[TheReckoner]

I am modeling turbulent flow in a simple straight pipe. I have confirmed that this flow is turbulent with a Re of 11,000, however the velocity profile I get looks laminar.

I have tried all k - epsilon models, and the SST model. Still nothing.

What could be wrong? I have been trying to figure this out for days.

[AlexanderZ]

Originally Posted by TheReckoner

I am modeling turbulent flow in a simple straight pipe. I have confirmed that this flow is turbulent with a Re of 11,000, however the velocity profile I get looks laminar.

I have tried all k - epsilon models, and the SST model. Still nothing.

What could be wrong? I have been trying to figure this out for days.

what y+ value do you have?

Best regards

[TheReckoner]

Originally Posted by AlexanderZ

what y+ value do you have?

Best regards

Sorry, what is y+?

[CeesH]

Dimensionless wall distance, relevant for boundary layer development. it shouldn't matter too much when working with the k-e model though, assuming you use wall functions.

Can you show the velocity profile?

[LuckyTran]

Are you simulating the entrance problem? I.e. How long is your pipe?

What are your inlet BC's? Did you give the inlet a reasonable turbulence intensity and length scale?

[TheReckoner]

Originally Posted by CeesH

Dimensionless wall distance, relevant for boundary layer development. it shouldn't matter too much when working with the k-e model though, assuming you use wall functions.

Can you show the velocity profile?

Here is the velocity profile. For calculated Re of about 11,000! https://ibb.co/ngPBxn

[TheReckoner]

Originally Posted by LuckyTran

Are you simulating the entrance problem? I.e. How long is your pipe?

What are your inlet BC's? Did you give the inlet a reasonable turbulence intensity and length scale?

My pipe diameter is 0.37 in, and 10 in long. I have tried it even with 40 in of pipe.

I have not introduced any inlet pressure, just an inlet velocity of 1.143 m/s which is 45 in/s. The working fluid here is liquid water.

I used the setting which asks for intensity (which i calculated to be 5%) and hydraulic length, so my pipe inner diameter of 0.37in.

[CeesH]

It already looks more flat than a true laminar profile. But judging from the image, you have a very crude mesh (6-7 cells in the diameter?), first, refine the mesh. I am quite certain that with something like 20 cells in the diameter the solution would already be quite different. Since gradients are to be expected near the wall, a finer mesh near the wall would be beneficial (this does link to the y+, although when using wall functions your y+ does not have to be nearly as low as without). Also, as LuckyTran mentions, it matters where you measure - keep into account your starting profile is uniform, and will need some time (= distance) to develop to turbulent.

[TheReckoner]

Originally Posted by CeesH

It already looks more flat than a true laminar profile. But judging from the image, you have a very crude mesh (6-7 cells in the diameter?), first, refine the mesh. I am quite certain that with something like 20 cells in the diameter the solution would already be quite different. Since gradients are to be expected near the wall, a finer mesh near the wall would be beneficial (this does link to the y+, although when using wall functions your y+ does not have to be nearly as low as without). Also, as LuckyTran mentions, it matters where you measure - keep into account your starting profile is uniform, and will need some time (= distance) to develop to turbulent.

Okay, thank you!

[TheReckoner]

Originally Posted by CeesH

It already looks more flat than a true laminar profile. But judging from the image, you have a very crude mesh (6-7 cells in the diameter?), first, refine the mesh. I am quite certain that with something like 20 cells in the diameter the solution would already be quite different. Since gradients are to be expected near the wall, a finer mesh near the wall would be beneficial (this does link to the y+, although when using wall functions your y+ does not have to be nearly as low as without). Also, as LuckyTran mentions, it matters where you measure - keep into account your starting profile is uniform, and will need some time (= distance) to develop to turbulent.

I do not think it is 6 to 7 cells, just 6 to 7 divisions in the chart. I made a line across the diameter of the pipe and made a velocity profile of it.

[flotus1]

Too little information.

Show the mesh you used, the solver settings, convergence criteria, where did you extract that line plot, use more points for the plot, use coordinate instead of chart count on the x-axis...
And sorry to bring this up: did you make sure that your imperial units were used correctly throughout the whole simulation process?

[TheReckoner]

Originally Posted by flotus1

Too little information.

Show the mesh you used, the solver settings, convergence criteria, where did you extract that line plot, use more points for the plot, use coordinate instead of chart count on the x-axis...
And sorry to bring this up: did you make sure that your imperial units were used correctly throughout the whole simulation process?

https://ibb.co/mjtrf7
https://ibb.co/g9cWDS
https://ibb.co/kOZHnn
https://ibb.co/hs4TtS
https://ibb.co/dOknL7
https://ibb.co/erqRDS

[flotus1]

Any particular reason why you are using mesh interfaces for this case? If not, drop them.
My psychic abilities tell me that the line you used for post-processing might only evaluate results in the center domain of your mesh. This could be visible if we had a coordinate instead of a chart count on the x-axis

[TheReckoner]

Originally Posted by flotus1

Any particular reason why you are using mesh interfaces for this case? If not, drop them.

I do not really know what that is, I am assuming it was automatic. (I am very new to CFD).

[flotus1]

This should help you get a nice looking mesh without interfaces: Pipe bend meshing using MultiZone

[TheReckoner]

Originally Posted by flotus1

Any particular reason why you are using mesh interfaces for this case? If not, drop them.
My psychic abilities tell me that the line you used for post-processing might only evaluate results in the center domain of your mesh. This could be visible if we had a coordinate instead of a chart count on the x-axis

What do you mean my center domain?

I also forgot to add, I added in streamlines and started an animation and the flow was still not turbulent. :/


Thank you guys for helping me out, much appreciated!

[flotus1]

It seems like your mesh consists of at least two domains. One in the center of the pipe and one around the wall. If your plot only contains results from the center of the pipe, it could kind of make sense. But again, several domains with interfaces are not beneficial here.
We could confirm this theory if you put a coordinate instead of a chart count on the x-axis which we could compare to the dimensions of the pipe.

I also forgot to add, I added in streamlines and started an animation and the flow was still not turbulent. :/

I think there are quite a few misconceptions here. What you did was a steady-state RANS simulation. So transient turbulent fluctuations are not to be expected because you only have one result and there would not be any fluctuations if you had several time steps from an unsteady RANS simulation.

[TheReckoner]

Originally Posted by flotus1

It seems like your mesh consists of at least two domains. One in the center of the pipe and one around the wall. If your plot only contains results from the center of the pipe, it could kind of make sense. But again, several domains with interfaces are not beneficial here.
We could confirm this theory if you put a coordinate instead of a chart count on the x-axis which we could compare to the dimensions of the pipe.


I think there are quite a few misconceptions here. What you did was a steady-state RANS simulation. So transient turbulent fluctuations are not to be expected because you only have one result and there would not be any fluctuations if you had several time steps from an unsteady RANS simulation.

Ah okay, I see what you mean by center domain. You are correct.

Yes, I am assuming steady state.