Best Practices for Compressible Flows

Using ANSYS CFD

1 © 2015 ANSYS, Inc. June 12, 2018

 Outline
• Introduction to Compressible Flows
• Modeling Compressible Flows in ANSYS FLUENT
• Solver algorithms (Pressure-based, Density-based)
• Steady-state solutions with the pressure-based
coupled solver
• Transient solutions with pressure-based solver
• Steady-state solutions with density-based implicit
solver
• Transient solutions with density-based explicit
solver Supersonic exhaust flow from a rocket engine.

• Summary
https://en.wikipedia.org/wiki/Draco_(rocket_engine_family)

• Q&A

2 © 2015 ANSYS, Inc. June 12, 2018

 Compressible Flows
• What distinguishes compressible flows from incompressible?

• Fluid density is no longer constant but varies throughout the fluid

• Density is usually related to pressure and temperature through an Equation of State (e.g. ideal
gas law)
𝜌 = 𝑓(𝑃, 𝑇)

• Because of the link between density, pressure, and temperature, we now need to solve the
Conservation of Energy Equation along with the Conservation of Mass and Momentum
equations.
– Conservation of Energy Equation = First Law of Thermodynamics

 ( E )
t
 
+   V( E + P ) =   keff T −  j h j J j + ( τ eff  V) + S h

3 © 2015 ANSYS, Inc. June 12, 2018

 Mach Number
• For a compressible fluid, pressure waves propagate at a finite speed (speed of sound).
• Speed of sound (c) for an ideal gas is related to the temperature by
𝑐= 𝛾𝑅𝑇 )
• The ratio of the fluid velocity (V) to the speed of sound is called the Mach Number (M)
𝑀 = 𝑉/𝑐
• As the Mach number increases, we characterize the flow as follows:

• Subsonic, incompressible flow (M < 0.3)

• Subsonic, compressible flow (0.3 < M < 1.0)
• Transonic, compressible flow (M ~ 1.0)
• Supersonic, compressible flow (M > 1.0)
• Hypersonic, compressible flow (M > 4.0)

4 © 2015 ANSYS, Inc. June 12, 2018

 Compressible Flow - Physics
• Unsteady flows (wave propagation, acoustics)
• Viscous effects (separation, wakes, jets, shock-boundary layer interaction)
• Turbulence modeling
• Steady-State – RANS and transition Modeling
• Unsteady – Scale-Resolving Simulation
• Compressibility effects on turbulence

• Fluid properties
• Calorically-perfect gas (ideal gas law)
– Specific heats are constant
• Thermally-perfect gas (ideal gas law)
– Specific heats are functions of temperature only
• Transport properties (viscosity, thermal conductivity) are usually considered constant or
functions of temperature only. Can also be functions of composition for chemically
reacting flow.
– Example: Sutherland’s Law for viscosity of high temperature air.
• Other physics not consider here
• Real gas effects (non-ideal gas equations of state)
• Rarefied gas flows (e.g. slip BCs)
• High temperature effects and heat transfer (e.g. radiation, ionization)
– Properties function of chemical composition of fluid (pressure and temperature)
– Primarily a consideration for hypersonic flows

5 © 2015 ANSYS, Inc. June 12, 2018

 ANSYS CFD Solvers for Modeling Compressible Flows
Subsonic Transonic Supersonic/Hypersonic

ANSYS CFD solvers provide robust and efficient approaches for modeling
compressible flows through the entire compressible flow spectrum from low
subsonic to hypersonic re-entry problems

6 © 2015 ANSYS, Inc. June 12, 2018

 Computational Domain

• For compressible flows with shock waves, always try to shape the domain such that
shocks pass through an outlet and do not strike an inlet
− Prevents spurious reflections at an inlet boundary

outlet outlet

Not Recommended Good

7 © 2015 ANSYS, Inc. June 12, 2018

 Meshing
• All CFD good meshing practices should be adhered to for compressible flows!
✓ Resolve high gradient regions of the flow with smooth, high quality meshes
• For flows with shock waves, the mesh should be smooth along the shock direction, avoiding
jumps in cell size
✓ Mesh growth rates above 20% should be avoided

