ABP Fluent 2021R2 External Aero Automotive v0

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 94

• DECEMBER 2019

Application Best Practices


External Aero Automotive

For Ansys Fluent 2021R2


Application Best Practice - Informations

• Application Best Practices: External Aero Automotive

‐ Software: Ansys Fluent


‐ Version:2021R2
‐ Keywords and Tags : Automotive, Aerodynamics

2 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Outline : Automotive External Aero best practice
Geometry Preparation
Geometry Cleanup , Sliding mesh/ MRF zone preparation

Surface mesh
Part Management , Part Replacement, Gap Closure, Wrap Tech

Prism
Split Prism, Adaption , Stair Step

Volume mesh
Refinements, Poly-Hex, Hexcore , Quality mgt

Solver Models
Steady GEKO , SBES/LES, New NW Treatment, GEKO Tunings

Solver Numeric
PMN/DTS/OCM , URF/DS/CFL

Postprocessing Topics covered !

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Geometry Preparation &
Meshing
External Aero Meshing Strategy
• For accurate prediction of drag and lift forces we need to accurately resolve styling surfaces and Heat-exchangers (cooling
system)
• It is important to review the styling model and heat exchanger interfaces for good quality mesh and hence solver accuracy,
based on it decide mesh sizes, check for proximity in regions. This will help to resolve surfaces more accurately with less
iterations.

Reference: DrivAer Model : “Geometry”. Technical University of Munich.


https://www.epc.ed.tum.de/en/aer/research-groups/automotive/drivaer/

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Typical CAD preparation before going to FTM
• SpaceClaim/CAD package : STL Separation
STL Separation
‐ General issues while working on
STL/JT/faceted data
• Sometimes entire assembly comes as single stl
and it’s very difficult to work on it.
‐ STL separation for avoiding over
refinement during automation

‐ MRF creation and adjustment to have


clear intersection between rotating
bodies and MRF

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Typical CAD preparation before going to FTM
Geometry
Creation
• SpaceClaim/CAD package :
‐ Tyre step creation to avoid sharp angle between
tyre and tunnel. (make sure the height of step
above ground will be in between 8-12mm)
‐ Tunnel creation if tunnel is of non-rectangular
shape or has slip/No-slip patches on ground
‐ Heat-exchanger preparation Geometry Creation
• Remove rounds/sharps angles between frame and H
porous region this will help to get good 1:1 X
interface between porous region and wrap bodies Frame

H
X-section X
It is very important to have 1:1 pairing for hxc in/out zones. Sharp angles,
roundness sometimes gives trouble to get right rezoning of wrapper,
which creates issue in solver

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


FTM : Pre-processing
• FTM workflow for mesh generation.
‐ Automated, Easy learning curve
‐ Settings can easily adjusted
‐ For certain complex meshing cases custom journal tasks can be
added in workflow
‐ Can save and repeat workflow for variants
‐ Gap closure, Hole closure operations are included as a part of
workflow automation
‐ Ability to combine wildcard with operators
• Will be available from 2022R1
• Operators
▪ ^ : indicates a NOT function, or negation
▪ | : indicates a Boolean OR function
▪ & : indicates a Boolean AND function
Wheel and
front

Wheel and not


tire

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice on part management
• File import : FTM workflow allows to import
below formats
‐ Ansys Formats :
• From SpaceClaim : Can be use to import *.scdoc or *.fmd files
from SpaceClaim
• Mesh files : Can export *.msh from ANSA
‐ Other CAD Formats :
• *.jt, *.stl, *.CAT, *.prt and other neutral formats are also
supported

• Meshing Model :
• Under part management task, we can use
existing “External flow” template which will
create pre-filled objects and user can arrange
model/parts/assemblies into these objects
• Apart from existing objects user can also create
new objects in case one wants to have multiple
objects as per design

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice on part management
• Object Settings :
• Object settings are important while importing CAD model in FTM workflow
• For getting correct face-zones and correct objects we need to assign correct “object settings”
• Object setting always works on object so make sure if settings are applied on objects only
• Understanding of object, parts and assembly in part management

Object • Objects are made of assemblies and parts


Assembly
• Object settings can not be applied on assemblies/parts
Parts • Object settings for parent object and child object can be different

Object

Parts

Object
Assembly

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice on part management
• Object Settings to get correct output • External Object has different parts, We wants to import part
names as zone names and all zones should associated to single
object
• In Advance Object settings use
• Create One Zone Per – Part : This will assign part names to
face zones
• Turning on “Prefix Object Name to Zone Name” is recommended
for further automation in solver setup

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice on part management
• Object Settings to get correct output • In this example, Miscellaneous object has weird part names, repeating part names and
important part names. Arrange them into 3 more objects which are under miscellaneous
Unusual part name
object.
• Apply Object settings
• Miscellaneous object :
Repeating part name • Create one zone per – part
• Merge all children : this will merge children objects into one and finally we will get
single object
• EngineBayFlow object :
• Create one zone per – object
Important part name • Exhaust object : Children object settings get override
• Create one zone per – object
• Wheel_attaching_parts object :
• Create one zone per – part

• Object setting for EngineBayFlow and Exhaust object


creates single zone and assign object name to it
• Object setting for Wheel_attachment_parts object
creates individual zones as per part names

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice on Vehicle Body of Influence
Offset BOI is local refinement region
Below settings are recommended for full scale vehicle

Flow Direction : direction around which BOI should scaled


Wake Levels : Number of BOIs
Wake Growth Factor : BOI scaling factor in flow direction
Cross Wake Growth Factor : BOI scaling factor in other two
directions

Note: Defeaturing size depends on model size. Defeaturing size used here is
for actual car model
If model is scaled down then defeaturing size will change Vehicle BOI created based on above setting

