Residual Convergence on Meshes is poor

SleeperService · July 23, 2011, 20:54

Hello all, I'm trying to carry out a study involving a 2D axisymmetric simulation of a conical model in Mach 5 flow with a small spherical plasma spot some distance ahead of it. I've currently meshed the problem using a mix of quad pave and map meshing depending on the zones, but am seeing extremely poor convergence for my residuals. They tend to oscillate. I am lost as to why this is happening, and am a little bit out of my depth here - would someone be willing to look at my case files and help me out? I'll gladly PM them to any who offer their assistance. Thanks.

vxv · July 23, 2011, 21:58

".....but am seeing extremely poor convergence for my residuals. They tend to oscillate......"

As it is a Mach 5 flow, I assume you are using a limiter. Which limiter are you using?

Some limiters never allow the residual to go to machine zero, even though the solution looks ok. It may happen because of one bad cell, where the residual keeps oscillating. Try Venkatkrishnan limiter!

SleeperService · July 24, 2011, 03:30

I'm not familiar yet with the concept of limiters, but the residual oscillations are very high in magnitude - between 10^2 and 10^-1. I think if they were lower in magnitude, I could ignore them safely and assume the simulation has converged, but such a degree of uncertainty seems to be indicative of another problem - would you agree?

vxv · July 24, 2011, 10:52

I have never run a Mach 5 case but for supesonic cases around Mach 2, my residuals used to stall around 1e-2. But as you say, your residuals are order of 1e+2. In that case there certainly is something wrong elsewhere.

Which code are you using? What boundary conditions have you specified?

SleeperService · July 25, 2011, 06:17

Er...fairly new to this, so apologise in advance if this isn't the answers to the questions you're asking!

I know the Solver is density based, implicit, second order upwind. Simulations are being checked for grid dependency, laminar vs. turbulent cases (using K-omega SST), and finally steady vs. unsteady solutions.

Boundary Conditions at the inlet are:
P_inf = 1220 Pa
T_inf = 62 K
M_inf = 5.0

The flow is in the axial direction only, and enters from a pressure_far_field condition on the left (also present at the top of the grid, but of course, with zero flow contribution), and exits via a pressure_outlet condition with zero backflow.

The walls of the simulation are also set to the default condition. The small plasma spot, like the rest of the flow is simply modelled as air behaving as an ideal gas (viscous modelling is sutherland where applicable) but of course with a considerable energy source term totalling around 7W of power across it's 0.5 mm diameter.

I apologise if I've missed anything out - if you prefer, I'd be happy to send you my case or data files, or upload them to a filesharing account. Otherwise, please let me know what other information you need and I'll do my level best to find it - I just wasn't sure what you meant by Limiter or code.

vxv · July 25, 2011, 11:33

As the inlflow is supersonic, the flow at the outlet will also be supersonic after passing through the oblique shock wave created by the cone. So at the outlet use "extrapolation" boundary condition.

Let me know if this works!

SleeperService · July 25, 2011, 12:52

Hmm, I tried to do what you said, but it appears Fluent is already extrapolating that information. The outlet BC is of type pressure_outlet, and while this requires the specification of a outlet gauge pressure, this is ignored automatically when the local flow is supersonic as calculated from upstream values. It is also impossible to set the condition to force extrapolation in any case, since this setting is incompatible with compressible flow problems and fluent immediately reports an error.

vxv · July 25, 2011, 13:35

Can you send me your case and data files. Although I have not touched fluent in more than 3 yrs, I will give it a try. My id (varunvikas@gmail.com)

SleeperService · July 25, 2011, 14:32

Sure, sending you them now.

July 23, 2011, 20:54	Residual Convergence on Meshes is poor	#1
SleeperService New Member Join Date: Feb 2011 Posts: 10 Rep Power: 15	Hello all, I'm trying to carry out a study involving a 2D axisymmetric simulation of a conical model in Mach 5 flow with a small spherical plasma spot some distance ahead of it. I've currently meshed the problem using a mix of quad pave and map meshing depending on the zones, but am seeing extremely poor convergence for my residuals. They tend to oscillate. I am lost as to why this is happening, and am a little bit out of my depth here - would someone be willing to look at my case files and help me out? I'll gladly PM them to any who offer their assistance. Thanks.