Not Recommended Acceptable

8 © 2015 ANSYS, Inc. June 12, 2018

 Mesh Adaption
• Fluent’s mesh adaption is a very effective tool for resolving shock waves, expansion fans and related
compressible flow phenomena!
• Can be applied to steady state and unsteady flows (dynamic adaption)

Unadapted baseline mesh (supersonic diffuser) Adapted mesh (supersonic diffuser)

9 © 2015 ANSYS, Inc. June 12, 2018

 CFD Solver Algorithms for Compressible Flows
Pressure-based (PB) methods have evolved from
Pressure-Based Solvers (PBS) solution techniques developed for low speed flows.
FLUENT’s PB solvers have been gradually enhanced
• Segregated methods (SIMPLE, to permit transonic and super sonic flows.
SIMPLEC, PISO)
• Coupled solver (continuity and ✓ Use PB solvers for lower speed compressible
flows and steady flows with embedded
momentum) transonic/supersonic regions and internal flows.

Density-Based Solvers (DBS) Density-based methods evolved from solution

techniques developed for high speed aerodynamics.
• Implicit algorithm FLUENT’s DBNS solvers have been gradually
enhanced to permit low speed to incompressible
• Explicit algorithm Mach numbers.

✓ Use DB solvers for high speed steady and

unsteady compressible flows, especially those
involving strong shock waves.

10 © 2015 ANSYS, Inc. June 12, 2018

 Pressure Based Solver Algorithms
Segregated Coupled

• Two different PB algorithms are available Solve U-Momentum

with the pressure based solver
Solve V-Momentum
Solve Mass
Solve W-Momentum & Momentum
– Pressure Based Segregated Solver
– Pressure Based Coupled Solver (PBCS) Solve Mass
Continuity;
Update Velocity

• Pressure Based Segregated Solver solves Preferred solver for

individual governing equations sequentially unsteady compressible Solve Energy
(loosely coupled) flows (e.g. LES, acoustics)
Solve
Species

• Pressure Based Coupled Solver solves Solve Turbulence

Preferred solver for Equation(s)
momentum and pressure based continuity steady-state compressible
equations in a coupled manner (energy Flows (RANS) Solve Other Transport Equations as
equation is not coupled) required

11 © 2015 ANSYS, Inc. June 12, 2018

 ANSYS FLUENT - Pressure-Based Coupled Solver (PBCS)

The PBCS offers a robust and efficient methodology for solving compressible
flows ranging from low subsonic to supersonic
• Elliptic nature of pressure-velocity coupling algorithm is very efficient for low speed, subsonic
compressible flows
• Algorithm can also handle supersonic flow regions, and is ideal for mixed
subsonic/supersonic flows.
• Fully implicit coupling between momentum and continuity equations provide improved
robustness versus Segregated approaches
▪ Implicit discretization of pressure gradient terms in momentum equations
▪ Implicit discretization of face mass fluxes including Rhie-Chow pressure dissipation terms
• Coupled algebraic multigrid (AMG) scheme is used to solve the system of discretized
algebraic equations

12 © 2015 ANSYS, Inc. June 12, 2018

 PBCS: General Settings
• Double precision FLUENT
• Green-Gauss node-based or Least-squares Cell-
based calculation of gradients
• Second Order scheme for pressure
• Second Order Upwind discretization for
convection terms
• 1st-order upwind scheme can be used during the
initial iteration stage
• 1st-to-higher order blending can be applied to the
2nd-order upwind discretization if the 2nd-order
solution becomes numerically stiff. This is
accomplished by a TUI command:
solve/set/numerics
and set 1st-to-higher order blending factor to a value
below 1.0
• Pseudo-Transient Option can be advantageous
for steady-state solutions (see later slides)

13 © 2015 ANSYS, Inc. June 12, 2018

 PBCS: Advanced MG Solution Controls
• In general, you should use default MG control
settings for most problems.
• However, if the convergence becomes stiff,
you can increase Pre-Sweeps to 1 and Post-
Sweeps to 4 under Coupled Parameters
• This permits the AMG solver to work harder
within each iteration
• Stabilization algorithms are also available for
improving linear equation convergence
• If you suspect a problem with MG
convergence, set Verbosity = 1 to help
diagnose the issue (e.g. which equations are
not converging)

