Intro to CFD and Multiphysics Simulations

2014 European Altair Technology Conference

June 24-26, 2014 | Munich, Germany

Join, Contribute, Exchange Technical Session #2: Multiphysics & CFD

Moderator: Steve Cosgrove

14:00-16:15 @ Bodensee 1+2

Copyright © 2014 Altair Engineering, Inc. Proprietary and Confidential. All rights reserved.

Agenda

• Opening Remarks and Presentation: Steve Cosgrove (15 min)

• Presentations: (20 min + 5 min Q/A)

• Dr. Fotis Konias, Altair Greece, “CFD Analysis on Gulfstream Landing Gear”

• Dr. Michael Arrigonia, ENSTA, “Indent Tests for Explosive Equation of State

Determination”

• Stephan Pitzing, Knorr-Bremse Systeme, “Simulating Thermal Balance of

Electronics – An AcuSolve Approach”

• Aymen Slimani & Dr. Jorg Sorensen, Magna Car Top Systems GmbH, “

Aerodynamic Simulation of a Cabriolet Soft Top Roof System with VWT”

• Mathias Reichert, Westinghouse Electric Germany GmbH, “Numerical

Simulation of Liquid Sloshing with SPH Method in Nuclear Power Plant

Facilities”


Biomedical- Airflow in Canine Noses

• AcuSolve validated for sniffing in canine nasal passage

• Finest mesh used 58 mm element size (100 Mio elements)

CFD Mesh

Journal of Biomechanical

Engineering SEPTEMBER 2009,

Vol. 131


Studying fluid flow

• Fluid flow can be studies in three ways

• Experimental fluid dynamics

• Theoretical fluid dynamics

• CFD: Computational fluid dynamics


Equations (Navier Stokes)

continuity

momentum

energy

• The Navier Stokes equations are general, describing the majority of flows

• Non linear terms make them difficult to solve

• Another non-linearity is that most real world flow problems involve

turbulence in the fluid

Non linear terms


Turbulence

• Turbulence is a phenomena which includes flow structures of

• large range of length scales (small and large eddies)

• Large range of time scales (low and high frequencies)

• There is no theorem relating Reynolds number to turbulent, but usually

• Low Re number -> laminar

• High Re number -> turbulent (e.g. pipe flow Re > 2500) Parallel layer, no

interaction

between layers

CFD code solves for Ui(mean velocity), Turbulence model gives impact of u’


AcuSolve Turbulence Modeling Slide: Scales of Energy

Transfer in Turbulence Modeling

DES

LES

DNS

(SA,k-eps, k-omega, SST,…etc, etc..)

Co

mp

uti

ng

Co

st

DDES

Resolved=Computed by code exactly, Modeled=approximation

(LES)

RANS, URANS

(RANS)

(LES) (RANS)


Turbulence

• Why is turbulence modeling important ?

• The turbulence determines the point of flow separation

• Different separation point different flow pattern, different performance (e.g. drag

coefficient of a car, lift coefficient of an airfoil)

Same BC, different turbulence model

AcuSolve DES DDES Exp. RANS (SA)

Separation 0.641 0.662 0.665 0.663

Reattachment 1.36 1.18 1.11 1.22


CFD solver technologies

• Primary technologies:

• Finite difference – Phoenix (Cham)

• Finite volume – StarCD, Fluent, CFX, SCTetra, CFD++, OPENFOAM

• Finite element –

• Galerkin formulation: CFDesign, Flotran, Fidap,

• Galerkin Least Squares formulation: AcuSolve

• Others include:

• Panel method – aerospace codes for external aero

• Spectral method – Nekton, Polyflow

• Lattice Boltzman Method (POWERFLOW, XFLOW)


Challenges of CFD

• Pre-processing – Still the bottleneck

• 80% of the CFD analysis time is spent in pre-processing (cleanup, meshing)

• Up to 6 weeks for modeling an external automotive flow

• Fast and automated meshing algorithms are required (e.g. multi CPU)

• Model size from 0.5 Mio – 300 Mio cells (now up to 1,000 Mio)

• Large variation in element size (1e-5 … 20)

• CFD analysis (Solving) – Scalability, Quality, Robustness

• High CPU time for transient analysis (e.g. aero acoustics)

• CPU expensive models (e.g. turbulence, chemical reactions, multiphase)

• Clusters with up to 512 cores (e.g. Formula One teams)

• Mesh and numerical model have a strong impact on the solution

• Large differences in time/space scales

• Post-processing - Interactivity

• Huge amount of data (e.g. fine mesh, transient GB range)

• High computing time (streamlines, contour plot, math. operations)

• Post-processing on multi CPU client server machines


Multiphysics – A New Generation of Simulation

Multiphysics

At Altair

{



Block

Diagram

System

Modeler

ScicosPro



• Several levels of MP technology available to the users so you

can select the right level of coupling.

