multiphysics modeling

1

��

��

��

��

M. Cross, T.N. Croft, D. McBride, A.K. Slone, and A.J. Williams

Centre for Civil and Computational Engineering

School of EngineeringUniversity of Wales, Swansea

[email protected]

1

��

��

� !��"#��$��%��#��"��&��"��"��'

� �(�"��

� ��$��#��"��) ��*��

� ��) ��$��%��

� ��$��

� ��%�%��

��

��

��

��

��

��

� �� "��"��"��

� ��$�%��"��$��+

� �,�� &��%��

� �� &��%��

� ��-� ��$��.��./�.�,0,��&��

�� %��'

� ��$��"��#��

� ��%��1��

2

��

��

KKff KKfsfs KKftft

KK == KKsfsf KKss KKstst

KKtftf KKtsts KKtt

aaff

aa == aass

aatt

ffff

ff == ffss

fftt

KaKa = = ff

A single code for allA single code for allphenomena phenomena –– coupling directcoupling direct

��

�(��(�� "��"��

KKffnnaaff

nn = = ffffnn -- gg11((aass

nn--11,,aattnn--11))

KKssnnaass

nn = = ffssnn -- gg22((aaff

nn,,aattnn--11))

KKttnnaatt

nn = = ffttnn -- gg33((aass

nn,,aattnn))

Impact of usingImpact of usingdistinct solversdistinct solversfor each phenomenonfor each phenomenon

�� !!

1

��

��

� ��

� ��!!��

� ��" ��#�� " ��$��

%�&�&��!�� '

� �� " �$�$��(� ��!��!��)�$ �� $��

%�&�&�� '

� �� " ��(� ��!��

%�&�&��!��'

��

��

Phys-A Phys-B

• Interpolation from one setof variables to another =>compatibility of mesh

• Virtual single database of mesh data & simulation variables

• Solver strategy- direct vs iterative- Eulerian vs Lagrangian

• Is coupling strategy compatiblewith scalable parallelism, EVEN if software components are parallel

2

��

��

� �(��*��+ � ��!�$��"��

� ��!��!��#��,��(� ��!��!�� #��$�� !��

� -�� !��!��!�� " ��.��)��./��!�� /

� /��#��0�� !��

� 1�� #��!��#��

��

��

� ��!��2��3��" ��#��-.��-21 ��

� www.scai.fraunhofer.de/mpcci.0.html

� ��#��!�� /!��,

� �4�52��6��72��8�!�9��/

� �8�1� 9�6��81��!�9��/

� �� / ��!��"

��(��#��!��)�#��&�&

3

��

��

� ��

� !"�#�$ � ��%%��&�� &��

� �#'�#�"($� ��%%��&�� &��

� ��

� )*("!� � ��%%��&��&��

� �+ *!("!'*� ��%%��&�� &��

� ("!'*, � ��%%��& ��-��&��

� .��

� �� %%��&��&��

� :��.��

� �.�/*0)� ��%%��&��&��

� ��

� (�� %%��&��&��

��

��#*�� -��$��

� ANSYS/Multi-physics http://www.ansys.com� ABAQUS http://www.abaqus.com� ADINA http://www.adina.com� ALGOR http://www.algor.com� AUTODYN http://centurydynamics.com� CFD-ACE http://www.esi-group.com� DYNA http://www.lsc.com� COMSOL http://www.comsol.com/� LMS software http://www.lms.com� MSC- NASTRAN http://www.mscsoftware.com� PHYSICA+ http://www.physica.co.uk� STAR-CCM+ http://www.adapco.com

1

��

��

� ��

� ��

� �� !��"��

� ��

� #$!�� %��

� �&�$�!�� $��

� !�� $' ��(�)��

� ��

� ��% ��%��

� ��

� �*�+$,#��"��

� ��

� !��

��

��$��(�� -�� )��("�� -��

� ANSYS/Multi-physics http://www.ansys.com� ABAQUS http://www.abaqus.com� ADINA http://www.adina.com� ALGOR http://www.algor.com� AUTODYN http://centurydynamics.com� CFD-ACE http://www.esi-group.com� DYNA http://www.lsc.com� COMSOL http://www.comsol.com/� LMS software http://www.lms.com� MSC- NASTRAN http://www.mscsoftware.com� PHYSICA+ http://www.physica.co.uk� STAR-CCM+ http://www.adapco.com

