scieee Science in your language
[en] (orig)

Influence of the turbulence representation at the inlet on the downstream flow pattern in LES of backward-facing step

Abstract

The impact of the choice of inlet boundary condition treatment on the fluid flow is studied in this work. The correct representation of the turbulence at the inlet to the domain is essential for the accuracy of Large Eddy Simulation. The inappropriate specification of the inlet velocity has significant effect on the downstream flow pattern. The case of backward-facing step was used as a test case. Three different approaches of the inlet boundary conditions were studied: uniform velocity profile, mean velocity profile of the fully developed channel flow and velocity obtained from mapping velocity from plane positioned behind the inlet back to the inlet. The results of the simulations were compared with experimental results. It has shown, that using uniform velocity profile on inlet and even prescribing turbulent mean velocity profile without proper representation of turbulence fluctuations leads to unrealistic results.

Read accessible full text

Influence of the turbulence representation at the inlet on the downstream flow pattern in LES of backward-facing step

Author: Volavý, Jaroslav; Forman, Matěj; Jícha, Miroslav
Publisher: EDP Sciences
Year: 2012
DOI: 10.1051/epjconf/20122501099
Source: https://dspace.vut.cz/bitstreams/d487620c-36aa-48e7-aa8e-627957046067/download
Influence o he u bulence ep esen a ion a he inle on he
downs eam flow pa e n in LES o backwa d- acing s ep
Ja osla VOLAVY*, Ma ej FORMAN**, Mi osla JICHA***
Abs ac : The impac o he choice o inle bounda y condi ion ea men on he fluid
flow is s udied in his wo k. The co ec ep esen a ion o he u bulence a he inle
o he domain is essen ial o he accu acy o La ge Eddy Simula ion. The inapp o-
p ia e specifica ion o he inle eloci y has significan e ec on he downs eam flow
pa e n. The case o backwa d- acing s ep was used as a es case. Th ee di e en
app oaches o he inle bounda y condi ions we e s udied: uni o m eloci y p ofile,
mean eloci y p ofile o he ully de eloped channel flow and eloci y ob ained om
mapping eloci y om plane posi ioned behind he inle back o he inle . The esul s
o he simula ions we e compa ed wi h expe imen al esul s. I has shown, ha using
uni o m eloci y p ofile on inle and e en p esc ibing u bulen mean eloci y p ofile
wi hou p ope ep esen a ion o u bulence fluc ua ions leads o un ealis ic esul s.
1INTRODUCTION
In he simula ion o u bulen flow, he mos impo an ac o seems o be p ope choice o u bulence
model ha will p o ide good ep esen a ion o he flow. Ano he issue o g ea impo ance is specifica ion
o bounda y condi ions, especially inle bounda y condi ion. The eloci y and ano he inflow da a linked
wi h u bulence p esc ibed by bounda y condi ion should be consis en wi h chosen u bulen model.
Desc ip ion o mean eloci y p ofile and o he u bulence a iables ( u bulen kine ic ene gy, ...) ob ained
om analy ical solu ion o om expe imen s can be ega ded as su ficien o he RANS simula ions.
This app oach was jus ified in [1]. I was shown ha RANS model each uni e sal asymp o ic beha io
i espec i e o he ini ial condi ions. Fo La ge Eddy Simula ion, whe e he field on he inle is u bulen ,
he si ua ion is mo e p oblema ic. Desc ip ion o he flow is usually limi ed by knowledge o s a is ical
quan i ies such as mean eloci y p ofiles and mass fluxes. In LES, he da a gene a ed by inle bound-
a y condi ion should include an uns eady u bulen eloci y signal ep esen ing u bulence a he inle .
Ideally, he simula ion o ups eam flow en e ing he compu a ional domain will p o ide a good flow ep-
esen a ion. Howe e , unlimi ed ex ension in he ups eam di ec ion is no possible because o high
compu a ional cos . The e o e app oxima e inle condi ions mus be specified.