14 © 2015 ANSYS, Inc. June 12, 2018

 PBCS: Steady-State Solution Controls
• If you are NOT using the pseudo-transient
solver, the following are recommended solution
controls for the PBCS
• Courant Number between 10 (hypersonic) and
200 (low subsonic)
– Courant number is a global relaxation
parameter
– Unlike the density-based solver, low values of
CFL in PBCS can lead to instability, and CFL
should be greater than 1.0
• Explicit relaxations for pressure and momentum:
– 0.75 – low subsonic flows
– 0.50 – transonic flows
– 0.25 – supersonic and hypersonic flows
(lower values in hypersonic regime)
• Under-relaxation factor (URF) for density is the
same as relaxations for pressure & momentum
• All other URFs at default values

15 © 2015 ANSYS, Inc. June 12, 2018

 PBCS: Improved Convergence with Pseudo Transient Under-relaxation

• Traditional steady-state PBCS

• Under-relax the transport equations by applying under-relaxation factor to the diagonal coefficient.
Introduces a CFL-type variable

1− 1−

( )
a p  pn +1 −  pn + a p pn +1 −  aiin +1 = S p
1
CFL
=


• Pseudo-transient (PT) method

• Under-relax the transport equations by adding pseudo-transient term. Depends on global scales rather than
local scales and therefore often converges better on anisotropic meshes

 pVol p
( n +1
p −  pn )+a  n +1
−  ai n +1
= Sp t =
length _ scale
t
p p i velocity _ scale

16 © 2015 ANSYS, Inc. June 12, 2018

 PBCS: Pseudo Transient Settings for PBCS

• Recommended PT settings for

external aero problems:
‒ Time Step Method: Automatic,
Timescale Factor = 1
‒ Length Scale Method: User-
specified
‒ Length scale: Mean MAC
Aerodynamic Cord (MAC)
‒ Gradient: Least Squares or
Green-Gauss Node Based
Use
gradients Automatic/User
‒ For skewed meshes, PT Explicit Specified
Relaxation factor values may Length Scale
Method for
need to be reduced external flows

17 © 2015 ANSYS, Inc. June 12, 2018

 High Order Term Relaxation (HOTR)
• What is HOTR?
• HOTR is an under-relaxation which is applied to
higher-order discretization terms (pressure-based
and density-based solvers)
• What does it do:
• Improves solution convergence behavior for
second order discretizations
• Helps prevent apparent convergence stalling
• Improves solution startup robustness
• How to use it:
• Activate it from the Solution Methods task page
• Extra Options are available

18 © 2015 ANSYS, Inc. June 12, 2018

 HOTR Example: Supersonic Jet Impingement

Convergence stalls without HOTR

With HOTR Without HOTR

19 © 2015 ANSYS, Inc. June 12, 2018

 Turbulence Modeling 10.000
DDES fine - LLG

1.000 UF experiment

• Wide range of turbulence models are available for PBS and DBS, and all

prms2/St
0.100
are applicable to compressible flows.
• Steady-State Models 0.010

– Generally use one or two-equation RANS models 0.001

• Spalart-Allmaras 1.00 St 10.00
• RNG and realizable k-e (RKE)
• Standard and SST k-w
• Transitional four-equation SST k-w, and three-equation k-kl-w
– Reynolds Stress Model (RSM) is also available (less often used)
• Unsteady Models
– Scale-Resolving Simulation (SRS) Methods (LES, DES, DDES, SAS, ELES)
• The choice of a turbulence model depends on your simulation objectives
✓ RANS models are normally used for steady-state problems
✓ Unsteady problems may use unsteady RANS (URANS) or an SRS-based model
(SRS for acoustics and complex turbulent mixing) SRS model of landing gear acoustics

20 © 2015 ANSYS, Inc. June 12, 2018

 Compressibility Effects Option for RANS Turbulence Models

• “Compressibility Effects” option is available for both k-e and k-w

families of RANS turbulence models
– Models effect of compressibility on turbulence kinetic energy
– Accounts for “dilatation dissipation” which is normally neglected for
low speed flows
– Available with ideal-gas option only and is turned off by default
– Improves predictions for high-Mach number free shear flows by
reducing turbulence spreading rates

Dk    t  k   t  k
 = 
  +   +  S 2
− e − g − 2 e
Dt x j 
  k  x j  t i
 Pr t x i  RT

Turbulence kinetic energy equation for k-e model

Added dissipation term, prevents overprediction of spreading rate in compressible flows

21 © 2015 ANSYS, Inc. June 12, 2018

 Boundary Conditions For Compressible Flows
Proper flow Boundary Conditions (BCs) are essential for compressible flows!