2

��

� .�� %�� %��(��/� ��0�� (

� ��!��"��#

$ ��(��(��"��%�"��&%��

$ ��-��%��&&��%��

��%�"��%��&%��

$ �("��%��-��1�� %��

��"��'��""��

(��)��&��

$ "��(�� 2��("� "

(��*�+�+�(��"��,

$ ��(��-��("��&%��

Alternative approach:Single Software Framework

��

�("��3��

� -��-.$ ��%��)��%��&%��&��"��

$ ��/��

� �0�1$ 2�*'��0�1$ �3��,

$ ��%��4�&��%��

$ �2�(��)��%��%��

� �56�7 �

$ ��8��*"�"�)4��,�'��*"��,

3

��

9��%�")��: ��$&%��

�2

��

��

��%��!��&��%��&%��

��!��#$5��$��"��$��

��

� ��%��("

� �%��1�!��

� ��)

� 5��"�

� ��"��;&%��%��

� ��

� ��$��

� ��

� <��

� �&��8

� .��

�=.07$�56�7 �

=��01= 0=1�2

��1�..�.

.�� !#��-��

1

��

��

� Mixture of liquid steel and argon injected into rectangular mould

� Liquid metal flux sits on top of mould

� Water cooled mould extracts energy forming a solid steel shell

� Continuous withdrawalB.G. Thomas

��

��

2

��

��

� ��

� � ��!��" ��"!�

��#�"��!��

� ��$�� #��"�

��#��!��!�"��

� ��"!��!"��%��"!��#��"�#��

� &&�"!��!��#'!��"!��& ��&��"

0. =∇+∂∂ φφ u

t

��

z

Solid regions appear in blue

EndView

Top view

��

3

��

��'�''��!�(��"!��&��$)��&&��

L particles

Incorp in contCFD code

Embed withmass flux

effects

Evalintegrated

path

��

��

1

��

� Computations were also performed to estimate the effects of EMB on the free surface . For this the Maxwell equations were solved, which with the usual MHD assumptions, lead to:

� Continuity of magnetic flux:

� Ohm's Law for conducting metals

� Magnetic Transport, or Induction equation

� Lorentz force:

� Note: Terms containing the velocity U, are only important when Rm(=LU/h)> 1

0 = B.∇

φσ ∇× - = E where),BU + E( = J

µσηη

m

2 1 = where,B + )BU( =

t∇××∇

∂∂

egslonea

�� ˘ˇˇˆ � Computations were also performed to estimate the effects of EMB on the free surface . For this the Maxwell equations were solved, which with the usual MHD assumptions, lead to: � Continuity of magnetic flux: � Ohm's Law for conducting metals � Magnetic Transport, or Induction equation � Lorentz force: � Note: Terms containing the velocity U, are only important when Rm (=LU/h)> 1 0 = B . — f s — ¥ - = E where ), B U + E ( = J m s h h m 2 1 = where, B + ) B U ( = t — ¥ ¥ — ¶ ¶

1

��

B=0TB=0.4T

Flow suppressedhere

Fluid behaviour under EMB conditions

��

Coupled EM-flow calculations

� For most practical calculations in metals processing:� The EM field influences the flow and

thermal fields� BUT the thermo-fluid phenomena has little

influence of the EM fields� Hence, essentially one way coupling � So calculate the EM field and calculate the

thermal and flow loads in the CFD calculation

� Can implement above model in any good CFD code!