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice on feature edge extraction
• Feature edge extraction is very subjective and model dependent process
• Make sure that prominent features like styling surfaces, heat exchangers, under covers, wheels, MRF are extracted nicely.
• General guidelines are as below
• External/styling surface, Heat exchangers, undercovers/shielding, wheels, MRF : 20-40 degree
• Grill : 10-20 degree
• Underhood, Underbody components, platform, brake assemblies, suspension, Exhaust and others : 40-60 degree
• Intersection edge creation :
• Intersecting edge extraction between wheel-step, ground, wheel, MRF is must.
• Sometimes if external bodies are intersected, it is important to create intersection edge between those
bodies to get good feature capturing for wrapping operation

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice for high quality interface
• Heat Exchanger boundary rezoning issue
‐ If sharps angles, roundness, proximity regions are not resolved in cad level for HXC
region, it leads to wrapper rezoning issue and creates trouble in solver

• Small re-construction on Frame can help getting high quality mesh

Poor Clean 1:1 pairing


rezoning of till end
wrapper

With geometric simplification of Heat exchanger frame gives


More detailed slides in appendix better overlap interface

16 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice on assigning sizes
• We generally use two types of sizes during FTM process
• Local sizes : this takes part in wrapping and surface remeshing
• BOI sizes : This takes part in volume meshing.
• Optimized BOI settings for complex models with good solver accuracy :
• For typical car model first BOI value should not exceed above 7mm. You can go below this value for simpler models or for more accuracy,
but it will increase cell count.
• BOI : 7, 14, 28, 56, 112…
• For scaled model multiply scaling factor by above value to get similar count with similar accuracy
• Surface mesh sizes
• Use fine curvature sizing on styling surface, wheels, heat exchangers, MRF, undercovers/shields, cables
• Use coarse curvature size on under-hood, underbody components.
• For grill, mirror, fan and other small feature we sometime go much lesser value. Also, for better curvature refinement we
recommend to change curvature angle from 18 degree to 12 degree.
• Proximity settings are important in various locations such as in between MRF and wheels, fan and mrf, HXC and surrounding bodies
to capture geometry accurately

Curvature (target) Proximity –face (target) Proximity-edge (target)


min-max-curvature angle-GR min-max-prox-GR min-max-prox-GR
styling surfaces 2-7-18-1.3 - -
Grill 0.5-3-12-1.3 - 0.5-3-1-1.3
HXC 4-7-18-1.3 0.7-6-2-1.3 (self prox)
wheel 4-7-18-1.3 1-7-1-1.3 (between MRF and tyre)
fan 0.4-3-12-1.3 1-7-1-1.3 (between MRF and fan)
This sizing is for getting idea about how to apply
MRF - - -
UB,UH 7-16-1.3-18 - -
sizes for typical car model. It can be different
Tunnel 448 - - model to model.

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice for cell count reduction
• Effect of STL separation at hood

• Such unwanted refinements


from roof and hood can be After separation of STL surfaces
removed by separating the stl We can assign sizes to top hood to
files avoid over refinement

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Best practice for closing gaps
• Automatic Gap Closure :
‐ It’s an automated process available from 2022R1. in 2021R2 we can use script that will help closing gaps
‐ From Object names and the maximum gap threshold definition required as input. FTM will try to fill all the gaps
within provided range

STL Input file Gap closure How to do it in FTM?

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Meshing strategy
• Poly-Hexcore is suggested (it reduces cell count/solve time) however for FTM meshes it is not suggested to use
high AR cells with Poly-Hex meshes, e.g. y+1. For such cases high AR cells or when robustness is a concern, e.g.
with poor quality FTM meshes, Hexcore is suggested (it will raise cell count by ~20-30% at same sizing field)
• Y+ < 4-5 to have one point in the viscous sub-layer
‐ Y+1 is ideal but can lead to less robustness when wrapped surface meshes and hexa-poly meshes are used
• Experience shows that Y+ ~10 is a good Accuracy vs Speed vs Robustness compromise (in particular for FTM Poly-
Hexcore meshes) – this has been validated with comparison to exp. data pls see “Typical study of different variants using SBES”
slide in this presentation, in the “Solver Setup: Transient SBES” section

• BL mesh needs to cover as much as possible the physical BL. For typical 39 m/s speed BL generally grows up to
about 20 mm towards the end of the car – compromise is needed here 12-15 mm mesh BL height can be sufficient
in most cases
• It is suggested to have a constant BOI region embedding car surfaces, underbody up to floor and first part of the
wake

20 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


General guidelines for meshing with Poly-Hexcore mesh for
precursor RANS and SBES
Mesh refinement estimated is from quick steady state calculation.
Mesh Metrics Values
Poly Hexcore (hanging nodes)
Mesh count ~100 mill
Minimum cell size 7 mm

First layer height on styling 0.2 mm 14 mm 7 mm

# Prisms on Styling 8 (12 preferred, 8 for


compromise accuracy vs speed)
# Prisms on UB 3 (8 is preferred)
Growth rate 1.6 - 1.2
Total Prism Layer height on ~15 mm (depends on Reynolds
styling number, typical for 39 m/s)
Y+ (avg styling) 4-5 preferred.
10 for compromise accuracy vs
mesh count
Mesh quality Inverse orth Q <0.98
Hanging node mesh preferred
Max AR <1000
over conformal polyhex
• For underbody components : Last ratio setting with first cell height
0.5mm used.
• For ground & Tire : Aspect Ratio set to 10 with 4 prism layers