• Subsonic Inlet, Subsonic Outlet

– Total conditions upstream and static pressure downstream
– Static pressure can propagate upstream
– Transonic pockets can exist within domain (including choked conditions!)
• Supersonic Inlet, Supersonic Outlet
– All flow properties required upstream – no BCs required downstream!
– Static pressure cannot propagate upstream due to supersonic flow
• Subsonic Inlet, Supersonic Outlet
– Usually seen in choked internal flows (rocket nozzles)
– Flow chokes at throat and expands to supersonic Mach numbers
• Supersonic Inlet, Subsonic Outlet
– Flow Mach number reduces in a converging duct through a series of oblique shock waves, J-2X rocket motor
https://en.wikipedia.org/wiki/J-2X
terminating in a normal shock
• Farfield Boundary Conditions
– Mach number, static pressure and temperature defined at a farfield boundary
– Typically used for vehicle flow at a fixed speed (e.g. aircraft and missiles in flight)

22 © 2015 ANSYS, Inc. June 12, 2018

 Isentropic Relations for Inlet BCs
• Inlet BCs are determined using isentropic relations
• For ideal gases:
𝑃 = 𝜌𝑅𝑇 Eq. (1)

𝑇0 𝛾−1 2 Eq. (2)

= 1+ 𝑀 SUBSONIC INLET: Total pressure and total temperature inputs required.
𝑇 𝛾 Static pressure obtained from CFD solution.
𝛾
𝑃0 𝛾−1 2 𝛾−1
Eq. (3) SUPERSONIC INLET: Total pressure and total temperature inputs required.
= 1+ 𝑀
𝑃 𝛾 Static pressure must ALSO be input! Use isentropic relation
to compute static pressure for the desired Mach number.

BASIC INLET CALCULATION

1. Knowing total pressure and static pressure, compute Mach number from Eq. (3)
2. Compute static temperature from Eq. (2)
3. Compute static density from ideal gas law, Eq. (1)
4. Other flow properties (speed of sound, velocity magnitude) can be computed from other
compressible flow relations.

23 © 2015 ANSYS, Inc. June 12, 2018

 Setting Flow BCs – Subsonic Pressure Inlet

Inlet Total Temperature

Inlet Total Pressure

This input is not used during computation!

Flow direction vector

Inlet turbulence BCs

✓ Inlet static pressure obtained from CFD solution

24 © 2015 ANSYS, Inc. June 12, 2018

 Setting Flow BCs – Supersonic Pressure Inlet

Inlet Total Temperature

Inlet Total Pressure

This required input! Set such that total/static
pressure ratio provides desired Mach number

Flow direction vector

Inlet turbulence BCs

𝛾−1
2 𝑃0 𝛾
𝑀2 = −1
𝛾−1 𝑃

✓ Appropriate inlet static pressure is a required input.

25 © 2015 ANSYS, Inc. June 12, 2018

 Setting Flow BCs – Pressure Outlet

Outlet Static Pressure

This input is not used for supersonic flow at outlet*

Reverse flow BCs

* For mixed flow (subsonic + supersonic) the pressure

will be used for the subsonic portion of the boundary

26 © 2015 ANSYS, Inc. June 12, 2018

 Mixed Subsonic-Supersonic Exit Flow
Contours of Mach Number
• There may be both subsonic and
supersonic velocities at an outlet supersonic
boundary flow outside
boundary
− For example, in a near-wall viscous layer
boundary layer with supersonic external
flow
• In this case, the static pressure prescribed
at the outlet will be used to define the
subsonic part of the boundary subsonic
boundary
• Sometimes reverse flow may be induced layer
in the subsonic region, which may lead to
instability wall

