Advanced Turbulence Modeling for engine applications
description
Transcript of Advanced Turbulence Modeling for engine applications
1
GM UWCollaborative Research Lab
Advanced Turbulence Modeling for engine applications
Chan Hee SonChan Hee SonUniversity of Wisconsin, Engine Research Center University of Wisconsin, Engine Research Center
Advisor: Professor Christopher J. RutlandAdvisor: Professor Christopher J. RutlandSponsor: General MotorsSponsor: General Motors
2
GM UWCollaborative Research Lab
Motivation
Linear k-Linear k- model model widely used, but compromise between expense and accuracy Inherently unable to account for secondary flows Poor predictions for separated or curved streamline flows
Non-linear modelsNon-linear models Able to predict secondary flow of the second kind Numerical instability leads to excessive computational expense Wallin-Johansson's explicit Algebraic Reynolds Stress Model as
a representative case
vv22-f model-f model Two turbulence scales are used More accurate representation of the physics (eddy viscosity)
close to the wall Very good performance in flow separation regions
3
GM UWCollaborative Research Lab
Model formulation
Turbulence governing equations of vTurbulence governing equations of v22 - f - f
Tt
k
k U k k P
2
1 2T
t U C P Ck k
2 2 2 2v v v vTt
k
U kfk
223
2 21 2
v
f f
k Pf L f C C
T k
1 28
0.251.3 1.9
1 dl
C C
1 20.4, 0.3f fC C
2
0.19 0
v
.3 70LT
L
L C l
C C
C
C
T
1 21.44, 1.92C C
ProductionP
1
2max ,6
kT
13 3 2
2 22
max ,
wall distance
kl C
d
4
GM UWCollaborative Research Lab
Sandia National Lab Optical engine
SpecificationsSpecifications Bore – 79.5mm, Stroke – 85.0 mm CR = 18.7
1500 RPM RS = 1.5 ~ 3.5
Cold flow (no spray or combustion)
Measurement locationsMeasurement locations 3 clusters of 5 points located in a vertical plane bisecting the
exhaust valves The 3 center points are at r= 13.6 mm with all neighboring
measurement points being 1mm away.
5
GM UWCollaborative Research Lab
Radial and tangential velocities @ 5 ATDC with swirl ratio 3.5
v2-f W-J
6
GM UWCollaborative Research Lab
TKE history for case with swirl ratio = 3.5
7
GM UWCollaborative Research Lab
Conclusion
For the Sandia National lab optical engine simulation, W-J For the Sandia National lab optical engine simulation, W-J eARSM does not show any improvement for the mean flow. eARSM does not show any improvement for the mean flow. Even the Even the k-k-model is better.model is better. Potential reason: the W-J ARSM is originally derived for 2D flow. 3D
version is quartic order. Thus, too complex for practical use. Increased levels of turbulence is predicted by the WJ model.Increased levels of turbulence is predicted by the WJ model.
At swirl ratio 2.5 and 3.5, TKE prediction over time is very similar to k-model in trend, but about 50% higher in turbulence level.
This is not due to the ability of this model to capture turbulence anisotropy, as the trend is almost exactly the same as k-t high swirl anisotropy increases.
The The vv22-f-f model consistently shows improved results. Still it fails model consistently shows improved results. Still it fails to catch the trends of the experimental turbulent kinetic energy to catch the trends of the experimental turbulent kinetic energy results.results.