Residual hgih but stable, acceptable?

Robson_Aguiar · January 6, 2021, 09:18

Dear experts,

I am running a simple fluid flow through cocentric pipes wich are both inlets.

They have a communication in the end of the more external pipe and my pressure outlet is in the end of the internal pipe, the end of the external pipe is simply a wall.

I run my simulation using implicity unsteady and K-epsilon but my residuals seems not to decrease much and they look like the image attached

Capturar.jpg

normally I see my residuals going as low as 1e-4 but this time it does not goes low, I believe that it is because of the small communication between the cocentric pipes. Should I consider it valid because it seems "stable" despite being high?

fluid23 · January 6, 2021, 11:00

These do not appear to be normalized. Did you turn that option off for some reason?

Robson_Aguiar · January 6, 2021, 12:12

Quote:

Originally Posted by fluid23

These do not appear to be normalized. Did you turn that option off for some reason?

not that I am aware of... I will double-check, it should be disabled in the plot itself?

edit: Just checked in the monitors and it is as automatic normalization.

for all the curves the factor is 5. Maybe automatic is not giving a good one?

Maddin · January 7, 2021, 07:08

Y+ fits? Temp/Pressure fits?
Depending on your "simulation target" you may can tweak the mesh (coarser) to reduce the turbulence, you could also modify a little bit the coefficient of your model - but it depends on your target.

5-10 years ago it was easier cause you didn't had so much CPU power easy available, fine mesh and near wall stuff brings you back to the theory work

ping · January 8, 2021, 02:26

a common reason for higher residuals is the flow is not stable and so you need to investigate this using a few monitors and writing scenes to disk every few time steps to inspect for the flow changes - they might be quite normal.
also you are using the unsteady solver so you should also test using a few more inner iterations and lowering the timestep since you want to get each timestep converging too - this is not apparent in your plot above.

Robson_Aguiar · January 8, 2021, 08:47

Quote:

Originally Posted by ping

a common reason for higher residuals is the flow is not stable and so you need to investigate this using a few monitors and writing scenes to disk every few time steps to inspect for the flow changes - they might be quite normal.
also you are using the unsteady solver so you should also test using a few more inner iterations and lowering the timestep since you want to get each timestep converging too - this is not apparent in your plot above.

Hello, thanks a lot!! I have already reduced the time step and I got the same result, I will try increasing the inner iterations to check what happens.

ping · January 8, 2021, 09:31

maybe you have not gone small enough with timesteps since the residuals for each timestep should look like a mini steady state convergence ie dropping a few orders of magnitude then jumping back up at the first iteration of the next timestep ie like a saw tooth. and it is rare to need more than 10 inner iterations if your timestep is small enough, other than in force balances, mass conservation etc cases

January 6, 2021, 09:18	Residual hgih but stable, acceptable?	#1
Robson_Aguiar New Member Robson Join Date: Dec 2020 Posts: 4 Rep Power: 6	Dear experts, I am running a simple fluid flow through cocentric pipes wich are both inlets. They have a communication in the end of the more external pipe and my pressure outlet is in the end of the internal pipe, the end of the external pipe is simply a wall. I run my simulation using implicity unsteady and K-epsilon but my residuals seems not to decrease much and they look like the image attached Capturar.jpg normally I see my residuals going as low as 1e-4 but this time it does not goes low, I believe that it is because of the small communication between the cocentric pipes. Should I consider it valid because it seems "stable" despite being high?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
laplacianFoam with source term	Herwig	OpenFOAM Running, Solving & CFD	17	November 19, 2019 14:47
Segmentation fault when using reactingFOAM for Fluids	Tommy Floessner	OpenFOAM Running, Solving & CFD	4	April 22, 2018 13:30
chtMultiRegionSimpleFoam turbulent case	Aditya Patil	OpenFOAM Running, Solving & CFD	6	April 24, 2017 23:13
simpleFoam error - "Floating point exception"	mbcx4jc2	OpenFOAM Running, Solving & CFD	12	August 4, 2015 03:20
calculation stops after few time steps	sivakumar	OpenFOAM Running, Solving & CFD	7	March 17, 2013 07:37

January 6, 2021, 11:00		#2
fluid23 Senior Member Matt Join Date: Aug 2014 Posts: 947 Rep Power: 18	These do not appear to be normalized. Did you turn that option off for some reason?

January 7, 2021, 07:08		#4
Maddin Senior Member Join Date: Mar 2009 Posts: 260 Rep Power: 18	Y+ fits? Temp/Pressure fits? Depending on your "simulation target" you may can tweak the mesh (coarser) to reduce the turbulence, you could also modify a little bit the coefficient of your model - but it depends on your target. 5-10 years ago it was easier cause you didn't had so much CPU power easy available, fine mesh and near wall stuff brings you back to the theory work

January 8, 2021, 02:26		#5
ping Senior Member Ping Join Date: Mar 2009 Posts: 556 Rep Power: 20	a common reason for higher residuals is the flow is not stable and so you need to investigate this using a few monitors and writing scenes to disk every few time steps to inspect for the flow changes - they might be quite normal. also you are using the unsteady solver so you should also test using a few more inner iterations and lowering the timestep since you want to get each timestep converging too - this is not apparent in your plot above.

January 8, 2021, 09:31		#7
ping Senior Member Ping Join Date: Mar 2009 Posts: 556 Rep Power: 20	maybe you have not gone small enough with timesteps since the residuals for each timestep should look like a mini steady state convergence ie dropping a few orders of magnitude then jumping back up at the first iteration of the next timestep ie like a saw tooth. and it is rare to need more than 10 inner iterations if your timestep is small enough, other than in force balances, mass conservation etc cases