− Good initial conditions can help prevent

this from occurring outlet

− If reversed flow persists, try extending

the exit boundary position

27 © 2015 ANSYS, Inc. June 12, 2018

 Subsonic Inlet Example - Laval Nozzle
Mach number
contours

Pexit/Ptin = 0.95 (subsonic) Pexit/Ptin = 0.8 (choked, normal shock) Pressure inlet –
Pressure outlet
BCs

Pexit/Ptin = 0.4 (choked, oblique shocks) Pexit/Ptin = 0.12 (supersonic exit)

28 © 2015 ANSYS, Inc. June 12, 2018

 Supersonic Inlet Example – Supersonic Diffuser

Mach number
contours

Mach 3 inlet flow prescribed

using appropriate pressure ratio

Pressure inlet –
Pressure outlet
BCs

29 © 2015 ANSYS, Inc. June 12, 2018

 Pressure Far-Field Boundary Condition
Far-field boundary
• Set freestream P, T, M, flow M = 1.6

direction, turbulence, other scalars

• Far-field boundary should be at least
10L or greater away from body (body
length scale = L)

Mach number
contours

30 © 2015 ANSYS, Inc. June 12, 2018

 Other Flow Boundary Conditions for Compressible Flows

• Velocity Inlets, Mass Flow Inlets

− Both can be used with compressible flows
− Be cautious if you are using these with choked
internal flow or supersonic inlets (may be less
stable than a pressure inlet)
− Best used for mostly subsonic flows
• Non-reflecting and Impedance BCs
− Very useful for unsteady acoustics
− Density-based solver
• NRBCs are applied only on pressure-outlet
BCs only.
− Pressure-Based Solver
• NRBCs can be used with pressure-inlet,
pressure-outlet, velocity-inlet and mass-
flow-inlet BCs

31 © 2015 ANSYS, Inc. June 12, 2018

 Large Pressure Ratio Internal Flows inlet

• Problems with very large differences in boundary pressures

can be challenging to converge
• A straightforward approach of dealing with these
problems: outflow

✓ Increase the exit pressure (to near the inlet total Ball diverter valve example (pressure ratio = 500 psi)
pressure) to achieve subsonic low speed flow initially -
use stagnation conditions to initialize
✓ Gradually decrease the exit pressure in stages,
resolving any supersonic (choked) flow as you go with
adapted meshing
✓ Adjust solution controls if needed and be aware of any
potential for unsteady flow (e.g. shock-boundary layer
separation)

Mach number

32 © 2015 ANSYS, Inc. June 12, 2018

 Compressible Flow Solution Initialization
• Solution initialization can often be crucial for obtaining stable solutions, especially for high speed
compressible flows!
• When possible, use Full Multigrid (FMG) initialization to obtain the initial solution (otherwise hybrid).
• FMG initialization is launched by TUI command:
solve/initialize/fmg-initialization
• For supersonic and hypersonic flows, it is recommended to reduce FMG Courant number from 0.75 to
0.25:
solve/initialize/set-fmg-initialization

Initial solution after FMGI Final converged solution

33 © 2015 ANSYS, Inc. June 12, 2018
 Solution Bounding
• Proper limiting of pressure & temperature
fields will help in preventing solution
divergence
• Positivity Rate Limit is an auto-localized explicit
relaxation
• If the resultant temperature at a given cell
changed by more than a “specified
percentage” (20% i.e. PRL=0.2 by default)
then the solution vector for that cell will be
under-relaxed by a factor:

 Tcell 
q new = q old +  pl q with  pl = PL.  if Tcell  PL.Tcell
 Tcell 

• For very high speed flows, you may need to reduce the positivity limits to 0.05
or even 0.02 to prevent solution from diverging
34 © 2015 ANSYS, Inc. June 12, 2018
 Pressure-Based Solvers and Transient Flows
• Transient problems frequently arise for compressible flows
− Acoustics
− Unsteady separated flows and wakes
− Moving shock waves
− Forced unsteadiness due to transient boundary conditions
• For lower speed compressible flows, you can typically use either the Pressure-
based Segregated Solver or the Non-Iterative Time Advancement (NITA)
approach
− Why? For sufficiently small time steps, you don’t need added expense of equation coupling
− Segregated solvers (SIMPLEC, PISO) and NITA are much more CPU and memory efficient than
the Coupled solver
− HOWEVER – for high speed flows, where the equation coupling becomes more important,
the PBCS can be used effectively for transient flows

