hybrid urans-les method for internal combustion engine...

1

HYBRID URANS-LES METHOD FOR INTERNAL COMBUSTION ENGINE APPLICATIONS

LES4ICE - RUEIL-MALMAISON 11-12 DECEMBER 2018

HASSAN AFAILAL

2

PLAN

Introduction of common turbulence modeling approaches

Consistent hybrid turbulence model for ICE flows: HTLES

Validation on three cases

Attached boundary layer: Channel flow

Highly separated flow: Hill flow

Flow downstream a fixed valve: Steady Flow Rig

Conclusions and perspectives

3

INTRODUCTION

4

S U S T A I N A B L E M O B I L I T Y

4 | © 2 0 1 8 I F P E N

Huge interest to benefit from the simulation tools by developing accurate modeling approaches with affordable cost

DNS: Solve Navier-Stokes equations for all turbulent length & time scales => expensive

Solution: Limit the resolved scales by using turbulence modeling

2 commonly used approaches:

OVERVIEW OF COMMON MODELING APPROACHES

• All turbulent motions are modeled • Large framework • Affordable • Compatible with wall-functions • Provides only phase averages

• Explicit resolution of turbulent motions larger than the grid size

• Adapted for fully time-dependent ICE flows • Challenging for wall-bounded flows

Introduction Hybrid model: HTLES Results Conclusions & perspectives

Buhl et al. [1]

Possible solution

Use URANS near the wall and LES inside the domain

[1] Buhl, S. (2018). Scale-resolving simulations of internal combustion engine flows (Doctoral dissertation, Technische Universität). Affordable accurate scale resolving simulation

5

HYBRID MODEL FOR ICE FLOWS

HYBRID TEMPORAL LARGE EDDY SIMULATION (HTLES)

6


6 | © 2 0 1 8 I F P E N

PRESENTATION OF THE HTLES METHOD

d

dtρktot = P − ρktotω + D (Eq. 1)

The original k-ω SST URANS model

where all turbulent motions are modelled ktot is the energy relative to integral scales

HTLES

d

dtρkm = Pm − ρ

km

T+ Dm (Eq. 2)

Scale resolving model adaptive to the grid resolution km is the energy relative to the grid resolution

By making the dissipation term in the energy equation sensitive to the temporal filter width T

𝜔 is replaced by 1/𝑇


T =r

1 +Cε2Cε1

− 1 1 − rCε1Cε2

×ktot

ε (Eq. 3)

T is the temporal filter width is computed from:

The ratio r is defined as km

ktot that has to be modelled

ktot = kres + km

HTLES transforms the k-𝜔 SST URANS model (Menter et al. [2]) to hybrid model

The model switches automatically between URANS & LES depending on the grid resolution

[2] Menter, F. R. Influence of freestream values on k-omega turbulence model predictions. AIAA journal, 30(6), 1657-1659.

7


7 | © 2 0 1 8 I F P E N

MODELING THE RATIO R


r = min 1,1

β

Uc ε

ktot3/2

23

ωc

−23

(Eq.4)

r varies between 0 and 1:

The ratio r is modelled using the multi-scale approach by comparing (Manceau et al. [3]):

the highest frequency that can be resolved by the grid resolution

and the frequency of the integral scales of the turbulence

The wall-distance is added to the model using the Elliptic Blending framework [4] which:

Forces the URANS mode near the wall

Guarantees a smooth transition between URANS and LES regions near the wall

0 1

DNS URANS Scale resolving mode

[3] R. Manceau et al. (2010) Recent progress in hybrid temporal LES-URANS modeling, V European Conference on Computational Fluid Dynamics, 1525–1532. [4] Fadai-Ghotbi et al.(2010). Temporal filtering: A consistent formalism for seamless hybrid URANS–LES modeling in inhomogeneous turbulence. International Journal of Heat and Fluid Flow, 31(3), 378-389.