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


High level GUI→ batch execution workflow

• One time process for


Generate workflow mesh generation using save *.wft File
GUI based FTM
file (FTM) in GUI workflow.
Before Batch meshing, user first need to get
workflow ready manually and same workflow can
be useful for batch execution
Create a main- • Main_journal.jou
journal.jou file to file to be read in • Batch file creation :
fluent will carry out • We can save wft (FTM workflow) file
read the FTM file and all meshing from FluentMeshing
other supporting files operation • You can create another journal which
in fluent batch mode automatically
will call above wft file and execute all
the tasks
• Benefits:
• User can generate meshes without UI.
• If some changes are • User can tweak wft file with text editor
Mesh will be required in prism and can generate meshes for different
generated and sizing then we variants
automatically. can modify the
setting files

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Meshing Workflow Automation

Files required for workflow automation


‐ Main journal for batch execution
• main-journal.jou
‐ Input Setting files
• delete-objects.jou (Creating variants)
• Prism-settings.jou (Modify if one wants to change prism settings)
‐ Other Supporting files
• workflow.wft (File having all FTM tasks)
• TGTools_2021R1_13December2020.bin (Advance settings for mesh generation required for V2021R1/R2)
• Gap_cover_21R1.bin (settings for gap closure required for V2021R1/R2)

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Step 1 : Modify delete-objects.jou file as per variant
Delete-objects.jou a

• This file is used to change below b


settings
a. Global size controls
b. Close gaps
• Provide inputs of objects which we want
use for closing gaps c
c. Delete unnecessary objects
• We are importing all objects and as per
variants we are deleting unnecessary objects

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Step 2 : Modify main-journal.jou file
• In our meshing process we have two file
extensions
1. .stl (all vehicle CAD files)
2. .msh/.fmd (file for tunnel)

• Please provide path along with msh file in line #3


• Please provide only path for stl files line #9

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Step 3 : Load main-journal.jou file along with FluentMeshing
• Launch FLuentMeshing
• Select “main-journal.jou” and launch
FluentMeshing It should execute workflow
journal

• For batch execution with GUI use below


command
/nfs/…./ansys_inc/…/v211/fluent/bin/fluent* -meshing
3ddp -i main-journal.jou

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Mesh details
Wrapped surface mesh can shows good
feature capturing

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Mesh details

• Volume mesh Stats


‐ Mesh type : Poly-Hexcore without hanging nodes
‐ Regions : 1 main fluid, 4 MRF fluids
‐ Prism settings :
• 8 layer : Uniform setting - (height=0.2mm) with 1.6 Growth Rate
on Styling surfaces
• 3 layers : Last ratio setting = 0.4 and height = 0.5mm for
underbody and underhood components
• 5 layers : Aspect ratio setting = 10 (first AR) on Tunnel and Tyre
‐ Cell count : 95 million
‐ Cell Quality :
• Inverse Orthogonal Quality: <0.98
• Fluent Aspect Ratio: <1000

Prism Settings are for full scale model


©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.
Typical Mesh statistics for industrial external Aero case

• Quality targeted = Inv Ortho <0.98, AR<1000


• Meshing time with
‐ Import raw cad = ~0.5 hour
‐ Gap closure = ~ 0.5 hour
‐ Surface mesh = ~5 hours
‐ Volume mesh = ~2 hours (32 cores)
‐ Improve quality = ~1 hour
‐ Total Time = ~9 hours
‐ Mesh count = ~ 100 million

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Solver Setup: RANS
Pseudo Transient Strategy
• Mesh check & quality imrpove
• Initialization
‐ Fast Hyb-Init to locate low velocity cells
‐ Hyb-Constant-Velocity-Init with low velocity cells patching
• Phase I – Conservative (20 it)
‐ HOTR ON
‐ 1/10th timescale
‐ Lower URF
• Phase II – Aggressive (300 it)
‐ HOTR OFF
‐ 2* timescale
‐ Higher URF
The timestep is estimated as a timescale= Car_Length/Velocity. In Phase one 1/10 and
smaller timestep can be used depending on the convergence rate. In the aggressive
phase you can go to 2*timescale and more (with high URF, II order etc.)

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Mesh check & quality imrpove
• Report mesh size and quality in transcript
‐ Useful for debugging Journal script commands
• Mesh check
‐ Find out the quality issue before running simulation /mesh/size-info
/mesh quality
‐ Optional step is to add repair improve command (No
/mesh/check
harm)
;;/mesh repair-im repair q q ;; optional
• Poor mesh numeric (pmn) ;; /solve set pmn enable yes 0 q q ;; optional
‐ level set to 0 can provide better convergence in very /parallel partition set laplace yes 10 q q
bad quality meshes. However, it’s optional step and
not recommended for good quality meshes
• Parallel partitioning Original mesh Repaired mesh
‐ Laplace partitioning is reccomonded

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Turbulence model settings
• Recommended model
‐ GEKO model Journal script commands
• Typical GEKO Parameters values
/define/models/viscous/kw-geko yes
‐ Csep=1 /define/models/viscous/geko-options csep yes 1.0
‐ Cmix, Cj, → default
‐ Above setting should give good accuracy, However if
even higher accuracy is required, then further tuning
of GEKO parameter is possible. Please see appendix
slide # for more details

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Reference values
• How to chose reference values
‐ Need to specify accurate values of reference Journal script commands
properties
‐ Directly affect the Cd, Cl coefficient values /report ref density 1.205
/report ref length 2.775
/report ref velocity 38.888
/report ref viscosity 1.822e-5
/report ref area 2.32 q q

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Boundary conditions
• Inlet : Velocity , Outlet : Pressure outlet,
• Tunnel Top & Side wall : Symmetry