35 © 2015 ANSYS, Inc. June 12, 2018

 Transient Settings

• Recommended PBCS Solver Controls

• Set Courant Number to a very large
• Smaller Courant numbers can be used if solution
becomes stiff or diverges
• All transport equation relaxation factors are 1.0
– Stability is controlled by the physical time step
• Second-order implicit time marching scheme
• Time step size is based on smallest resolved
time scale
• About 20 time steps per smallest time scale
• 10 - 20 sub-iterations per time step

36 © 2015 ANSYS, Inc. June 12, 2018

 Transient Compressible Flow Example – Supersonic Screech

Pressure field showing predominantly upstream radiation of screech tone.

Mj = 1.2, nozzle lip thickness = 0.625”

37 © 2015 ANSYS, Inc. June 12, 2018

 Supersonic Screech Example

1.4
test
1.3 CFD
1.2

ps / pambient
1.1

1.0

0.9

0.8

0.7

0.6
0 1 2 3 4 5 6 7 8 9
x/D

Comparison of calculated and experimental time-averaged centerline pressure

Instantaneous density field near nozzle exit
showing unsteady shock cell structure of
the jet
Mj = 1.2, nozzle lip thickness = 0.625”

38 © 2015 ANSYS, Inc. June 12, 2018

 Supersonic Screech Example – Acoustic Signals

150 Mj = 1.2 A1 mode 150 Mj = 1.22 A1 mode

microphone at r = 0.662" microphone at r = 0.662"
140 microphone at r = 0.889" 140
microphone at r = 0.889"
A2 mode
130 130
SPL, dB

SPL, dB
120 120

110 110

100 100

90 90
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
f, kHz f, kHz

Calculated noise spectra at two microphones on the nozzle lip located at r = 0.642 and 0.889 inches

39 © 2015 ANSYS, Inc. June 12, 2018

 Density-Based Solver (DBS)
Density-based Coupled
Density-Based Solver algorithm
• The governing equations of continuity, momentum, and
(where appropriate) energy and species transport are solved Solve Mass,
simultaneously (i.e., coupled together) Momentum,
& Energy

• Additional scalar equations are solved in a segregated fashion

• The density-based solver can be run implicit or explicit
• Implicit Solver – Typically preferred for steady-state DBS Solve Energy
solutions
Solve Turbulence

• Explicit Solver – Used for transient problems such as moving Equation(s)

shock capturing
Solve Other Transport Equations as
required

40 © 2015 ANSYS, Inc. June 12, 2018

 DBS – General Settings
• Double precision Fluent
• Green-Gauss node-based calculation of gradients
• Roe-FDS flux for subsonic and transonic flows
• AUSM+ flux for supersonic and hypersonic flows
• 2nd-order upwind scheme
• 1st-order upwind can be used during initial iterations
• 1st-to-higher order blending can be applied if the 2nd-order
solution becomes numerically stiff. This accomplished by a
TUI command:
solve/set/numerics
and set 1st-to-higher order blending factor to a value < 1.0
• Reduce explicit under-relaxation to a value from 0.25 to
0.15 for supersonic and hypersonic flows using a TUI
command
solve/set/expert

41 © 2015 ANSYS, Inc. June 12, 2018

 AUSM+ Flux Scheme

• When both High mach and very low mach number zone are
• Quad. Mesh of 8.6K cells
present, default Roe Scheme exhibits flaw known as the • Inviscid solution
carbuncle effect. The AUSM+ flux scheme is needed for low- • M=30
Mach number preconditioning which corrects this issue.