A lo o me hods we e de eloped in he pas . Kal enbach e al. [2] used ecycling me hod. Tu bulen
fluc ua ions a e specified by unning p ecu so simula ion, whose only ole is o p o ide main simula ion
wi h accu a e bounda y da a. Ano he app oach is o add syn he ic u bulence a he inle . Lund e
al. [3] modified inle eloci y field by adding a andom e m o all eloci y componen s. Klein e al. [4]
p oposed digi al signal p ocessing p ocedu e o emedy lack o la ge- scale dominance in he inflow da a
gene a ed by he andom me hod.
*Ja osla Vola y, Facul y o Mechanical Enginee ing, B no Uni e si y o Technology, Technická 2896/2, 616 69 B no, Czech
Republic, ja osla . [email p o ec ed]
**Ma ej Fo man, Facul y o Mechanical Enginee ing, B no Uni e si y o Technology, Technická 2896/2, 616 69 B no, Czech Re-
public, o man@ me. u b .cz
***Mi osla Jicha, Facul y o Mechanical Enginee ing, B no Uni e si y o Technology, Technická 2896/2, 616 69 B no, Czech
Republic, [email p o ec ed].cz
EPJ Web o Con e ences , 010 (2012)
DOI: 10.1051/epjcon /201225010
© Owned by he au ho s, published by EDP Sciences, 2012
This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License 2.0, which
pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed.
A icle a ailable a h p://www.epj-con e ences.o g o h p://dx.doi.o g/10.1051/epjcon /20122501099
2GOVERNING EQUATIONS
Fo he solu ion o he fluid flow in his a icle was chosen La ge Eddy Simula ion. The main idea o
La ge Eddy Simula ion is o sepa a e la ge scales (g id-scales) om small scales (subg id-scales) o
lowe compu a ional cos . The subg id scales a e modelled using subg id model. The scale sepa a ion
is done by applying fil e ope a o on Na ie -S okes equa ion. I we apply he fil e ope a o on Na ie -
S okes equa ions we ob ain fil e ed Na ie -S okes equa ions:
∂¯ui
∂ +∂
∂xj
(¯ui¯uj)=−1
ρ
∂¯p
∂xi
+ν∂2¯ui
∂xk∂xk
−∂τij
∂xj
.(1)
Fo e alua ion o subg id s ess enso τij is used Smago insky model:
τij −1
3δijτkk =−2ν ¯
Sij,ν
=(CSΔ)2|¯
S|(2)
¯
Sij =1
2∂¯ui
∂xj
+∂¯uj
∂xi(3)
whe e |¯
S|=|2¯
Sij ¯
Sij|1/2.
3DESCRIPTION OF THE TEST CASE
Fo es ing o di e en app oaches o he inle bounda y condi ion was chosen backwa d- acing s ep
flow. This case o flow is good benchma k o alida ing a ious models because i con ains massi e
sepa a ion and consequen e-a achmen . Sepa a ion bubble on he wall opposi e o he s ep could
appea o some geome ical and flow pa ame e s.
The geome y and flow p ope ies was chosen acco ding o he expe imen al s udy done by [6]. The
Reynolds numbe o he inle channel flow was 13 800, based on he bulk eloci y o 10 m/s and he
channel hal -wid h o 20 mm. The expansion a io was 5/3. The flow pa ame e s o he inle channel and
backwa d acing s ep flow a e in able 1.
Table 1: Flow pa ame e s
Channel flow Backwa d- acing s ep flow
Channel hal -wid h, h20 mm S ep heigh , H26.7 mm
Channel bulk eloci y, U010 m/s Expansion a io 5:3
Reh=U0h
ν13 800 ReH=U0H
ν18 400
uτ, ic ion eloci y 0.5 m/s τ , la ge eddy ime scale, 5H/U012.7 ms
Viscous leng h scale 31 μm
Kolmogo o leng h scale, η170 μm
The compu a ional mesh consis s o 2.1 million hexahed al cells. The mesh is block-s uc u ed and
becomes fine owa ds he wall in o de o sa is y condi ion o y+≈1. The de ail o he mesh nea he
ailing edge is in he Figu e 1.
EPJ Web o Con e ences
01099-p.2
Figu e 1: The de ail o he mesh nea he ailing edge.
