Turbulence Model Effects - Home | NESC Academy Online · 2016. 1. 3. · Israeli CFD Center RANS...

Post on 24-Jan-2021

4 views 0 download

Transcript of Turbulence Model Effects - Home | NESC Academy Online · 2016. 1. 3. · Israeli CFD Center RANS...

Israeli CFD Center

Turbulence Modeling Effects

Aeroelastic Prediction Workshop 2

January 2 2016San Diego

Yuval Levy, Yair Mor-Yossef, and Daniella RavehIsraeli CFD Center, and Technion

Israeli CFD Center

Outline• Overview of turbulence model classes

and models used by workshop analysts• Overview of Reynolds stress models• Effect of turbulence models on one of

the static cases• Case 3A (Mach = 0.85, AOA = 5 deg,

Gas: R-134a)• Effect of turbulence models on one of

the dynamic cases• Case 2 (Mach = 0.74, AOA = 0 deg,

Gas: R-12)• Summary and future outlook

Israeli CFD Center

RANS Turbulence Model Classes

• Linear eddy viscosity models (LEVM)

• Based on the Boussinesq assumption

• Involves the solution of one or more partial differential equations (could also be algebraic)

• Non-linear eddy viscosity models (NLEVM) and algebraic explicit Reynolds stress models (EARSM)

• Based on a linear model as the background model

• Reynolds stress models (RSM, also known as second-moment closure models)

• Hybrid RANS/LES (sometimes also known as DES)

Israeli CFD Center

Turbulence Models Employed

Code Team CFD code Model Class

A FOI EDGE SA LEVM

B Embraer CFD++ SST LEVM

C NASA east FUN3D SA LEVM

D Technion EZNSS SA, SST, TNT LEVM

E UMich SUMad SA LEVM

F ZHAW EDGE, SU2 SA, ? LEVM, EARSM

G ANSYS Fluent SA, SST LEVM

H ATA Loci SA, SST LEVM

Israeli CFD Center

Turbulence Models Employed

Code Team CFD code Model Class

I NRC OpenFoam

J NLR EDGE

K ITA SU2

L NASA west LAVA

M CDADAPCO STAR-CCM+ SST LEVM

N Milano Various None

O Rafael EZAir TNT LEVM

P Strasbourg

Israeli CFD Center

Spalart Allmaras Turbulence Model

• The standard model heavily relies on calibration to a wide range of experimental data

• Has advantage over other models when applied to attached flows

• Suffers from excessive separation at junction flows and has shortcoming when simulating unsteady flows involving considerable separation

• Has many versions, each developed to address certain issues

Israeli CFD Center

Example of Excessive Separation around Junction: Static Case 3A

Israeli CFD Center

k-ω-SST Turbulence Model

• The original version is considered the standard version

• Also has many versions

• The most popular two equation model

• Like all linear models, the main shortcomings are its difficulty to accurately predict unsteady flows involving massive separation and flows involving streamline curvature

Israeli CFD Center

Reynolds Stress Models

• RSM are not based on the Boussinesq assumption and therefore the assumption that the turbulent shearing stress is proportional to the rate of mean strain is dropped

• Exact transport equations for the Reynolds stresses are derived from the Navier-Stokes equation and the models are based on the solution of these equations

• The production term does not require approximations• The production term is primarily responsible for the

anisotropy and the selective response of turbulence to different strain types

• RSM are becoming more affordable

Israeli CFD Center

RSM Varieties Considered

• SSG/LRR-ω (AIAA J. 2015)

• Omega based model

• Uses a blend of two pressure-strain models, the LRR model is activated in the near wall region while the SSG model is activated further away

• MCL (AIAA J. 1999)

• A modification for compressible flow of the Craft-Launder closure model (TCL)

• Employs a cubic pressure-strain model

• Topology free (no need for wall distance)