• Coupled ability in a single solver (AcuSolve + thermal)

• Sequential simulation (AcuSolve + OS for linear FSI))

• Co-Simulation (AcuSolve + RADIOSS or Abaqus for non-linear

FSI)

14


OptiStruct,

RADIOSS

(Mechanics)

OptiStruct,

RADIOSS

(Thermal)

AcuSolve,

HyperXtrude

(Flow)

AcuSolve,

HyperXtrude

(Thermal)

FEKO

(HF Emag)

JMAG

(LF Emag)

ScicosPro

(Controls,…)

[coming soon]

MotionSolve

(MBD)

F-Tire

(Tire

Dynamics)

DesignLife,

FEMFAT

(Fatigue)

DSH Plus

(Hydraulics)

RadTherm

(Human Cft)

AcuSolve

FWH

Simulink,

CD Tire,

RMOD-K Tire

APA Partners

3rd Party


AcuSolve Flow + Thermal Coupling in HW12

• Coupled Flow-Temp

Solve

• Unique to

AcuSolve!

• Nominal coolant flow

rate = 115 liters per

minute (LPM)

• Coolant – constant

properties

• Solids – temperature-

dependent conductivity

• Block - Steel

• Head - Aluminum

• Gasket – Stainless

Steel=

T


Powertrain Boundary Conditions on AcuSolve model

• Boundary Conditions

• Inlet – 105 C, 115 LPM

• Outlet – Pressure = 0.0

• Cylinders – Fixed

Temperatures

• Most Solid Boundaries have

Convective Heat Flux

Coefficient with Sink

(Reference) Temperature

• Outside Surfaces

• 30 W/m^2-C with 30 C

• Major Inner Surfaces

• 100 W/m^2-C with 135 C

• Exhaust Ports

• 623 W/m^2-C with 790 C

• Cylinder Covers

• 625 W/m^2-C with 980 C


Powertrain Model Description

• Surface mesh from SIMLAB, volume mesh in AcuConsole

• Total Mesh Count: ~7 Million nodes / ~39 Million tetrahedral elements

• Fluid: 4. million nodes 24 Million elements (all tet)

• Solid: 3. million nodes,15 Million elements

• AcuSolve accuracy and runtime if f(# of nodes) NOT # of elements

• Solving Fluid Flow / Turbulence / Energy equations

• Solution Time for Flow / Turbulence/Energy about 3.0 hours on

120 cores (10 nodes, 6 core Intel Westmere cpus)

• Effect of

varying

flow rate -

15%

• 115 LPM –

model

max temp

= 271.7 C

• 97.75 LPM

– model

max temp

= 276.0 C


First cylinder in block starved for cooling flow!


AcuSolve + OS or RADIOSS Coupling in HW12

AcuSolve

Optistruct

(linear, static)

Radioss

(non-linear, implicit, dynamic)

P-FSI

(eigenmodes)

DC-FSI

(co-simulation)

Loads

(T,p,…)

Down force for rigid and elastic

4%


New in HW13:

• AcuSolve + MotionSolve Coupling

• AcuSolve/MotionSolve communicate using AcuSolve’s code coupling

interface (CCI)

• Wetted surfaces are “paired” with rigid bodies

• Loads/displacements exchanged at run time


AcuSolve + Motionsolve Coupling

• Example of Riser with Fairing in an ocean current

• Fairing rotates (windmills) about riser with friction between the riser and Fairing

Water

Flow


AcuSolve Mesh Motion for MS Coupling Example


Coupling AcuSolve + Structural

• Interface Conditions

• Traction continuity

• Displacement continuity

• Key Issues

• Spatial coupling

• Non-matching meshes

• Projection and interpolation

• Temporal coupling

• Explicit coupling schemes

• Implicit coupling schemes

Fluid

Solid

pf

tf

:

:

d d

P P

Displacement continuity

Traction equilibrium

on

on

f s fsi

f s fsi


Explicit Coupling Schemes

• Explicit Coupling Schemes

• Conventional Sequential

Staggered (CSS)

(Felippa and Park, 1980)

• Generalized Sequential

Staggered (GSS)

(Farhat et al., 1995)

• Combined Interface Boundary

Conditions (CIBC)

(Jaiman et al., 2007)

• At each time step

Page 24

Apply fluid force

Advance s

tructu

re

2

3

1

Advance flu

id

CFD Solver

4

CSD Solver

t = tn

t = tn+1

Apply fluid force

:

:

*

*

d d + d

P P + P

Displacement prediction

Traction correction

on

on

P

f s fsi

C

s f fsi

Does not work for structures

in liquids


Implicit Coupling Scheme

• Linearized fluid-structure system:

• Off-diagonal terms are not explicitly known

• Explicit coupling effectively has set to zero

• Can we replace by pseudo inverse

• Altair Invented Multi-Iterative Coupling (MIC) Scheme

• Predictor-corrector iterations

• Combined fluid+interface solver

ss sf s s

fs ff f f

A A q R

A A q R

1

sfA1

sfA

sfA

http://www.altairhyperworks.com/html/en-us/rl/ACUSIM/papers/OMAE2009-79804.pdf


Last Slide ! HyperMesh 13.0 (CFD optimization)

• Exhaust system / design challenge

• Significant integration and ease of use improvement in 13.0

Min. pressure drop

Max. uniformity

Morphing

(design space)

initial optimized