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Boundary conditions
Journal script commands (Example)
• Velocity inlet
o Typically, velocity magnitude in normal direction is specified /define bc zt kwt*inlet* () v-in q q
o For turbulence specification: Turbulence intensity and /define bc set v-in kwt*inlet* () vmag no 28 vel-spec
turbulence length scale option is recommended n n y ke-spec n y turb-in 0.2 turb-len-scale 0.01 q q q
• Pressure outlet
• Zero-gauge pressure with similar turbulence specification as of
inlet 1m
• Slip boundary in front of car
• Try to model Test chamber boundary as accurate as in
experiments
• To achieve Top hat velocity profile at inlet of Test chamber,
Apply No slip or Symmetry conditions
• The upstream ground wall need to split at exactly same Uniform Velocity
location where Test chamber starts
• The ground distance in front of car = 1 meter

slip

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Heat exchangers (Porous zones)
Journal script commands (Example)
‐ Use alternate formulation for better convergence /define bc fluid hxc-rad-cells no no no no no 0 no 0 no 0 no 0 no 0 no 1 no no
no yes no no 1.0 no 0 no 0 no 0 no 1.0 no 0 yes no 1.6e7 no 1.6e9 no 1.6e9
no no 81.37 no 1800 no 1800 0 0 no 1 constant 1 no

Radiator

Radiator
Velocity 250
dp [Pa] 1
[m/s] y = 2.4411x2 + 14.52x p= xDu + xCu 2
200 R² = 0.9988 2
1.3 17.6
150
2.4 47.5 D = 14.52 / (μ * 0.05)
3.6 85.2 100 = 1.6e+07
C = 2.4411 / (0.5 * ρ * 0.05)
4.8 129.4 50 = 81.37
6.1 179.4 0 Where
7.3 234.6 0 2 4 6 8 μ = 1.8265e-5 [kg/m-s]
ρ = 1.2 [kg/m3]

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Interface treatment for heat exchangers and MRF

Journal script commands (V2021R2)


/define/mesh-interface/remove-left-handed-
interface-faces? yes , q q
• Recommended to use latest fluent version /define mesh-interface one-to-one no yes q q
/define mesh-interface aut-pair yes no yes iff , , , q q
• The commands in older Fluent version are different
• The TUI command to remove left handed faces provide
better convergence. Need to executed only in TUI
‐ /define/mesh-interface/remove-left-handed-interface-faces?
yes , q q

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Solver setting for steady state
Step :1 (Conservative setting) for first 20 sub-iterations

;;; ------------------------- robustness settings ------------------------;;


/solve/set/gradient-scheme no yes, ;; NBG LSQ
/define/models/viscous/turbulence-expert kato yes
/solve/set/high-order-term-relaxation/enable? yes
/solve/set/warped-face-gradient-correction/enable? yes yes
/solve/set/p-v-coupling 24
/solve/set/discretization-scheme/k 1 ; 1= second order
/solve/set/discretization-scheme/omega 1 ; 1= second order
/solve/set/discretization-scheme/mom 1 ; 1= second order
/solve/set/discretization-scheme/pressure 12 ; 10= standard, 12=
second order, 14=presto
;; Pseudo Solver Settings and Under Relaxation
/solve/set/p-v-controls 0.25 0.25 ;; Low implicit URF for pressure
and momentum
;; Modifying default Multigrid
/solve/set/multi-grid-controls/k 3 0.1 0
/solve/set/multi-grid-controls/omega 3 0.1 0
/solve/set/multi-grid-controls/pressure-coupled 3 0.1 1
;; Aggressive coarsening ON
/solve/set/amg-options/aggress-amg-coars y y ;; Aggressive
coarsening for faster AMG convergence

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Solver setting for steady state
Step :2 – Aggressive settings for next 380 sub-iterations

;; Define Reports and monitors


/solve/report-files/add forces report-defs cd_total cl_total
cd_hx_porous cl_hx_porous clf clr () file-name ./forces.out q
/solve re-plot add cd_total report-def cd_total () q q q
/solve re-plot add cl_total report-def cl_total () q q q
/solve re-plot add cme_total report-def cme_total () q q q
/solve re-plot add clf_clr report-def clf clr () q q q

it 380
;; total 400 iterations observed to be sufficient for most of the cases

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Solver Setup: Transient SBES
Guidelines for fast transient SRS External Aero
• Optimizing mesh size and refinement regions is fundamental
• Hexcore/Hexapoly mesh looks a better compromise of accuracy vs total cell count
• Time-step ramp down approach
− Starting from larger time-step size to accelerate development of turbulent structures, e.g. start with 1 order
of magnitude larger time-step then final time-step size
− In many cases this has shown to give 20-40% speed up
• Solver Set up: we have 2 main strategies:
− Using SIMPLEC and reducing the number of inner iter up to 2-3, at the same time rising up URF (0.95-1 if
possible), seems to work on most cases. Use small time-step size in order to have Convective Courant
number of order unity. This is the suggested strategy.
− PBCS with high URF (allows higher URF on low quality mesh then SIMPLEC in some cases) with 2 inner-iter
giving similar results of SIMPLEC with 4
• PBCS approach seems to be able to handle larger time-step size. Running with larger time-step size (e.g.
Convective Courant number about 5-10 and larger) can give consistent speed up but this is risky and not
CFL consistent.
• PBCS is more robust and can be the choice in case of bad quality mesh and poor convergence with
SIMPLEC
• On high core count LSQ is generally faster than NBG (about 15%)
− LSQ is suggested/default on poly meshes
− NBG in some cases show to be more robust

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


LES Mesh Resolution Estimate – (2021R2)
Two new output fields have been added to help you in the estimation of the mesh
resolution quality in a scale-resolving simulation of a turbulent flow
• LES Resolution Estimate available for precursor steady state RANS
• LES Resolution Quality part of the “Unsteady Statistics”