Israeli CFD Center

Predicted Junction Flow by the SSG/LRR-ω Model: Static Case 3A

Israeli CFD Center

Turbulence Model Effects on Shock Prediction: Static Case 3A

0 0.25 0.5 0.75 1x/c

−1.25

−0.75

−0.25

0.25

0.75

1.25

−Cp

Inviscid splitter (Medium)Inviscid splitter (coarse)Viscous splitter (coarse)Exp

SA-Edwards

y/b = 0.6

Israeli CFD Center

Turbulence Model Effects on Shock Prediction: Static Case 3A

0 0.25 0.5 0.75 1x/c

−1.25

−0.75

−0.25

0.25

0.75

1.25

−Cp

Inviscid splitter (Medium)Inviscid splitter (Coarse)Viscous splitter (Coarse)Exp

SST-2003

y/b = 0.6

Israeli CFD Center

Turbulence Model Effects on Shock Prediction: Static Case 3A

0 0.25 0.5 0.75 1x/c

−1.25

−0.75

−0.25

0.25

0.75

1.25

−Cp

Inviscid splitter (Medium)Inviscid splitter (Coarse)Viscous splitter (Coarse)Exp

SSG/LRR-ω

y/b = 0.6

Israeli CFD Center

Turbulence Model EffectsConvergence of Static Case 3A

0 2000 4000 6000 8000 10000Iterations

−8

−6

−4

−2

0

Mea

n flo

w re

sidu

al

SA modelSST modelSSG/LRR−ω model

Israeli CFD Center

Turbulence Model EffectsConvergence of Static Case 3A

0 2000 4000 6000 8000 10000Iterations

0

0.2

0.4

0.6

0.8

Lift

coef

ficie

nt

SA modelSST modelSSG/LRR−ω model

Israeli CFD Center

Turbulence Model Effects on Static CaseInviscid Splitter Plate, MCL Model, Time-Accurate

0 2000 4000 6000

Time step

0.3

0.4

0.5

0.6

Lift

coef

ficie

nt

Israeli CFD Center

Turbulence Model Effects on Static CaseInviscid Splitter Plate, MCL Model, Time-Accurate

1 1 1 333

0 2000 4000 6000

Time step

0.3

0.4

0.5

0.6

Lift

coef

ficie

nt

4 42 22

Israeli CFD Center

Turbulence Model Effects on Static CaseInviscid Splitter Plate, MCL Model, Time-Accurate

0 2000 4000 6000

Time step

0.3

0.4

0.5

0.6

Lift

coef

ficie

nt

0 0.25 0.5 0.75 1x/c

−1.25

−0.75

−0.25

0.25

0.75

1.25

−Cp

1234

Israeli CFD Center

Turbulence Model Effects on Static CaseInviscid Splitter Plate, MCL Model, Time-Accurate

Israeli CFD Center

Turbulence Model Effects on Static CaseInviscid Splitter Plate, MCL Model, Time-Accurate

Israeli CFD Center

Turbulence Model EffectsFlutter, Mach = 0.74, AOA = 0.0

0 0.5 1 1.5 2−0.05

0

0.05

0.1

0.15

0.2

SATNTSST

First mode

Israeli CFD Center

Turbulence Model EffectsFlutter, Mach = 0.74, AOA = 0.0

0 0.5 1 1.5 2−0.08

−0.07

−0.06

−0.05

−0.04

−0.03

−0.02

−0.01

0

0.01

SATNTSST

Second mode

Israeli CFD Center

Summary

• Results concerning the prediction of the most challenging static case may significantly vary (case 3A, Mach = 0.85, AOA = 5)• All LEVM converge to a steady flow• SA and SGG/LRR-ω fail to predict the correct shock

locations• SST predicts the (average?) shock location• MCL model fails to converge to a steady flow

• Time accurate simulations result in shock oscillations that are similar to the experiment

• LEVM have very little effect on flutter case 2

Israeli CFD Center

Future Outlook

• Recent results show that adding a second nonlinear term to the linear Reynolds stress tensor — the so called quadratic constitutive relation (QCR) — may alleviate the excessive junction flow separation problem

• SA-Edwards-QCR2000

• SST-2003-QCR2000

• Since Reynolds stress models become more affordable, they may provide other means for accurately simulating complex, unsteady, massively separated flows

• Hybrid models?

Israeli CFD Center

Thank You