4INLET BOUNDARY CONDITION
Fo he gene a ion o he eloci y on he inle o he domain we e used h ee di e en app oaches. Le
hem deno e as Case A, B and C. All cases ha e bulk eloci y o 10 m/s.
Case A: Uni o m eloci y p ofile is p esc ibed a he inle o he compu a ional domain. No eloci y
fluc ua ions a e p esen .
Case B: Mean eloci y p ofile o ully u bulen channel flow is p esc ibed on he inle . No eloci y
fluc ua ions a e p esen .
Case C: Fo he gene a ion o he eloci y on he inle o he domain was used di ec mapping app oach
[2]. The eloci y on he inle is ob ained by di ec mapping o he eloci y om he plane wi h 3h(60
mm) o se om he inle plane. The schema ic pic u e o he mapping is shown in he Figu e 2 ( he
placemen o he mapping plane in his figu e is only schema ic). Fo he flow ini ializa ion was used
ollowing p ocedu e: Fi s was done simula ion o inle channel only using pe iodic bounda y condi ion.
Fo as e ansi ion o he ully de eloped u bulen flow was used o cing scheme based on O ns ein-
Uhlenbeck p ocess p oposed by [7]. When he ully u bulen egime was eached hen he esul s o
his p e-simula ion was mapped o he inle channel o he backwa d- acing s ep geome y.
Figu e 2: Scheme o inle da a mapping.
5RESULTS
In his sec ion a e desc ibed esul s ob ained by a ious app oaches o he inle bounda y condi ion.
The simula ions we e s a ed om ze o eloci y ini ial condi ion o cases A and B. The ini ial condi ion
o case C could be seen in he Figu e 2. The simula ions an o 0.5 s in o de o allow de elopmen
o u bulen s uc u es in he domain. Then a e aging was u ned on and he simula ions con inued o
ano he 1 s. This ime was long enough o each s a is ical s eady s a e. The cen e o he coo dina e
sys em is posi ioned in he middle o he ailing edge.
EFM11
01099-p.3
In he figu e 3 is ins an aneous eloci y magni ude o uni o m eloci y p ofile a he inle (Case A). F om
his pic u e is e iden ha he eloci y fluc ua ions a e gene a ed a e he sepa a ion o he s eam
(behind he ailing edge). The lack o hese fluc ua ions leads o he o ma ion o he sepa a ion bubble
on he op o he domain. The o he consequence o he missing fluc ua ions is di e en shape o he
eci cula ion egion. Fo he mean eloci y p ofile o ully u bulen channel flow desc ibed a he inle
(Case B) is si ua ion simila . The ins an aneous eloci y magni ude o ully de eloped u bulen flow a
he inle (Case C) is in he pic u e 4. The flow pa e n is much di e en om he p e ious one. The e is
no sepa a ion bubble on he op and he ea achmen poin is much u he om he s ep.
Figu e 3: Ins a eneous eloci y o uni o m in-
le p ofile (Case A)
Figu e 4: Ins a eneous eloci y o ully u bu-
len flow a inle (Case C)
In he figu e 5 a e eloci y p ofiles in sho dis ance om he s ep. I could be no iced ha he uni o m
inle eloci y gi es non ealis ic esul s. F om he fla ness o he p ofile abo e he s ep could be s a ed
ha he leng h o he inle channel is no long enough in o de o de elop u bulen eloci y p ofile.
Veloci y p ofiles u he om he s ep (x/H = 5) a e in he figu e 6 . Nega i e alues o eloci y a he
op o he domain p edic ed in cases A and B indica es ha sepa a ion bubble has been o med in his
egion. I is because o lack o wall-no mal eloci y fluc ua ion, which leads o absence o u bulen
mixing mechanism and he main s eam is bend down due he s ong ad e se p essu e g adien . Case
A also ails in p edic ion o e-a achmen o he flow behind he s ep.
