Optimising turbulence model for TKE
 

Hello all,

I have experimentally measured in-cylinder velocities inside an engine and trying to model the same using CONVERGE. I used measured intake and exhaust manifold pressures and temperatures as boundary conditions and cylinder pressure to validate the model. After post-processing (in ANSYS EnSight), the velocity levels from the model (in the x-tumble plane) are in good agreement but the TKE values are much lower when compared to the experimentally measured data.

Kindly help me suggesting how can the turbulence (RNG k-eps) model be improved.

Thanks

Hello,

Can you post some plots on how well you match in-cylinder pressure and velocity against measured data? And also post a plot comparing the TKE from simulation and experiments?

Can you also describe how you measure velocity and TKE in the experiments and what simulation quantities (from which files) you use to compare against these data?

Thanks,

If your TKE is too low, maybe your mesh resolution is not sufficient?

Hello Sameera,
Thank you for the reply.
I have tried to answer all your questions to the best of my abilities but in case you need more information about anything, feel free to contact me.

Question: Can you post some plots on how well you match in-cylinder pressure and velocity against measured data?
The figure below shows comparison of cylinder pressure traces obtained using experiment and simulation. The measured and predicted values of peak cylinder pressure are 10.4 bar and 10.7 bar, respectively, with a deviation of 2.9%.
https://www.researchgate.net/profile...ation_W640.jpg

The figure below shows a comparison of the velocity fields obtained using PIV (left) and simulation (right) over the tumble measurement plane at different crank angle positions. Note that the experimental velocity field at each crank angle is ensemble average of 200 instantaneous velocity fields.
https://www.researchgate.net/profile...-over_W640.jpg

Question: And also post a plot comparing the TKE from simulation and experiments?
The figure below shows a comparison of the TKE fields obtained using PIV (left) and simulation (right) over the tumble measurement plane at different crank angle positions.
Note that-
I. The scale for simulation TKE is different in the first three images 84, 114 and 175 CAD i.e. during the intake stroke. This is the main problem as velocities are comparable but TKE values are very low.
II. Beyond which (compression stroke) TKE seems to have more-or-less similar range as that of experiment thought distribution is not matching well.
https://drive.google.com/file/d/1YPP...ew?usp=sharing

Question: Can you also describe how you measure velocity and TKE in the experiments
I have performed laser-based particle image velocimetry (PIV) expriments to measure velocity inside the engine over a tumble measurement plane bisecting the intake and exhaust valves. Note that the grid spacing for vector fields, and hence for TKE, is around 1.9 mm. The 2D velocity fields thus obtained are used to measure TKE as follows -
TKE at each grid location, from the experiments, is calculated from velocity fields as tke = 1/2*(<u'^2>+<v'^2>+<w'^2>) = 3/4*(<u'^2>+<v'^2>). Note that <w'^2> is estimates as (<u'^2>+<v'^2>)/2 as suggested in literature.

Question: what simulation quantities (from which files) you use to compare against these data?
As experimental TKE values are measured along a plane to obtain 2D spatial variation of TKE the CFD model is post-processed in Ensight instead of directly using any output file.
The tumble plane bisecting the intake and exhaust valves is clipped from model to get 2D velocity fields (along that plane) from the simulation as the same had been used during PIV experiments. And the clip plane is colored with velocity magnitude.
Whereas to obtain tke variation from simulation same procedure as velocity field is used (except the clip plane is colored with TKE instead of velocity).

Hello Tobias,
Thank you for the reply.
The grid spacing for velocity field and TKE obtained using PIV experiments is around 1.9 mm.
I have used fixed embedding of scale 2 with the base grid size of 4X4X4 mm in the cylinder region (from IVO to EVO) that gives grid size of 1 mm.

Will it cause any problem?

TKE comparison image
https://www.researchgate.net/profile...31/45_W640.jpg