8

M O B I L I T É D U R A B L E

8 | © 2 0 1 6 I F P E N

PRESENTATION OF THE CFD SOLVER

Cell centered CFD code based on Finite Volume methods

Multiphysic solver: Turbulence – Diphasic flows – Combustion – Chemistry …

Automatic meshing for mobile geometries Hexahedral mesh with cut-cell Technique

Automatic mesh refinement (AMR)


(Converge CFD presentation video)

9

• 𝑅𝑒𝜏 = 1 000 • Attached boundary flow

• 𝑅𝑒ℎ = 10 595 • Separated flows &

reattachment • Recirculation

• 𝑅𝑒𝐷 = 30 000 • Complex & industrial flow • Realistic engine configuration

Channel flow Steady flow rig Hill flow

HTLES PHYSICAL VALIDATION & COMPARISON WITH URANS/LES

10


10 | © 2 0 1 8 I F P E N

REMARKS:

All simulations use the same setup:

Focus only on the single impact of the models on the results.


Approach URANS Hybrid LES

Models k-𝜔 SST HTLES 𝜎-model

Numerics 2nd order temporal and spatial schemes

Grid resolution Same grid without AMR

Time resolution Same time step

Near-wall treatment Automatic wall function Werner & Wengle

wall function

11


11 | © 2 0 1 8 I F P E N

CHANNEL FLOW SETUP

Reference data: DNS data of Moser et al. [5]

𝑈𝑏𝑢𝑙𝑘 =28.32 m/s

𝑇 = 320 𝐾

𝑅𝑒𝐷 = 40 000


Case setup:

Height of the Channel Ly = D = 25 mm

Length of the Channel Lx = 10 × D

Width of the Channel Lz = 2 × D

Grid size is set to 0.5 mm in the 3 directions (500x50x100)

𝑦+~30

D

[5] Lee, M., & Moser, R. D. (2015). Direct numerical simulation of turbulent channel flow up to 𝑅𝑒𝜏 ≈ 5200. Journal of Fluid Mechanics, 774, 395-415.

Periodic

Slip wall

Wall (wall function)

Periodic + fixed axial momentum source

Q-criterion iso-surface given by HTLES simulation

12


12 | © 2 0 1 8 I F P E N

COMPARISON OF THE VELOCITY PROFILES

Velocity profiles along the altitude of the channel


LES combined with wall-function over-estimates the flow rate

URANS and HTLES give accurate flow rate compared with DNS-data

13

HILL FLOW

COMPARISON OF HTLES TO URANS AND LES



14


14 | © 2 0 1 8 I F P E N

2D slice

HILL FLOW SETUP

• Hill height h = 28 mm • Length of the domain Lx = 9h

• Height of the domain Ly = 3.035h

• Depth of the domain Lz = 4.5h

• 𝑅𝑒ℎ = 10 595 • Reference data: well-resolved LES data (Temmerman et al. [6])

Slip wall


Periodic Periodic

+ axial momentum

source

Wall: Wall function

Wall: Wall function 𝑦+ < 4

• 5.8 million cells

[6] Fröhlich, J., Mellen, C. P., Rodi, W., Temmerman, L., & Leschziner, M. A. (2005). Highly resolved large-eddy simulation of separated flow in a channel with streamwise periodic constrictions. Journal of Fluid Mechanics, 526, 19-66.