LES resolution Estimate in RANS: LES resolution quality in SRS:


LRE = 𝑙𝑡 ΤΔ𝑥 should be >= 5 for LES region Q= 𝑘𝑟 Τ 𝑘𝑟 + 𝑘𝑚 >= 0.8 good resolved flow for LES
Estimate shows good resolution Adequately resolved kinetic energy in wake region
around the car

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Turbulence Hybrid SBES
Stress-Blended Eddy Simulation (SBES) is hybrid RANS-LES
turbulence closure model.
• Main difference between SBES model compared to
Detached Eddy Simulations (DES) and Delayed DES
(DDES) models is a much stronger shielding function fSDES
for the RANS region. Has faster transition from RANS to
LES
• Better control of RANS behavior through GEKO model
(e.g. tuning for separation)
• More acurate consistent results than DDES models
• Turbulence stress tensor blending, and eddy viscosity LES
formulation are defined as shown below RANS

Ʈ𝑺𝑩𝑬𝑺
𝒊𝒋 = Ʈ𝑹𝑨𝑵𝑺
𝒊𝒋 𝒇𝑺𝑫𝑬𝑺 + Ʈ𝑳𝑬𝑺
𝒊𝒋 𝟏 − 𝒇𝑺𝑫𝑬𝑺 (1)

Ѵ𝑺𝑩𝑬𝑺
𝒊𝒋 = Ѵ𝑹𝑨𝑵𝑺
𝒊𝒋 𝒇𝑺𝑫𝑬𝑺 + Ѵ𝑳𝑬𝑺
𝒊𝒋 (𝟏 − 𝒇𝑺𝑫𝑬𝑺 ) (2)

44 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


SBES vs DDES, IDDES

45 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


• Add Mesh resolution for LES
Fast transient setup example and Time resolution slides

• GEKO Csep=1, SBES, NBG


• CFL<4 in first refinement region
• Wall Yplus<10 on styling surfaces
• Stages
1. Steady State iterations PBCS =400
2. PBCS or SIMPLEC (in V2021R2)

B. PBCS (Transient) B. SIMPLEC (Transient)


1.Δt= 5e-03 , steps =100 , subitr=15, URF = 0.50, HOTR, flushing 1.Δt= 5e-03 , steps =100 , subitr=15, URF = 0.40, HOTR, flushing
2.Δt= 1e-03 , steps =150 , subitr=8, URF = 0.75, HOTR, flushing 2.Δt= 1e-03 , steps =150 , subitr=8, URF = 0.6, HOTR, flushing
3.Δt= 4e-04 , steps =375 , subitr=4, URF = 0.90, flushing 3.Δt= 4e-04 , steps =375 , subitr=4, URF = 0.7, flushing
4.Δt= 2e-04 , steps =1000, subitr=2, URF = 0.95, flushing 4.Δt= 2e-04 , steps =1000, subitr=2, URF = 0.8, flushing
5.Δt= 2e-04 , steps =5000 , subitr=2, URF = 0.95, sampling 5.Δt= 2e-04 , steps =5000 , subitr=2, URF = 0.8, sampling

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Typical study of different variants using SBES
Accuracy
(compromise accuracy vs speed, sim time < 24hrs)

Outcome: 6 variant out of 8 within 6 counts.

Matching trends with Experiments

CD
All trends on 8 variants are matching well

PIV Contours
Captured correctly 10 counts

Meshing & Solver Automation var-1 var-2 var-3 var-4 var-5 var-6 var-7 var-8
Variants
Fully scripted automated workflow CAD to Results Cd Test Cd SBES
Meshing Speed (64cores)
(Gap closure + Wrapping+ vol. mesh + prisms): 11 Hrs

Solver Speed (504 cores)


PBCS: 38 Hrs
SIMPLEC: 23 Hrs

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Solver Troubleshooting
Some latest tips for solver robustness
vehicle PMN= 0 PMN=0 &
Variants (Ortho-EM-default)
For steady state: Variant 1 Diverging: BUG 427594

Converging with OrthoEM


Ortho Enhancement Metrics (OEM) for Variant 2

robustness Variant 3

Variant 4

Variant 5

Variant 6

For Transient: Variant 7

Variant 8
“Dual time stepping” + SIMPLEC + Use Variant 9
PMN 1

Dual time
stepping &
Use Auto select for Flux type: PMN=1
Simulation with many industrial cases have
showed improvement in convergence using
RC2 (momentum based)

Tire modelling common BP:


Use of Pylon step (Box or tire cutting ground
shape extruded)

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Some latest tips for solver robustness: script example

For steady state: In steady state journal


•First turn on beta : /define beta yes q q
Ortho Enhancement Metrics •Ortho Enhancement Metric
(OEM) for robustness • /solve/set/poor-mesh-numerics orthogonality-enhancing-metrics? Yes 0.5 q q q

In Transient part of journal


For Transient: •DTS command for V2021R2, For later versions the commands are changed
•Commands available only for transient and SIMPLEC/SIMPLE solver
“Dual time stepping” + • (enable-feature 'pbns-dual-time-stepping)
SIMPLEC + Use PMN 1 • /solve set pmn enable yes 1 q q ; ; reverting default 1
• /solve/set/advanced/dual-time-stepping enable? yes courant-number 200 q q q
• DTS (CFL200, URF- P 0.6, Mom 0.7, k 0.8, omega 0.8) + RC2 seems to be better
solution. You can relax CFL as small as 5 for bad meshes
Use Auto select for Flux type:
Simulation with many Using correct flux schemes (RC1/RC2)
From 2021R2 we have auto select option available using below command
industrial cases have showed
• /solve/set flux-type yes ;; Automatically selecting RC scheme
improvement in convergence
using RC2 (momentum based)