Example: High-speed flow over 2D-cylinder

Roe-FDS AUSM+

42 © 2015 ANSYS, Inc. June 12, 2018

 DBS - Pseudo Transient Under-relaxation
• Standard discretization already has a
time derivative term
• Strategy of convergence is different
between the two methods
• CFL value is specified in Standard
method
▪ Time step is computed for each cell and
solution is advanced at the local time step
▪ For anisotropic meshes convergence may MAC
slow down due to the very small local time
steps on high aspect ratio cells
• Pseudo time is specified In Pseudo
Transient method which effectively vary
the CFL value spatially and make the
implicit solver converge at optimal
settings

43 © 2015 ANSYS, Inc. June 12, 2018

 DBS – Solution Controls for Steady-State

• Use the largest Courant Number

allowed by solution stability
• Smaller CFLs at initial iteration stage
• Solution Controls – Solution Limits
• Sometimes, the initial conditions force
the provisional solution to predict
unphysical pressures and
temperatures e.g. larger than inlet
total values.
• Set sensible limits for pressure and
temperature.
• Reduce Positivity Rate Limit to ~ 0.05
for supersonic and hypersonic flows

44 © 2015 ANSYS, Inc. June 12, 2018

 Gradient Limiters
Gradient limiters (or slope limiters) are used on the higher order
discretization schemes to prevent spurious oscillations
• Standard Limiter (Cubic MinMod)
– Uses Minimum Modulus (MinMod function) to limit
and clip reconstructed solution
• Multi-dimensional limiter
– Similar to standard limiter
– Clips only the normal components of gradient to the cell faces
– Less dissipative than standard limiter
• Differentiable limiter
– Based on the work of Z.J. Wang (Int J. num Mech. Fluids 2000). It is a
Modified form of Venkatakrishnan limiter (AIAA-93-0880)
– Overcomes the disadvantage of non-differential limiter which tend to
stall residual convergence
– Uses a smooth function to impose monotonicity condition while
allowing residuals to converge

45 © 2015 ANSYS, Inc. June 12, 2018

 Gradient Limiters Example: Transonic flow over RAE2822 Airfoil

Standard Limiter Differentiable Limiter

46 © 2015 ANSYS, Inc. June 12, 2018

 Convergence Aiding
1st-to-2nd Order Blending Tools
Density Based Solver
• 1st-to-2nd order blending is a parametric variation of the reconstruction
gradients
RG
 f = C 0 +   C 0 • dr0

• A blending factor (β) of 0 reduces the reconstruction scheme to 1st-order

while blending factor of 1 will recover the 2nd-order reconstruction scheme
• In the DBS, the blending factor is applied directly to the reconstruction
gradients as shown above
• In the PBS, the blending factor is applied to the higher-order convective
terms of the transport equations
✓ Use 1st-to-2nd order blending to obtain more accurate solutions for cases
that would otherwise be unstable if full 2nd order gradients were used

47 © 2015 ANSYS, Inc. June 12, 2018

 1st-to-2nd Order Blending
Example: Supersonic projectile @ M=3.5
• 1st-Order converged solution is possible
• 2nd-Order solution exhibit slight
unsteadiness
• Blending of 0.65 was used to damp out
oscillation and achieve a converged
steady state solution
Mach Contours

1st-O
2nd-O
Blending

48 © 2015 ANSYS, Inc. June 12, 2018

 Solution Steering
• The solution steering feature in the DBS, and is a
convergence strategy that will navigate the
numerical solution from an initial flow field,
through the difficult initial phase of
computation, to a converged solution with
minimum user intervention
• The convergence strategy (expert-system) is
based on combinations of
• Initialization
• Solution scheduling & heuristics
• Solution monitoring
• Divergence control strategies

49 © 2015 ANSYS, Inc. June 12, 2018

 Solution Steering Strategy

Stage 0: FMG initialization

To obtain sufficiently good initial guess
Can skip FMG if you have already started with a good
initial guess
Stage 1: Initial iterations
Advance solution at a constant initial CFL value and
increase spatial accuracy gradually from First-Order to
maximum accuracy set by the user (typically Second-
Order)
Includes divergence prevention in case start up CFL is
aggressive
Stage 2: Iterate to convergence
Advance solution with CFL update (increase, decrease or
no change) based on monitoring continuity residual
To change Outer-loop divergence prevention
Solution Steering
default settings

50 © 2015 ANSYS, Inc. June 12, 2018

 Activating Solution Steering

• Set up your case the normal way

• Perform standard initialization
• typically from computation domain inflows
• Activate Solution Steering (SS)
• Select type of flow that best characterizes the
problem at hand
• incompressible
• subsonic
• transonic
• supersonic
• hypersonic
• Adjust settings if desired
• Start iterating
51 © 2015 ANSYS, Inc. June 12, 2018
 DBNS Solution Steering: Turbine Stage Example

