What is CFD and Why?
yiQPIQGRX gph veture I
1.1
Overview
gomputtionl pluid hynmis { use of omputers to study uid dynmisF olve the xvierEtokes iqutions @xiA X
r:u a H Du a rp C r2 u C F Dt
R equtions for R unknownsD oupledD nonElinerD nd often involve omplex geometriesF
R stge proess IF wodelling
xE equtionsD urulene gomustionD multiphseD shoks : : :
PF hisretistion QF olvingF RF snterprettion
u aq p
u a w1 q p
1.2
User's perspective
prom the user9s point of viewD gph modelling is Q stge proess X IF Problem de nitionF
geometryD meshD oundry onditions physil modelsD onstnts
PF Numerical solutionF
dierening shemesD soln tehniquesF tolernes tully running the solver
QF PostprocessingF
1.3 FD vs. FV (vs. FE)
pinite hierene methods { vriles represented t disrete points in speF herivtives evluted p p @p a j +1 j 1 @x Px { very omplex if the grid distorted or omplitedF pinite ilement @piA methods hve een usedD ut re not ommonF pinite olume @pA method most ommonF { used in most ommeril odes { pluentD eEghD gp etF
1.4 FV Method
ow region is divided into smll oxes {
mesh
cells
or control volumesD forming
equtions reexpressed in terms of ow into nd out of eh ellF his is done y integrting the equtions over the volume of eh ellF
he result is set of dierene equtions whih n e solved numerilly s eforeF he dvntge of doing things this wy is tht the ells n e ny shpe required E uesD tetrhedrD distorted uesD or more omplited struturesF
1.5
FV for transport equation
ter owing in river is ontminted y hemil lekF hetermine where the hemil hs rehed fter given time tF
u u
w W P e E
plow uniform throughout the hnnel X onstF speed u t ll pointsF gonentrtion q of ontminnt
@q @t
@uq @x
@ @x
@q @x
{ the IEd trnsport equtionF o pply the p methodD we X IF plit the hnnel into ells V a Ax PF integrte trnsport eqn over the ellF pirst term X
ZZZ
dV
@q
V @t
dV
d dt
ZZZ
V
qdV
d dt
@qV A
eond term X
ZZZ
a Adx
dV
@uq
V
@x
ZZZ
V
@uq @x
Adx
a @uqAAe @uqAAw
hese two terms re now eing evluted on the oundry fe etween the ellsF rowever we do not know the vlue of the vriles on w nd eF xeed to interpolate to get these vluesF pwind dierening @uqAw a @uqAW ; Q @uqAe a @uqAP
gentrl dierening @uqAW C @uqAP P hiusion n e treted similrly hisretised equtions X @uqAw a
d dt
@qV A C @uqAAe @uqAAw a
@q @x
e
@x A
@q
w
ernging it nd using entrl dierening to interpolteD we get
dq dt
Px
u
@qE qW A a
x2
@qE PqP C qW A
xumer the ells i a H : : : N
dqi dt
3X
x2
Px
@qi+1 qi1 A C
@qi+1 Pqi C qi1 A
1.6
Temporal discretisation
his prolem is parabolic one { solve y timestepping { disretistion of time into M timesteps qj F ime derivtive dq q j +1 q j a et wht timestep re the vlues on the rFhFsF evlutedc
dt t
ke vlues t timestep j D 3 ke vlues t timestep j C ID 3
ixpliit sheme X
qi
j
explicit
shemeF shemeF
implicit
+1 qj a
i
@q q Aj Px i+1 i1 t C 2 @qi+1 Pqi C qi1 Aj
x ut x ; D
ut
rite
C
x2
t
C a Courant number { signint prmeter in determining the stility of the shemeF
reltes to the diusionF
riting this out s n lgorithm X
j qi
+1 a
C
P
CD
j qi
1 C @I PDA
j qi
C
P
qi+1
e rule for dvning the vlues of q through one timestepD whih ould e written into spredsheetF smpliit sheme X write s mtrix eqution X
0F F B F B D B B L B B H @
U D L
U D
FFF
10 F 1 0 F 1 F F C B jF+1 C B jF C C Bqi1 C Bqi1 C C B j+1 C B j C C Bqi C a B qi C C B j+1 C B j C C Bqi+1 C Bqi+1 C A@ A @ A
F F F F F F
snverting this mtrix provides the solution without the stility prolemsF @lE though the gournt numer is still worth lulting if there re prolemsAF
1.7 3d transport equation
his sheme ould e derived y stndrd ph tehniquesF his is euse the mesh we re deling with ws regulr nd uniformF rowever this is not lwys the seF p pproh mkes it muh esier to evlute the vrious terms for irregulr shped ellsF he Qd trnsport eqution X
@q @t
C r:uq a r:rq C Sq
hen we integrte the seond term hereD we nd
ZZZ
V
r:uq dV a
X
A
uq
{ Gauss' theorem in vetor lulusF