2

��

Welding processes simulation - natural multi-physics

� Processes involve:� free surface flow� electromagnetic forces� heat transfer with solidification/melting� development of non-linear stress

� Ideal candidate for multi-physics modelling

��

T-Junction arc weld simulation

1

��

Experiment and simulation

T-junction section, highlighting HAZ region

X

Y

Z

Model: T_JCASE1: PHYSICA ResultsStep: 1 TIME: 0Nodal LFNMax = 1 Min = 0

.909E-1

.182

.273

.364

.455

.545

.636

.727

.818

.909

FEMGV 5.1-01 28 FEB 2000Greenwich University

G ERE

N

of

YT

I

WI

UI NV

E

the

HCS

R

theUNIVERSITYofGREENWICH

��

Distortion of T-junction due to heat source

Heat source

2

��

Weld pool dynamics

Velocity vectors in crossectionLorentz force distribution in the weld-pool

��

Distortion of T-junction due to heat source

Distortion

3

��

Welding – multi-physics BUT . .

� Welding involves:� free surface fluid flow� heat transfer and solidification/melting� electro-magnetic fields� non-linear stress

BUT . . no coupling back:� from thermo-fluids to EM field� from stress calculation to thermo-fluids

SO . . reasonably loosely coupled

��

Generic Dynamic Fluid Structure Interaction

� Closely coupled multi-disciplinary problem� Time & space accurate� Very challenging in every respect.� Issue of GCL

� Implementation of boundary conditions.

� Features of single software framework:� Consistency of mesh.� Single database & memory map.� Compatibility in the solution approaches FV-UM.

Traction boundary condition

CFD CSM

DeformationMeshadaptation

1

��

Three Phase Approach

CMD

1+� nn tt

mmm FdK =CSD

( )tsFKddCdM =++ ��

fsm Γon u

fsfsp Γ= on �t

fsΓon d

� ��

� ��

CFD

��

Vertex Based

MeshElement

GaussPoint

Finite VolumeCell Centred

ControlVolume

ControlVolume

Integration Point

x

Finite Element

x

Node

Finite Volume

x

x

x

x

x

xx

xxx

x

PHYSICA

� Spatial Discretisation for closely coupled multi-physics� Unstructured mesh

� CFD� Cell centred� Or mixed CC- VB� FV

� CSM� Vertex based� FV/FE

2

��

Dynamic fluid-structure interaction

� Targeted at problems involving flow induced vibrations

� Use dynamic structural equations and Navier-Stokes flow equations Wind direction Flow induced vibrations

��

Dynamic response of structure without flow

1

NAFEMS World CongressVancouver May 2007

Fluid Velocity and Pressure Movies

At tip of wing


Shear stress σσσσxy in wing

2


Bio-medical multi-physics modelling: the heart!

� Heart – a multi-physics system, featuring interactions between:- electro-chemical system- fluid- structure

Image from www.heartvalves.8m.com


Geometry

� 2D model� Right ventricle

Blood

0.005m

OutletInlet

Wall

0.045m

0.01m 0.01m

0.08m

3


Electro-fluid-structure interaction

Electrical

Structure

Mesh dynamics

Fluid flow


Multi-potential heart electrical activity model: comparison of results

PHYSICA

Clayton and Holden ‘02

Primary and secondary potentials Currents

1


Coupling electrical field to structural mechanics

� Assumption made that small strain model can capture behaviour of heart wall� Dominant behaviour of wall is contraction/expansion � Shearing effects negligible

� Elastic model� Change in potential results in a change in tension

in the heart wall� To model tension we introduce an electric strain

into structural mechanics equations


Electric potential, deformation & flow patterns

Results at various stages through a heart beat cycle

2

��

Parallel Multi-Physics Modelling


Exploiting parallel cluster technology: the challenge

� MpCCI and other filter technologies

� Upside – enables interaction at the code d’base level

� Downside – all data exchange must go via filter and is a compute bottleneck wrtscalability on parallel clusters

Code A (CFD)

Code B(FEA)

Filter tech:

Map onto Parallel cluster

Map ontoParallelcluster

BOTTLENECKBOTTLENECK

3


Parallel Multi-Physics Framework

�� !"�

�#$�% ��&��

'�� "�� (��)�)��

Simulations very Time Consuming – need Parallel capability


Parallelisation approach

� Partition of 3D unstructured mesh by JOSTLE

� Uses mesh partitioning SPMD strategy with non-uniform workload

� Assumes a homogeneous load balance across the mesh:� load balanced ( even no of cells per

node)� minimises sub-domain � interface elements� sub-domain connectivity� matches processor topology of the

parallel system

4


Multi-physics Simulation parallel issues� Sub-domains have

specific physics so partition must reflect this:� non-uniform load/node

� Distinct physics uses distinct discretisation procedures:� secondary partitions

� Sub-domains may change as problem develops:� dynamic load balance

Solid mechanics

Fluid flow

Heat transfer

Strategy needs to address all the above issues


Primary & secondary partitions

Primary & secondarymeshes

Good primary & poorSecondary partition

Good primary &Secondary partitionsfrom JOSTLE

5


Parallel multi-physics: two level approach

� Implement a generic parallel version of Multi-physics code/ MDA codes� without regard to in-homogeneity of the

computational work over the mesh(es) defining the analysis domain

� Dump the load balancing into the mesh (re)partitioning task -JOSTLE_DLB

� Process as straightforward as possible

� �!"�

�#$�% ��&��

Mesh Partition


Metal Forming - Extrusion

� Involves large scale deformation of metal work-piece through interaction with one or more dies

� Multi-physics problem� Flow/deformation of

work-piece� Heat transfer generated

by internal friction � Stress/strain in die(s)

6


Mixed Eulerian-Lagrangian Approach

Workpiece� Eulerian mesh

� Free-surface algorithm to track deformation

� Non-Newtonian material model

� Heat transfer plus energy generated by internal friction

Die� Lagrangian mesh

� Mechanical behaviour coupled with:

� Thermal behaviour in workpiece

� Fluid traction load from workpiece


Governing Equations - Extrusion

� Coupled Thermo - mechanical problem� Heat transfer significant factor in deformation process

� CFD� Non-Newtonian viscosity model – Plastic Norton Hoff law� Heat Transfer - Friction between die and workpiece.� Free Surface - Van Leer method

� CSM� Static equilibrium equation – linear elastic solid.

� Coupling at the workpiece/die boundary:� Die subject to fluid traction boundary condition.� Workpiece subject to a die velocity boundary condition.� Dynamic meshes – GCL. Fluid velocity relative to mesh

movement.

7


Extrusion through U-shaped die

4.76 mm

41.27 mm

47.62 mm

• Initial diameter = 200mm

• Bearing length = 2.5mm

• Punch speed = 5.85E-3m/s

� 63220 elements

� 69507 nodes•• Workpiece = 470Workpiece = 470°°CC

•• Die =Die = 450450°°CC

•• Air = 30Air = 30°°CC


Temperature contours in extruding work-piece

8


Effective stress contours and deformation of die


Parallel results

13.436.11610.927.5128.0310.284.4818.34

181.91

Speed-upRun time(hours)

Processors

� Itanium IA 64 cluster running Linux OS� Eight nodes, two 733MHz processors per node� Each node with 2 Gb memory & 2Gb swap space

Single phase mesh partitions on 16 processors

9


Challenges� Multi-physics OK where

discretisation methods are coherent

� What about distinct methods – FE-BE can be OK

� Continuum-particulate interaction – fracturing

� What happens when it involves combustion, heat transfer, fluid flows, etc

� Can I do this scalably in parallel?

��)��


Conclusion

� Multi-physics simulation is emerging in a commercially supported manner

� Most successful multi-physics is based upon loose or one-way coupling – even then, ‘heroic computing’

� Close coupling in time and space another ball game - Key here are procedures for time & space accurate simulations; DFSI a key exemplar

� Multi-physics essentially compute intensive – leads to challenge of parallel scalability for multi-physics simulation tools

� Can do for bespoke single software solutions, but for multi-code components, not so clear!

� Challenges for the future – integrating components using essentially distinctive model paradigms & solver strategies

multiphysics modeling

Documents

Transcript of multiphysics modeling