Blending Factor - Stage-1

a  f = 0 + (1 −  )  ( • dr )

CFL=200
CFL Update – Stage 2
CFL
Contours of Static Pressure
CFL=5
0 100 200 300 400
Stage-1
Stage-2

Mesh
CFL update Termination level

0 100 200 300 400

Solution steering used during a turbine

stage simulation

52 © 2015 ANSYS, Inc. June 12, 2018

 Convergence Acceleration For Stretched Meshes (CASM)
• Accelerates the convergence of the DBS-implicit lmax
method on highly-stretched meshes lmin
− 2 to 10 times faster than without using CASM
• Uses local cell CFL value equal to specified CFL
value times a factor proportional to cell aspect
ratio
− optimizes the solution advancement for implicit Standard time step
method V lmin
t  CFL  CFL
• Limitations  (  A ) f 

− Can be used with Solution-Steering but manual

CASM time step
schedule adjustment required (reduce max &
V lmax
min CFL values) t  CFL  AR  CFL
 (  A ) 
−
f
Steady-State solutions only

53 © 2015 ANSYS, Inc. June 12, 2018

 CASM Example: RAE Wing-Body Configuration )

CASM Standard

Solution speed-up of about

ten times is achieved with
CASM!

Cd CL
Standard Standard
CASM CASM

54 © 2015 ANSYS, Inc. June 12, 2018

 DBS – Explicit Solver Transient Settings

• Density-based explicit solver with explicit

time marching is used primarily for
simulating transient shocks and acoustic
waves
• This type of problems is often treated as
inviscid or laminar
• 4th order Runge-Kutta for time advancement
• Time step is determined by the CFL
condition
⁻ CFLs ~ 1.0 or smaller are typically used
• Dynamic mesh adaption can be used to
refine the mesh around propagating shock
waves

55 © 2015 ANSYS, Inc. June 12, 2018

 DBS – Explicit Transient Solution Example
2D and 3D Shock Tube Cases
4.25 m 1.5 m 4.25 m

Receiver 2 Receiver 1
Driver p1 / pD = 0.01
p2 / pD = 0.5 0.2 m
pD = 100,000 Pa 1 / D = 0.01
2 / D = 0.5 D = 1 kg/m3

Diaphragms

numerical numerical
1.0 1.0
analytical analytical
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
p/pD

p/pD
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1
t = 716  sec 0.1
t = 1194  sec
0.0 0.0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
x, m x, m

2D solution
56 © 2015 ANSYS, Inc. June 12, 2018
 3D Shock Wave Motion

707 sec

1158 sec

Mesh around left-traveling (weak) shock Mesh around right-traveling (strong) shock

57 © 2015 ANSYS, Inc. June 12, 2018

 Comparison of Shock Tube CFD with Analytical Solutions

numerical numerical
1.0 1.0
analytical analytical
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
p/pD

p/pD
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 t = 707  sec 0.1
t = 1158  sec
0.0 0.0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
x, m x, m

Static pressure distribution alone the centerline of the 3D shock tube at

two different time instances

58 © 2015 ANSYS, Inc. June 12, 2018

 Summary – Compressible Flows
✓ ANSYS CFD solvers can effectively and robustly simulate
compressible flow problems for all flow regimes, from low
subsonic to hypersonic
✓ The PBS or DBS solution algorithms may be used for all flow
regimes. However…
▪ PBS is preferred for lower speed flows
▪ DBS is preferred for supersonic and hypersonic flows

✓ Accelerated convergence rates significantly decrease run times

and reduce production cycle
✓ Solutions can be obtained with minimal user setup and
monitoring (e.g. solution steering)
✓ State-of-the-art post-processing capabilities provide all-inclusive
tools for graphical and quantitative presentation of the results

59 © 2015 ANSYS, Inc. June 12, 2018

 ANSYS Knowledge Database

Structures FAQs

Fluids Demos
17,000+ knowledge
Electromagnetics materials Tips & tricks

Installation/licensing Applications

Highlights Created, reviewed & maintained by ANSYS experts.

Over 500 video demos showcasing pre- and post-processing.

Used by over 40,000 people from around the world every month.

60 © 2015 ANSYS, Inc. June 12, 2018