July 23, 2011, 21:58		#2
vxv New Member vxv Join Date: Jul 2011 Posts: 6 Rep Power: 14	".....but am seeing extremely poor convergence for my residuals. They tend to oscillate......" As it is a Mach 5 flow, I assume you are using a limiter. Which limiter are you using? Some limiters never allow the residual to go to machine zero, even though the solution looks ok. It may happen because of one bad cell, where the residual keeps oscillating. Try Venkatkrishnan limiter! __________________ vxv

July 24, 2011, 10:52		#4
vxv New Member vxv Join Date: Jul 2011 Posts: 6 Rep Power: 14	I have never run a Mach 5 case but for supesonic cases around Mach 2, my residuals used to stall around 1e-2. But as you say, your residuals are order of 1e+2. In that case there certainly is something wrong elsewhere. Which code are you using? What boundary conditions have you specified? __________________ vxv

July 25, 2011, 06:17		#5
SleeperService New Member Join Date: Feb 2011 Posts: 10 Rep Power: 15	Er...fairly new to this, so apologise in advance if this isn't the answers to the questions you're asking! I know the Solver is density based, implicit, second order upwind. Simulations are being checked for grid dependency, laminar vs. turbulent cases (using K-omega SST), and finally steady vs. unsteady solutions. Boundary Conditions at the inlet are: P_inf = 1220 Pa T_inf = 62 K M_inf = 5.0 The flow is in the axial direction only, and enters from a pressure_far_field condition on the left (also present at the top of the grid, but of course, with zero flow contribution), and exits via a pressure_outlet condition with zero backflow. The walls of the simulation are also set to the default condition. The small plasma spot, like the rest of the flow is simply modelled as air behaving as an ideal gas (viscous modelling is sutherland where applicable) but of course with a considerable energy source term totalling around 7W of power across it's 0.5 mm diameter. I apologise if I've missed anything out - if you prefer, I'd be happy to send you my case or data files, or upload them to a filesharing account. Otherwise, please let me know what other information you need and I'll do my level best to find it - I just wasn't sure what you meant by Limiter or code. Last edited by SleeperService; July 26, 2011 at 17:29.

July 25, 2011, 11:33		#6
vxv New Member vxv Join Date: Jul 2011 Posts: 6 Rep Power: 14	As the inlflow is supersonic, the flow at the outlet will also be supersonic after passing through the oblique shock wave created by the cone. So at the outlet use "extrapolation" boundary condition. Let me know if this works! __________________ vxv

July 24, 2011, 03:30		#3
SleeperService New Member Join Date: Feb 2011 Posts: 10 Rep Power: 15	I'm not familiar yet with the concept of limiters, but the residual oscillations are very high in magnitude - between 10^2 and 10^-1. I think if they were lower in magnitude, I could ignore them safely and assume the simulation has converged, but such a degree of uncertainty seems to be indicative of another problem - would you agree?

July 25, 2011, 12:52		#7
SleeperService New Member Join Date: Feb 2011 Posts: 10 Rep Power: 15	Hmm, I tried to do what you said, but it appears Fluent is already extrapolating that information. The outlet BC is of type pressure_outlet, and while this requires the specification of a outlet gauge pressure, this is ignored automatically when the local flow is supersonic as calculated from upstream values. It is also impossible to set the condition to force extrapolation in any case, since this setting is incompatible with compressible flow problems and fluent immediately reports an error.

July 25, 2011, 13:35		#8
vxv New Member vxv Join Date: Jul 2011 Posts: 6 Rep Power: 14	Can you send me your case and data files. Although I have not touched fluent in more than 3 yrs, I will give it a try. My id (varunvikas@gmail.com) __________________ vxv

July 25, 2011, 14:32		#9
SleeperService New Member Join Date: Feb 2011 Posts: 10 Rep Power: 15	Sure, sending you them now.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Extrusion with OpenFoam problem No. Iterations 0	Lord Kelvin	OpenFOAM Running, Solving & CFD	8	March 28, 2016 11:08
How to write k and epsilon before the abnormal end	xiuying	OpenFOAM Running, Solving & CFD	8	August 27, 2013 15:33
Orifice Plate with a fully developed flow - Problems with convergence	jonmec	OpenFOAM Running, Solving & CFD	3	July 28, 2011 05:24
Differences between serial and parallel runs	carsten	OpenFOAM Bugs	11	September 12, 2008 11:16
Unknown error	sivakumar	OpenFOAM Pre-Processing	9	September 9, 2008 12:53