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Postprocessing
Postprocessing tools

• Fluent
‐ Basic level postprocessing & user friendly
‐ Forces and moments
Covered in this document
‐ Cumulative plots
‐ Partwise drags
‐ Simulation report (scriptable)
• Ensight
‐ Much comprehensive capabilities & Automation possible
• CFDPost
‐ Good rendering but has only basic postprocessing capabilities
‐ Not recommended for external aero postprocessing
• VRXPERENCE
‐ Used only to generate high quality images (Perceived quality)
‐ Can be used with Ensight

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Reporting Cd & Cl coefficient

53 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Clf and Clr

ΣM=0
Me+Clf*(L/2)-Clr*(L/2)=0
Cl= Clf + Clr
Clf = (Cl/2)- Cme
Clr = (Cl/2)+ Cme

Clf
Cme Clr

Make sure that the formulation is consistent with what


customer is using for wind tunnel testing

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Post-Proc for ext aero

• For Porous zones : drag, forces, moments


are now available in report definition
(2020R2)

55 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Cumulative Plot

The Cumulative Plot dialog box allows you to analyze the


development of force, moments, or coefficients for your
model in the any desired location or direction
Porous zones also supported after Version 2021R1
Car image can be superimposed in PowerPoint

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Post-Proc for ext aero
• Time-averaged forces (drag, lift, moments) for cumulative plots
(2021R1-beta)

57 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Partwise drag

• There are multiple ways to get partwise


(subsystem wise) drag
1. Using Fluent report definition
1. After finishing the steady state calculation, select the per zone
option and click compute. I
2. It will print the zone-wise drag coefficient. Then we copy data to
excel for grouping and further postprocessing
Part/Group Name cd_DR1
2. Using scheme file (not recommended due to complexity) Main Body 0.160
1. Group the walls in subsystem (e. g. Tires, Powertrain, cooling Front Tires 0.018
system, ..) Rear Tires 0.015
Front Suspension 0.006
2. Scheme script will calculate and export data in .csv file
Rear Suspension 0.006
Engine Powertrain 0.019
Exhaust System 0.002
Fuel Tank 0.003
Under Covers -0.002
Cooling Module 0.078
Heat Xchangers 0.077
Other Parts -0.056
Total 0.315

58 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Simulation report (scriptable)

• Most of the user prefer own scripts and


Ensight based postprocessing
• Simulation Report feature is introduced
from V2021 R1 (as a beta)
• Report be generated using GUI or scripted
using journal
• Can be exported in HTML and PDF file
format
• In HTML we can also visualize 3d AVZ files

59 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Visualizing tubulence structures using Lambda-2 & Q-criteria

L2 = 1e4 Q = 1e4

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Summary

• The best practices are based on combination of learnings from may customer
benchmarks on V2021R1 & V2021R2
• The slight variation in commands is expected in latest versions
• The preprocessing, Solver and postprocessing steps were discussed by considering a
DrivAer car
• In case of any questions please reach out to be
‐ Rohit (rohitkumar.sonawane@ansys.com)
‐ Tushar (tushar.Jadhav@ansys.com)
‐ Domenico (domenico.caridi@ansys.com)

61 ©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Appendix
USE THIS SLIDE TO DIVIDE PRESENTATION
CONTENT FROM SUPPORTING APPENDIX
CONTENT
Solver Testing – Pseudo Transient Strategy

Dr. Domenico Caridi


Senior Fluids Product Manager
Automotive

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Coupled CFL Solver

‐ FLUENT applies under-relaxation factor of the diagonal coefficient for each transport equation. This
is called an implicit relaxation of the equation. Relaxation can be different for different equations

1−

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

‐ FLUENT formally introduces CFL number to control implicit relaxation for coupled P-V system
instead of direct change of relaxation by user

1 1−
=
CFL 

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Pseudo Transient (Coupled)
‐ It applies diagonal dominance to the system adding pseudo-transient term or solving system as
unsteady for all equations. Time scales can be different for different equations.

 pVol p
( n +1
p −  pn )+ a  n +1
−  aiin +1 = S p
t
p p

‐ Explicit relaxation factors for all solution variables as alternative to traditional relaxation of
equations effect tuning for skew meshes and for high order numerics ( FLUENT using that in the
Coupled CFL Solver for pressure and momentum)

 new =  old + 


‐ Default setting in linear solver for pseudo-transient mode
• ILU smoother and F-cycle for scalar systems

• More uniform removal of errors for wide set of cell sizes compare to CFL approach. That can be beneficial for
unsteady problems modelled as steady-state.
• Overhead is up to 15% increase in CPU time per iteration
©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.
Pseudo Transient Strategy

• Initialization
‐ Fast Hyb-Init to locate low velocity cells
‐ Hyb-Constant-Velocity-Init with low velocity cells
patching
• Phase I – Conservative (20 it)
‐ HOTR ON
‐ 1/10th timescale
‐ Lower URF
• Phase II – Aggressive (380 it)
‐ HOTR OFF
‐ 2* timescale
‐ Higher URF
The timestep is estimated as a timescale= Car_Length/Velocity. In Phase one 1/10
and smaller timestep can be used depending on the convergence rate. In the
aggressive phase you can go to 2*timescale and more (with high URF, II order
etc.)

