[nskelly]

Hi all,

I was wondering whether anyone could help me regarding DNS simulation of the flow between the cone and plate.

I aim to look at the development of turbulent flow within an upper cone and lower plate (very similar to Taylor-Couette flow). At a certain Reynolds number (based upon the dimensions of the system) the flow should transition from laminar to turbulent flow at R = 4 (equivalent to Re = UL/v = 1375). However, no dominant turbulent fluctuations are present. So I simulated at a higher R = 12.3 (Re = 4230) which should be fully turbulent. To my joy, turbulent fluctuations were evident, but they died out and the flow looked laminar. So I am not too sure what going on.

Case set-up:
The geometry is relatively simple and I have made in blockMesh and only simulated a 45-degree wedge of a whole disc of fluid (see image). I apply rotatingWallVelocity at a certain RPM to simulate the rotation of the fluid and no-slip on the lower plate surface and slip on the outer edge as this a free surface. Zero gradient for pressure on all walls is used. Do I need to prescribe some sort of turbulent intensity? I am not sure how I can that with the rotatingWallVelocity BC I am using.

My mesh is 2.5 million elements with x+=0.566, y+=0.340 and z+=2.28 (For R=4). Do I need to go finer at the walls? I am using the discretization schemes similar to the dnsFoam solver as I have not done DNS before so was a bit unsure of the high order schemes would be best.

I guess my main questions are regarding what aspects of the case would be the main issue i.e. BCs, discretization, mesh etc.

Any help would be great and I'll happily upload my case file if needed.

[syavash]

Originally Posted by nskelly

Hi all,

I was wondering whether anyone could help me regarding DNS simulation of the flow between the cone and plate.

I aim to look at the development of turbulent flow within an upper cone and lower plate (very similar to Taylor-Couette flow). At a certain Reynolds number (based upon the dimensions of the system) the flow should transition from laminar to turbulent flow at R = 4 (equivalent to Re = UL/v = 1375). However, no dominant turbulent fluctuations are present. So I simulated at a higher R = 12.3 (Re = 4230) which should be fully turbulent. To my joy, turbulent fluctuations were evident, but they died out and the flow looked laminar. So I am not too sure what going on.

Case set-up:
The geometry is relatively simple and I have made in blockMesh and only simulated a 45-degree wedge of a whole disc of fluid (see image). I apply rotatingWallVelocity at a certain RPM to simulate the rotation of the fluid and no-slip on the lower plate surface and slip on the outer edge as this a free surface. Zero gradient for pressure on all walls is used. Do I need to prescribe some sort of turbulent intensity? I am not sure how I can that with the rotatingWallVelocity BC I am using.

My mesh is 2.5 million elements with x+=0.566, y+=0.340 and z+=2.28 (For R=4). Do I need to go finer at the walls? I am using the discretization schemes similar to the dnsFoam solver as I have not done DNS before so was a bit unsure of the high order schemes would be best.

I guess my main questions are regarding what aspects of the case would be the main issue i.e. BCs, discretization, mesh etc.

Any help would be great and I'll happily upload my case file if needed.

Hi,

I guess you should describe your problem more clearly, particularly about the boundary conditions. I personally cannot figure out the whole image of your configuration. You may address some reference paper/report for the case.

Regards,
Syavash

[nskelly]

Hi Syavash,

I will explain the problem a bit better.

So I am simulating the flow between a cone and plate rheometer. A paper (link at bottom) I have read has looked at this configuration and observed that once we go past a certain modified Reynolds number (R tilde) based upon the cone length scales (see image for modified Reynolds number definition) we enter turbulent flow.

Turbulent flow should begin when R tilde > 4, however, my simulations do not show that. I then tried to simulate a much higher R tilde = 12 (x3 greater) and some turbulent flow behaviour was seen in contours and from monitoring the velocity fluctuations through time. However, the fluctuations died down and the flow did not seem turbulent.

I applied the boundary conditions in the image provided. But I am unsure whether I have to apply some sort of turbulent intensity/perturbation to the flow field initially. And if I have to, I am not sure how.

I hope I made that a bit clearer.

Paper link: https://www.cambridge.org/core/journ...73A2014BBCD841