−0.2 0 0.2 0.4 0.6 0.8 1 1.2
−1
−0.5
0
0.5
1
1.5
y/H
u/U0
x/H=2
Case A
Case B
Case C
Expe imen
Figu e 5: Veloci y p ofile a posi ion x/H=2
−0.2 0 0.2 0.4 0.6 0.8 1 1.2
−1
−0.5
0
0.5
1
1.5
y/H
u/U0
x/H=5
Case A
Case B
Case C
Expe imen
Figu e 6: Veloci y p ofile a posi ion x/H=5
Figu es 7 and 8 show si ua ion a behind he s ep. Case B also ails in p edic ion o e-a achmen . On
he o he side, case C p edic s e-a achmen poin qui e well. Re-a achmen poin p edic ed by case C
is a posi ion x/H = 9.2 (expe imen p edic s his poin a posi ion x/H=8.4).
6CONCLUSIONS
The La ge Eddy Simula ion o backwa d- acing s ep was done. The simula ions we e done o h ee
di e en ep esen a ion o inle bounda y condi ion: uni o m eloci y p ofile, u bulen eloci y p ofile and
di ec mapping o eloci y om plane behind he inle . Only he las case gi es ealis ic esul s because
is capable ep esen also u bulence eloci y fluc ua ion. This has shown c i ical o he accu acy o he
EPJ Web o Con e ences
01099-p.4
−0.2 0 0.2 0.4 0.6 0.8 1 1.2
−1
−0.5
0
0.5
1
1.5
y/H
u/U0
x/H=7
Case A
Case B
Case C
Expe imen
Figu e 7: Veloci y p ofile a posi ion x/H=7
−0.2 0 0.2 0.4 0.6 0.8 1 1.2
−1
−0.5
0
0.5
1
1.5
y/H
u/U0
x/H=9
Case A
Case B
Case C
Expe imen
Figu e 8: Veloci y p ofile a posi ion x/H=9
simula ion. The fi s wo cases we e no able o p edic hese fluc ua ions wha esul ed in he inclina ion
o he incoming s eam downwa ds and o ming a sepa a ion bubble a he op o he domain. This is
no in acco dance wi h eal expe imen al s udy. The inclina ion o he s eam also a ec ed he posi ion
o he e-a achmen o he s eam behind he s ep, he s eam is e-a ached oo ea ly.
The only p ope way o ep esen a ion o he inle da a p esen ed in his wo k is unning p ecu so
simula ion o he inle channel using u bulence o cing and hen mapping esul ing fields o he main
domain and di ec mapping o eloci y o he inle . This is mos complica ed case bu gi es accu a e
esul s close o he expe imen . The p edic ion o he e-a achmen poin could be imp o ed by using
mo e ad anced subg id model, o example some a ian o localized Smago insky model [7].
ACKNOWLEDGMENT
The suppo o g an s GA CR 101/08/0096 as well as he p ojec FSI-S-11-6 is g a e ully acknowledged.
REFERENCES
[1] Geo ge W.K., Da idson L.: Role o ini ial condi ions in es ablishing asymp o ic flow beha io , AIAA
Jou nal, 42(3), 2004, 438446
[2] Kal enbach H.J., Fa ica M., Mi al R., Lund T.S., Moin P.: S udy o flow in a plana asymme ic di use
using la ge eddy simula ion, Jou nal o Fluid Mechanics, 1999, 151185
[3] Lund T., Wu X., Squi es D.: Gene a ion o u bulen inflow da a o spa ially-de eloping bounda y
laye simula ions, Jou nal o Compu a ional Physics, 1998, 233258
[4] Klein K., Sadiki A., Janicka J.: A digi al fil e based gene a ion o inflow da a o spa ially de eloping
di ec nume ical o la ge-eddy simula ions, Jou nal o Compu a ional Physics, 186, 2003, 652665
[5] Ghosal S., Lund T., Moin P., Aksel oll K.: A dynamic localiza ion model o la ge-eddy simula ion o
u bulen flows, J. Fluid Mech., 286, 1995, 229-255
[6] Fessle J.R., Ea on J.K.: Tu bulence modifica ion by pa icles in a backwa d- acing s ep flow, J. Fluid
Mech., 394, 1999, 97-117
[7] Vola y J., Fo man M., Jicha J.: Tu bulence o cing scheme in physical space based on O ns ein-
Uhlenbeck p ocess, CFD 2010, Lisbon, ISBN 978-889-96778-0-7, ECCOMAS
EFM11
01099-p.5