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Effect of start up phase

Run1

Run6

Run 6 is the baseline, in the II Phase:


p-v 0.4, Turb URF 0.8, dt=0.1s

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Effect of explicit URF

Run 6 is the baseline, in the II Phase:


p-v 0.4, Turb URF 0.8, dt=0.1s

From
p-v 0.4, Turb URF 0.8

To
p-v 0.3 Turb URF 0.6

No big difference can be noticed

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Effect of too aggressive explicit URF

From
p-v 0.4, Turb URF 0.8

To
p-v 0.7 Turb URF 0.95

Slightly faster but more unstable

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Effect of timestep

From
dt 0.05s

To
dt 0.1s

To
dt 0.2s

Slightly faster and stable

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Effect of too aggressive timestep

From
dt 0.5s

To
dt 1s

To
dt 2s

No speed up gain
and unstable

Convergence is
worsening

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Effect of higher timestep and lower URF

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Final Comments

• Effect of timestep on solution speed up is relevant, start up


conservative phase (small timestep) should be few iterations
• Too agressive URF can lead to larger instabilities but not so
much speed up
• Too aggressive timestep even if balanced by low URF leads to
instability
• Right balance of timestep and URF can be find, guidelines are
provided

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Summary of Settings to enhance robustness
Tested on many models

• Pseudo Transient solver (start with small timestep and HOTR) – BP described in
previous pages
• WFGC
• Poor Mesh Numerics
• Aggressive Coarsening (by default in 2019R1)
• Partitioning using Laplacian Smoothing can help in some cases
Journal File

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Fast Transient Best Practice
for External Aero and Aeroacoustics
Dr. Domenico Caridi
Fluids Product Manager
Europe

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Guidelines for fast transient SRS External Aero
• Optimizing mesh size and refinement regions is fundamental
• Hexcore/Hexapoly mesh looks a better compromise of accuracy vs total cell count
• Time-step ramp down approach
− Starting from larger time-step size to accelerate development of turbulent structures, e.g. start with 1 order
of magnitude larger time-step then final time-step size
− In many cases this has shown to give 20-40% speed up
• Solver Set up: we have 2 main strategies:
− Using SIMPLEC and reducing the number of inner iter up to 2-3, at the same time rising up URF (0.95-1 if
possible), seems to work on most cases. Use small time-step size in order to have Convective Courant
number of order unity. This is the suggested strategy.
− PBCS with high URF (allows higher URF on low quality mesh then SIMPLEC in some cases) with 2 inner-iter
giving similar results of SIMPLEC with 4
• PBCS approach seems to be able to handle larger time-step size. Running with larger time-step size (e.g.
Convective Courant number about 5-10 and larger) can give consistent speed up but this is risky and not
CFL consistent.
• PBCS is more robust and can be the choice in case of bad quality mesh and poor convergence with
SIMPLEC
• On high core count LSQ is generally faster than NBG (about 15%)
− LSQ is suggested/default on poly meshes
− NBG in some cases show to be more robust

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


External Aero Validation Case example

• Mesh of about 100M cells Hexcore Mesh, full detailed


underhood and underbody, with 15 prism layers everywhere
• Steady state in about 600 CPU Hours, e.g. 2 hours on 300 cores DrivAer
− About 250 CPU Hours on smaller model (70M cells)
• Transient in less than 6,000 CPU Hours, e.g. less than 20 hours on 300 cores
• High Accurate Drag and Lift prediction

0.01 0.01 Avg 0.05 Avg


Exp.

Ramp-time Ramp-time Sampling-time


Sampling-time

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Numerics SIMPLEC example

• P-V Coupling: SIMPLEC


• GEKO-SBES
• Gradient Reconstruction: LSQ Based* or NBG
• BCD discretization for Mom, II order upwind others
• Bounded II order Implicit Transient Formulation

URF
- Pressure to 0.8
(or 0.9 if convergence permits)
- Mom, Turb to 0.9
- All others 1
*reccomended for Poly mehses
including Poly-Hexcore
©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.
Single time-step
settings example
Time- No. of Sub
Flow No. of Discretiza
step-size Time- Iter/time URFs
Time Iterations tion
(sec) step -step
Pre, mom:
Pre, mom,
0.9
Initial time 0.96 0.00025 3840 2 7680 turbulence:
2nd order k, w:0.95
tur-visc:1

Sampling
1.28 0.00025 5120 2 10240 || ||
time

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Ramp time-step settings example
Can give very similar results to constant time-step being 20-40% faster

Flow Time-step-size No. of Time- Sub Iter/time- No. of


Discretization URFs
Time (sec) step step Iterations

Pre, mom:
0.65
Pre, mom, turb:
0.8 0.0025 320 6 1920 den, body:1
2nd order
k, w:0.85
tur-visc:0.9
Pre, mom:
Ramp time 0.8
0.096 0.0005 192 4 768 ||
k, w:0.85
tur-visc:0.95
Pre, mom:
0.9
0.024 0.00025 96 4 384 ||
k, w:0.9
tur-visc:0.95
Pre, mom:
Sampling time 0.9
1.28 0.00025 5120 2 10240 ||
k, w:0.95
tur-visc:1

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Example: comparisonof it/timesteps and ramp approach

Ramp Ramp No Ramp – Single Timestep


2subitr/time-step 4subitr/time-step
- 0.003
+ 0.003 + 0.002

Ramp-time Sampling-time Ramp-time Sampling-time


Sampling-time

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Recent ANSYS Developments and Innovations
• Laminar-turbulent transition
‐ 1-equation intermittency model
• Significant enhancement to γ-Reθ model

• Scale-Resolving Simulation
‐ Stress-Blended Eddy Simulation (SBES) model
• More accurate mixing predictions

• BSL2 (k-ω) defined as baseline model (R&D – Beta – R19.x)


‐ Wall distance free
‐ Can be combined with all other models (EARSM/RSM, CC, Transition)
‐ BSL2 Model can be tuned with free coefficients over a wide range of
flows

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


DDES vs SBES

DDES SBES

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Mesh Resolution in LES
• LES requires mesh and the time-step sizes sufficiently fine to resolve the energy-
containing eddies
‐ The mesh resolution determines the fraction of turbulent kinetic energy directly resolved

