Divergence for compressible flow with rough walls

turbvisc · September 16, 2022, 11:32

Hello everyone,

I'm currently doing a heat-transfer analysis an a nozzle with compressible flow with the k-w-SST turbulence model. Therefore I want to include rough walls with a roughness height of 129µm. But as soon as I include them, my solutions never converge and I get extreme values in the boundary layer. An inflation layer with an initial first cell height of 5,67e-3mm is included, but I also tried lower and larger values.

I'm running the pressure-based solver since I never got the solution to converge with density-based.

Other settings are coupled scheme, second order discretization, pseudo transient. But I also tried other inputs also for the relaxation factors, which are currently set to default.
I included a photo for rough walls and for smooth walls, all other settings are the same for both simulations.
With lower roughness heights (20µm) I reached convergence and plausible results.

If anybody has a guess on what might cause the divergence and extreme values I would be happy to here them.
Thanks

Jannis

Mach_roughwalls.jpg

Mach_smoothwalls.jpg

AlexanderZ · September 20, 2022, 04:22

roughness height of 129µm is bigger than initial first cell height of 5,67e-3mm

however, lower roughness heights (20µm) is bigger either
I would check this first

also you may try to simulate in transient to get better convergence

LuckyTran · September 20, 2022, 14:26

Most likely a general convergence error not really related to roughness. The roughness height just modifies slightly the wall shear stress. Just try to find a more robust approach or figure out where it is diverging (it's almost always where the shitty cells are).

turbvisc · September 23, 2022, 09:22

Thanks to both of you for your input. Sadly the advice you gave didn't really make a different in my calculations.

On thing that did work however was to use the high roughness (icing) model for the wall roughness. I get converged and plausible values using that approach.

LuckyTran · September 23, 2022, 11:18

Icing in a rocket nozzle is a plausible result? Well okay then

turbvisc · September 23, 2022, 11:32

Well Icing isn't, but the model is, even though primarily, not exclusively for Icing simulations, but also for high roughness walls.

high_roughness.png

LuckyTran · September 23, 2022, 16:38

Well I've never seen a rocket with rough walls either so there's that.

The expression grasping as straws comes to mind. Look, you got unlucky with your original settings and kept clicking buttons until something did work. Tomorrow you'll run a flat plate and it may or may not converge. And one day you'll run a pipe and it may or may not work. You'll continue to have these experiences until you figure out what makes stuff diverge in the first place and I'll tell you, it isn't wall roughness because you can simply turn off the wall roughness altogether by making it a slip wall.

turbvisc · September 24, 2022, 12:41

It's additively manufactured, hence the roughness.

I ran it with smooth walls, it converged. I ran it with lower roughness values, it converged. I also ran it with an other turbulence model (k-epsilon) and it did also converge. Just the combination of k-w and the high roughness gave problems. So I tried to understand how the roughness is taken into account by the turbulence model and stumbled over the high roughness model for the SST model in the theory guide, which uses a different modelling for k and w near the wall, which was the region where problems occurred, and why I tried it out. Is that such bad of a practice?

September 16, 2022, 11:32	Divergence for compressible flow with rough walls	#1
turbvisc New Member Jannis Join Date: Sep 2022 Posts: 7 Rep Power: 3	Hello everyone, I'm currently doing a heat-transfer analysis an a nozzle with compressible flow with the k-w-SST turbulence model. Therefore I want to include rough walls with a roughness height of 129µm. But as soon as I include them, my solutions never converge and I get extreme values in the boundary layer. An inflation layer with an initial first cell height of 5,67e-3mm is included, but I also tried lower and larger values. I'm running the pressure-based solver since I never got the solution to converge with density-based. Other settings are coupled scheme, second order discretization, pseudo transient. But I also tried other inputs also for the relaxation factors, which are currently set to default. I included a photo for rough walls and for smooth walls, all other settings are the same for both simulations. With lower roughness heights (20µm) I reached convergence and plausible results. If anybody has a guess on what might cause the divergence and extreme values I would be happy to here them. Thanks Jannis Mach_roughwalls.jpg Mach_smoothwalls.jpg Last edited by turbvisc; September 16, 2022 at 13:08.

September 20, 2022, 04:22		#2
AlexanderZ Senior Member Alexander Join Date: Apr 2013 Posts: 2,363 Rep Power: 34	roughness height of 129µm is bigger than initial first cell height of 5,67e-3mm however, lower roughness heights (20µm) is bigger either I would check this first also you may try to simulate in transient to get better convergence __________________ best regards ****************************** press LIKE if this message was helpful

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Error SIGSEGV using VOF and UDF	JERC_UTFSM	Fluent UDF and Scheme Programming	14	November 7, 2021 23:17
Pressure boundary condition on walls in Stokes flow (groovyBC)	wildfire230	OpenFOAM Running, Solving & CFD	15	November 26, 2020 12:56
divergence in two-phase air-water flow in vertical-horizontal elbow	E.Mrz	OpenFOAM Running, Solving & CFD	0	January 10, 2019 07:27
mesh at the walls for a laminar flow	stephane	FLUENT	1	April 27, 2015 11:33
Divergence in Two phase flow in impeller for Cavitation	joshghoun	Fluent Multiphase	2	November 5, 2014 09:33

September 20, 2022, 14:26		#3
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,674 Rep Power: 66	Most likely a general convergence error not really related to roughness. The roughness height just modifies slightly the wall shear stress. Just try to find a more robust approach or figure out where it is diverging (it's almost always where the shitty cells are).

September 23, 2022, 09:22		#4
turbvisc New Member Jannis Join Date: Sep 2022 Posts: 7 Rep Power: 3	Thanks to both of you for your input. Sadly the advice you gave didn't really make a different in my calculations. On thing that did work however was to use the high roughness (icing) model for the wall roughness. I get converged and plausible values using that approach.

September 23, 2022, 11:18		#5
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,674 Rep Power: 66	Icing in a rocket nozzle is a plausible result? Well okay then

September 23, 2022, 11:32		#6
turbvisc New Member Jannis Join Date: Sep 2022 Posts: 7 Rep Power: 3	Well Icing isn't, but the model is, even though primarily, not exclusively for Icing simulations, but also for high roughness walls. high_roughness.png

September 23, 2022, 16:38		#7
LuckyTran Senior Member Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,674 Rep Power: 66	Well I've never seen a rocket with rough walls either so there's that. The expression grasping as straws comes to mind. Look, you got unlucky with your original settings and kept clicking buttons until something did work. Tomorrow you'll run a flat plate and it may or may not converge. And one day you'll run a pipe and it may or may not work. You'll continue to have these experiences until you figure out what makes stuff diverge in the first place and I'll tell you, it isn't wall roughness because you can simply turn off the wall roughness altogether by making it a slip wall.

September 24, 2022, 12:41		#8
turbvisc New Member Jannis Join Date: Sep 2022 Posts: 7 Rep Power: 3	It's additively manufactured, hence the roughness. I ran it with smooth walls, it converged. I ran it with lower roughness values, it converged. I also ran it with an other turbulence model (k-epsilon) and it did also converge. Just the combination of k-w and the high roughness gave problems. So I tried to understand how the roughness is taken into account by the turbulence model and stumbled over the high roughness model for the SST model in the theory guide, which uses a different modelling for k and w near the wall, which was the region where problems occurred, and why I tried it out. Is that such bad of a practice?