# Divergence detected in AMG solver: temperature

 Register Blogs Members List Search Today's Posts Mark Forums Read

 July 29, 2013, 20:16 Divergence detected in AMG solver: temperature #1 New Member   Join Date: Oct 2012 Posts: 9 Rep Power: 13 Hi Guys, I'm working on a subsonic symmetric airfoil and I used the pressure far-field boundary condition. The speed of the airfoil is suppose to be between 6m/s to 10m/s and this correspond to a Mach number between 0.017 to 0.03. When I run the calculation, the following error messages "divergence detected in AMG solver: temperature" or "divergence detected in AMG solver: nut" appears and the iteration stops. But i noticed that there was no problem running the same calculation for a mach number of 0.08 and above. Please what is the possible explanation to this and how can I fix it.

 July 30, 2013, 01:31 #2 Senior Member   Paritosh Vasava Join Date: Oct 2012 Location: Lappeenranta, Finland Posts: 732 Rep Power: 22 The "divergence detected..." error could be due to coarse mesh or large time step or both. Also your mesh that works well for one case may not work for other case. You could check the temperature distribution to find the area where it goes wrong and then refine mesh in that region. You could use laminar model for steady conditions to obtain the initial condition. Then switch to turbulence model. chaitanyaarige, pdgoyani and Audrius like this.

 July 30, 2013, 02:01 #3 Senior Member   Paritosh Vasava Join Date: Oct 2012 Location: Lappeenranta, Finland Posts: 732 Rep Power: 22 At what stage of your simulation the error occurs?

 July 30, 2013, 06:31 #4 New Member   Join Date: Oct 2012 Posts: 9 Rep Power: 13 Thanks Vasava for your feedback. Actually the error occurs after some few iterations. So I cant check for the temperature distribution. My mesh seem ok because I checked for skewness and quality. Also I'm modelling compressible flow that's why am stuck with the pressure far-field boundary condition because am trying to avoid the velocity inlet boundary condition.

 July 30, 2013, 07:53 #5 Senior Member   Paritosh Vasava Join Date: Oct 2012 Location: Lappeenranta, Finland Posts: 732 Rep Power: 22 You could use 'hybrid initialization' instead of usual one. Also in the settings for 'hybrid initialization' increase the iteration to 50 instead of default 20. Lets see if this works. If you are very sure that your mesh is good then there must be something wrong with initial solution. How about the Courant number? Is it low enough?

 July 30, 2013, 08:14 #6 Senior Member   Paritosh Vasava Join Date: Oct 2012 Location: Lappeenranta, Finland Posts: 732 Rep Power: 22 One more question. During the iterations do you receive warning about 'turbulent viscosity' in the command window? If so then you still need to improve mesh. Your mesh might have fine skewness and quality but that does not mean that it is suitable for your experiment.

 October 25, 2013, 06:31 #7 Senior Member   Ashwani Join Date: Sep 2013 Location: Hyderabad Posts: 154 Rep Power: 12 Hybrid initialization worked for me. But is my mesh or any other thing wrong related to my mesh or BC's? What does hybrid initialization does?

 October 29, 2013, 08:39 #8 Senior Member   Paritosh Vasava Join Date: Oct 2012 Location: Lappeenranta, Finland Posts: 732 Rep Power: 22 Hybrid initialization solves equation for pressure-velocity for whatever number of iterations you give. Sometimes its difficult for fluent to get started from default initial conditions say when everything is zero. Although hybrid approach works most of the time it is always advisable to keep an eye on the mesh quality.