Coarser Mesh Finer Mesh

ln E ln E

Less Energy is
Modelled

Greater Resolved Area:


Isotropic
Anisotropic Homogeneous
More Turbulent Kinetic
Flow dependent Universal Energy is Resolved
Resolved Spectrum Modelled
 2   2   2   2 
ln   ln   ln   ln  
         
 f   f 

Energy spectrum against the inverted length scale (log scales)


(Please note, this is a local graph: would be different for each and every point in the domain )

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Mesh Resolution in LES (2)
• Integral length scale l0
‐ Turbulent kinetic energy peaks at integral length scale. This scale must be
sufficiently resolved
‐ Approximate estimation for l0
From the precursor RANS simulation l0 can be computed as
l0=k1.5/e or l0 = k0.5/ ( C  ) C = 0.09
l0

ln E • Smallest resolved length scale l


l – smallest resolved eddies – At least, a couple of cells needed in each
Integral Length direction to resolve an eddy with a
Scale, l0
Most Energy
length scale l
Containing
Eddies
Isotropic
Anisotropic Homogeneous
Flow dependent 𝒍 l
Universal
∆=
Resolved Spectrum Modelled 𝟐
 2   2 
ln   ln   ∆
    
 f 

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Mesh Resolution in LES (3)
• Suppose we want to resolve 80% of the turbulent kinetic energy
• Then, we need to resolve the eddies whose sizes are larger than
roughly half the size of the integral length scale l0.
1.0
• Approximately 5 cells will be needed across the
0.8
integral length scale l0.
Cumulative TKE against length-scale of eddies based on the
k ( )
Kolmogorov’s energy spectrum

k l / l0 l0 /∆
k ( l ) = 0.1k (10%) 6.10 0.33
k ( l ) = 0.5k (50%) 1.6 1.25
k ( l ) = 0.8k (80%) 0.42 4.8
0.1
k ( l ) = 0.9k (90%) 0.16 12.5
0.0
0.42 6.1  0

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Mesh Resolution in LES (4)
• It is a good practice to have at least 5 cells across lo to resolve 80% of the k spectrum and 10-12 cells
per lo can warrant a better 90% resolution of the energy spectrum in the regions of interest.
• There are a number of methods to perform an assessment of the resolved k spectrum. The quickest
and easiest assessment is based on the precursor RANS simulation*.

• The integral length scale lo is a local


quantity and a custom-field function (CFF)
can be evaluated using k, ε (or ω) values
from the RANS simulation.

• The number of cells per lo can be assessed


by assuming that

𝟑 **
∆≈
(𝑪𝒆𝒍𝒍 𝑽𝒐𝒍𝒖𝒎𝒆)
• Another CFF of lo /∆ will return the
number of cells per lo

* − Might not be accurate due to differences between RANS & LES.


**− Accurate if the aspect ratios are modest (not far from 1)

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Mesh Resolution in LES (5)

• Contours showing the ratio l0 /∆ need to be plotted at different


predefined surfaces.
• The upper values of l0 /∆ can be clipped so that the well-resolved
areas do not appear and the not-so-well resolved regions could be
identified easily.

Under-resolved Mixing Layer

• Critical regions of a particular interest will need remeshing or local


mesh adaption, if appear as under-resolved.
• The near-wall regions always pose challenges to LES and a number of
techniques are available to address that.

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Cost of LES: Time-Step Size

• The temporal resolution should match or exceed the spatial resolution in LES
‐ Let’s say we have a cell with largest edge length of ∆ x and local average velocity, U
‐ It takes a time interval ∆ t for the flow to travel across the cell
‐ The time step should be small enough to provide an adequate temporal resolution of the
flow as it passes through the cell
x
t 
U
‐ The real velocity can be higher than the averaged velocity
‐ Pre-cursor RANS simulation is used for the assessment of ∆ t
• Good practice to account for differences between instantaneous and averaged velocities as
well as for the errors introduced from RANS by

x Ut
t  Courant Number =  0.5
2U x

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Cost of LES: Time-Step Size (2)
• Custom-Field Function for
local time scales can be
‐ Evaluated
‐ Postprocessed
‐ Minimum time scale can
be calculated and the
rounded value for the
required time step can
be obtained

Please note that a larger ∆ t can be chosen if:


• The minimum time scales are detected far away from the
regions of main interest.
• The regions with smallest time scales are not posing
influence onto the rest of the domain.

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


SRS Timescale and Mesh Size Check: Giulietta case example
Convective Courant
number of order unity Courtesy of FCA - Italy
with dt=3e-5 s

Courtesy of FCA - Italy


L0 to delta ratio= L0 / (Volcell)^1/3
Time scale=(Volcell)^1/3 / (|V|)
L0 = k^1/2 / (Cµ*)

y+<3-4

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Useful References on the Customer Portal

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


Using prism adaption approach for poly-hexcore
meshes
• With Polyhexcore meshes there is possibility of generating non-prismatic cells
• When we try to do anisotropic adoptions on such non-prismatic cell in boundary layer mesh,
it may generate isotropic refinement and mesh count can increase exponentially.
• So, to address this issue we should execute below command just before volume fill operation
in FTM using custom journal
‐ (tgsetvar! 'poly/keep-nlayer-prism-stack 1)
‐ Here n=1 if you are growing one prism layer. Which is typically a case for adaption
approach
‐ As you know, for adaption we grow only one continuous prism layer.
‐ This will work only in latest (v221) versions

Adaption
script in solver

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.


domenico.caridi@ansys.com

©2022 ANSYS, Inc. Unauthorized use, distribution, or duplication is prohibited.

You might also like