The constant density, 0 (meaning the value entered in the Materials Panel), properly specifies the mass
of the domain

• For unsteady solver

• Boussinesq model or ideal gas law can be used.
• Mass conservation not guaranteed with incompressible ideal gas
• Initial conditions prescribe the mass in the domain.

24 © 2017 ANSYS, Inc. December 28, 2017

 Buoyancy Effects on Turbulence
• Transport equations for RANS models include terms
representing production of turbulence due to buoyancy

  k    ui k    T  k 
        Gk  Gb     YM  S k
t  xi  xj  k   x j 
      ui     T     
      
 1C G  C G
3 b   C 2    YM  Sk
t  xi  xj k   x j 
k
 k k

• Depending on which model is selected and which options are

selected in the Viscous Models panel, these terms may either be
activated or deactivated

25 © 2017 ANSYS, Inc. December 28, 2017

 The Full Buoyancy Effects Option: -based models

• For the three k–ε models (SKE, RKE, RNG) and

for the epsilon based Reynolds stress model
(RSM):
• The production term in the k equation is always active
– Also true for the Reynolds stress equations when using
RSM

• The only way to deactivate it would be through a UDF

– But there is no reason to do so

• The production term in the  equation is active only if the Full

Buoyancy Effects option is selected in the panel

– This option often has no discernible effect on the results

26 © 2017 ANSYS, Inc. December 28, 2017

 The Full Buoyancy Effects Option: w-based models

• For the three k–w models (SKO, BSL, SST),

Transition SST, Intermittency Transition and for
the Stress-Omega or Stress-BSL RSM:
• The production term in the k equation is not included
• The production term in the w equation is also not included
• Beta feature options available to activate these terms

These options appear after activating beta

features; without beta features solver
uses No Buoyancy Effects on Turbulence
Buoyancy production only in k equation

Buoyancy production in k and w equations

27 © 2017 ANSYS, Inc. December 28, 2017
 Turbulence Generation Due to Buoyancy

• The importance of the buoyancy term (Gb) can be seen in a mixing layer
example using the standard k–ε turbulence model.

  k    ui k    T  k 
        Gk  Gb     YM  Sk
t  xi  xj  k 
 j 
x

UDF is used to set Sk = -Gb in the k equation

T = 100 °C resulting in zero net buoyancy effect Default treatment
g V = 1 m/s

T = 2 °C
V = 1.2 m/s
Gb is a sink term for stable stratification in this
configuration with warmer fluid above the cooler
fluid and therefore turbulent mixing decreases
compared to a non-buoyant shear layer

28 © 2017 ANSYS, Inc. December 28, 2017

 Appendix

29 © 2017 ANSYS, Inc. December 28, 2017

 Custom Field Functions to Apply Hydrostatic Pressure Profiles

1. Define Custom Field Function for hydrostatic pressure 2. Write profile of CFF on outlet

3. Hook profile in boundary conditions panel

30 © 2017 ANSYS, Inc. December 28, 2017

 Grashof Number

• Grashof number (GrL)

• Flow regime in natural convection is governed by Grashof number
• Indicates the relative importance of buoyancy forces to viscous (damping) forces

Gravitational acceleration Characteristic length

Coefficient of thermal expansion Maximum temperature differential
1   
  
  T  P  g L3 ΔT
GrL 
2
Kinematic viscosity

31 © 2017 ANSYS, Inc. December 28, 2017

 Richardson Number

• To determine whether the flow is driven by natural convection, forced

convection, or both, we examine the Richardson number.

• Richardson number (Ri) represents the relative magnitude of natural

convection effects to forced convection effects.
Grashof number

Gr  g L ΔT
Ri  2 
Re U 02
Reynolds number Bulk velocity

Ri = 1  Free and Forced convection effects must be considered

Ri << 1  Free convection effects may be neglected
Ri >> 1  Forced convection effects may be neglected

32 © 2017 ANSYS, Inc. December 28, 2017

 Rayleigh number

• Indicates the relative importance of buoyancy forces to viscous

dissipation and diffusion forces.

• Ra is the main characteristic of natural convection flows.

• Large Rayleigh numbers indicate strong natural convection effects.

• Product of Grashof number and Prandtl number

Buoyancy force
Kinematic
viscosity
 g L3 T 
Ra L  GrL Pr  wher Pr 
 e

Kinematic viscosity Thermal
Thermal diffusivity diffusivity

33 © 2017 ANSYS, Inc. December 28, 2017

 Transition Conduction / Convection

• In a heated, closed domain, a fluid particle will rise only if the buoyancy
force is more important than the viscous forces and the thermal
diffusion.

• There is a critical Rayleigh number from which the buoyancy force is

high enough for the particle to start moving.

• Experimentally, it can be shown that free convection settles when Ra <

Ra*.
• In a closed cavity, Ra* ~ 1700.
• If Ra < Ra*, then there is no motion and heat transfer occurs only due to conduction.

34 © 2017 ANSYS, Inc. December 28, 2017

 Transition Conduction / Convection

• The experiments described above is the Rayleigh

Bénard problem (deals with free convection). Cold

• Fluid heated at the bottom expands and rises due to buoyancy

forces. Hot

• At the top, fluid is cooling and then sinks.

• Some longitudinal rolling can be observed.

• For high viscous material (silicon oil), these rolls are parallel as in
the image below.

• If the top is a free surface, we can observe hexagons due to the

surface tension effects.

35 © 2017 ANSYS, Inc. December 28, 2017

 Free Convection Boundary Layers – Example

• Consider a vertical heated wall

(temperature Tw) in contact with a
cold fluid. The fluid temperature far
from the wall is T∞. Temperature

Temperature
Or Velocity
• Close to the plate, there is a Velocity
disrupted zone called a free
convection boundary layer. The Distance from Surface

thickness and heat flux in this layer Profiles of Velocity and Temperature
Perpendicular to a Heated Plate
varies with the vertical coordinate.

36 © 2017 ANSYS, Inc. December 28, 2017

 Reference Temperature and Expansion Coefficient

• In the Boussinesq approach, the reference temperature is specified

instead of the reference density.

• Reference density is implicitly specified in the Materials panel.

• The reference temperature should be selected based on this density, as

follows:
T T Tmax  Tmin
• For a closed system (cavity): T0  max min  0 
2 2

• For an open system: T0  T  0  

• Thermal expansion coefficient


Tmax  Tmin 
• In general, 0 Tmax  Tmin 
1
• For an ideal gas, 
T
37 © 2017 ANSYS, Inc. December 28, 2017