[Ames]

Hi all,

I am currently running simulations in axisymmetric stenosis models (up to 90% diameter reduction) in Ansys Fluent. The domain is 10cm long and has a maximum radius of 1.6mm. I assume the fluid is incompressible. At the inlet I prescribe a parabolic velocity profile with maximum velocity 0.66 m/s. The outlet is stress-free and for the wall a no-slip condition is applied. I am using the Hybrid Initialization in Fluent.

In the region after the stenosis I observe relatively high Reynold's numbers (2600-3500), as the fluid is highly accelerated through the narrowing. I was running steady simulations (Time set to Steady in General Tab) with viscous model = laminar. As the value of Re is high in the region after the stenosis, I had some doubts about whether using the laminar viscous model was correct. My supervisor suggested to check this by running transient simulations with still the laminar model, to see if the transient simulations reach a steady-state. He rejected the idea of looking into turbulent/transition models.

Could someone elaborate on why using a transient simulation could show if my simulation is still in the laminar regime or already in the transitional/turbulent regime? Do I understand it correctly if by transient is meant to set Time in the General Tab to Transient?

Thanks in advance!

[LuckyTran]

transient does indeed mean set time to transient in general tab.

I don't think transient has anything to do with a search for turbulence. Your supervisor probably just wants to check the quality of your simulation and see if you reached the proper steady state solution.

Your supervisor probably isn't intersted at all in the turbulent solution, and I wouldn't be either. Turbulent flows are just soo impractical in these problems. As soon as transition to turbulence occurs, the pressure drop increases dramatically. The pumping vessel (the heart) cannot support this pumping power for long. That's how you end up dead. In a ghastly way... the turbulent solution is the trivial solution...

[Ames]

Originally Posted by LuckyTran

transient does indeed mean set time to transient in general tab.

I don't think transient has anything to do with a search for turbulence. Your supervisor probably just wants to check the quality of your simulation and see if you reached the proper steady state solution.

Your supervisor probably isn't intersted at all in the turbulent solution, and I wouldn't be either. Turbulent flows are just soo impractical in these problems. As soon as transition to turbulence occurs, the pressure drop increases dramatically. The pumping vessel (the heart) cannot support this pumping power for long. That's how you end up dead. In a ghastly way... the turbulent solution is the trivial solution...

Hi LuckyTran,

Thank you for your answer, I appreciate it very much! For Re = 3500, you also think that is right to still use the laminar viscous model? I am interested in the pressure drop over the domain. So if the difference in pressure drop between the steady-state and transient solution is small (~1%?), then I can assume that the steady state solution was fine?

I am wondering if my flow were in the transitional/turbulent regime, does a steady state solution actually exist?

[LuckyTran]

If the Reynolds number was 1 million, then certainly some type of turbulence model would be necessary. But it is really silly (not stupid, just silly) to simulate turbulent stenosis in general. So why? Academic curiosity?

To check for steady state, you need to check/plot your solution over time and show that it is asymptotic w.r.t time. If your steady simulation was properly converged, the solution won't go anywhere and you'll see that it is just constant (but that probably won't happen because the devil is rolling your dice).

Originally Posted by Ames

I am wondering if my flow were in the transitional/turbulent regime, does a steady state solution actually exist?

Yes and no.

No because it won't be steady state in the classical sense, it will be statistically stationary. If you want to be politically correct, then start using the term statistically stationary.

Yes because you will be doing RANS (the RA in RANS stands for Reynolds averaging). When you do RANS/URANS, you are solving for the statistically averaged variable. So despite turbulence and transition to turbulence being inherently chaotic processes, the Reynolds-averaged velocity (which is what you compute when you do RANS/URANS) will still be statistically stationary.

But actually, a transient laminar calculation is like DNS (except your grid is probably too coarse and schemes not accurate enough) and you could pick up some turbulence when you do this. So if you find your laminar calculation is unsteady, it could be due to turbulence. That could've been your supervisor's idea as well. Idk, you might want to ask them.

[Ames]

Originally Posted by LuckyTran

To check for steady state, you need to check/plot your solution over time and show that it is asymptotic w.r.t time. If your steady simulation was properly converged, the solution won't go anywhere and you'll see that it is just constant (but that probably won't happen because the devil is rolling your dice).

I have finished some transient simulations for the cases with the highest post-stenotic Reynold's number. I use CFD-Post for postprocessing. When I plot the velocity over time in several points in the domain, I see some start-up effects and then a flat horizontal line. Is it correct to say that there is a steady state?

Originally Posted by LuckyTran

But actually, a transient laminar calculation is like DNS (except your grid is probably too coarse and schemes not accurate enough) and you could pick up some turbulence when you do this. So if you find your laminar calculation is unsteady, it could be due to turbulence. That could've been your supervisor's idea as well. Idk, you might want to ask them.

So it could be that in my case the mesh is too coarse so that I can't capture some turbulence/unsteadiness in my simulations?

I will definitely discuss this with him, but with this forum I was looking for a better understanding of his suggestion! Thanks a lot for clearing things up on RANS/DNS models!