15


15 | © 2 0 1 8 I F P E N

QUALITATIVE ANALYSIS: MEAN VELOCITY STREAMLINES

URANS: limited accuracy Recirculation length is overestimated

The center of the recirculation is shifted away from the hill

HTLES: close results to the reference

Recirculation length is close to the reference

Good location of the recirculation center


UR

AN

S R

efer

ence

W

ell r

eso

lved

LES

HTL

ES

x

x

x

16


16 | © 2 0 1 8 I F P E N

MEAN AXIAL VELOCITY PROFILES OVER THE ALTITUDE

URANS shows discrepancies with the reference data

HTLES gives accurate results in term of:

Separation location

Reattachment location

Farfield velocity profiles

Introduction Hybrid model: HTLES Results Conclusions & perspectives In

flo

w

17


17 | © 2 0 1 8 I F P E N

MEAN TOTAL TURBULENT ENERGY PROFILES ALONG THE ALTITUDE

URANS under-estimates the turbulent kinetic energy profiles

HTLES gives satisfactory kinetic energy profiles compared to the reference data

Introduction Hybrid model: HTLES Results Conclusions & perspectives In

flo

w

18

STEADY FLOW RIG



19


19 | © 2 0 1 8 I F P E N

STEADY FLOW RIG SETUP

Parameters Value

Fluid N2

𝑅𝑒𝐷 30000

Mass flow rate (Kg/s) 0.055

𝑈𝑖𝑛𝑙𝑒𝑡 (m/s) 65

Outlet pressure (atm) 1

References: LDA measurements [7]

Fixed valve

Inlet Fixed flowrate

Outlet: Fixed pressure

Velocity field of the HTLES simulation


[7] Thobois, L., Rymer, G., Souleres, T., & Poinsot, T. (2004). Large-eddy simulation in IC engine geometries (No. 2004-01-1854). SAE Technical Paper.

21


21 | © 2 0 1 8 I F P E N

QUALITATIVE ANALYSIS: MEAN VELOCITY STREAMLINES

Close results between HTLES and LES

The first recirculation at the cylinder head is the same in the 3 simulations

The second recirculation length given by URANS is bigger than HTLES and LES predictions


UR

AN

S H

TLES

LE

S

22


22 | © 2 0 1 8 I F P E N

MEAN VELOCITY PROFILES ALONG THE RADIUS

20 mm

70 mm

velocity profiles are compared at two positions

(URANS) (LES) (HYBRID)

20

mm

• Good agreement with LDA measurements: velocity magnitudes are good located and well estimated

• LES and HTLES show the same radial profiles at 20 mm

• URANS simulations under-estimates

the radial velocity at 𝑟

𝑅∈ [0.6; 1]

• URANS over-estimates the axial velocity in the core region

• LES and HTLES show better assesment of the axial velocity at 70 mm

70

mm


23


23 | © 2 0 1 8 I F P E N

MEAN VELOCITY PROFILES ALONG THE RADIUS ARE BETTER PREDICTED BY HTLES AND LES

20 mm


20

mm

• All the simulations show good agreements of velocity magnitudes • LES and HTLES enhance slightly prediction of the radial velocity at 20 mm at

the recirculation region

24


24 | © 2 0 1 8 I F P E N


70 mm


70

mm

• URANS over-estimates the axial velocity in the core region • LES and HTLES show better results of the axial velocity at 70 mm than

the URANS

26


26 | © 2 0 1 8 I F P E N

PRESSURE PROFILES ALONG THE AXIAL DISTANCE

The head loss is over-predicted by LES simulation

The head loss is predicted correctly by HTLES and URANS simulations

ΔPLES ΔPHTLES


27


27 | © 2 0 1 8 I F P E N

CONCLUSIONS AND PERSPECTIVES


Consistent hybrid model was derived from k-𝜔 SST model

HTLES gave very good results compared to the reference data in three situations Attached boundary layers: channel fow

Highly separated flow: hill flow

simplified ICE configuration: steady flow rig

Results showed that HTLES combines the advantages of URANS and LES 1. Accurate prediction of the attached boundary layer flows by using URANS near the wall

2. Good predictions of separated flows by using the scale resolving mode far from walls

3. Fluctuation predictions as accurate as LES

Perspectives:

Explore the mesh dependency of the method

Extend the method to time-dependent flows and apply it to unsteady in-cylinder flows

hybrid urans-les method for internal combustion engine...

Documents