# Residuals not converging

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

 August 28, 2020, 07:00 Residuals not converging #1 New Member   Aniket Dalvi Join Date: Aug 2020 Posts: 4 Rep Power: 4 I am simulating Rayleigh Benard Convection between two horizontal plates for steady and laminar case. I am attaching a screenshot of my residuals along with convergence criteria. Any suggestions for reaching convergence will be highly appreciated. Screenshot (1329).jpg

 September 2, 2020, 01:01 #2 Senior Member   JSM Join Date: Mar 2009 Location: India Posts: 192 Rep Power: 19 Hello, There are two possible reasons. 1. If your initial solution is very close to final converged solution, then residual monitors will not go down. 2. If flow is in transient nature, then it may not converge much. In both scenario, it is better to put some monitor points at critical regions and monitor pressure or velocity. If it is flat, then solution can be considered as converged. If oscillation is too high at monitor points, then change to transient solver. Sai Krishna and Aniket Dalvi like this. __________________ With regards, JSM

September 4, 2020, 05:03
#3
New Member

Aniket Dalvi
Join Date: Aug 2020
Posts: 4
Rep Power: 4
Quote:
 Originally Posted by jsm Hello, There are two possible reasons. 1. If your initial solution is very close to final converged solution, then residual monitors will not go down. 2. If flow is in transient nature, then it may not converge much. In both scenario, it is better to put some monitor points at critical regions and monitor pressure or velocity. If it is flat, then solution can be considered as converged. If oscillation is too high at monitor points, then change to transient solver.

I tried using unsteady and turbulent k-epsilon model. In this case, I ran the simulation for a flow time of 20 mins but failed to see any flow occur. I am attaching the temp and velocity contours.

velocity.png

contour-1.png

 September 6, 2020, 19:51 #4 Senior Member   JSM Join Date: Mar 2009 Location: India Posts: 192 Rep Power: 19 Hello, For simulations like Rayleigh–Benard convection, at least transient simulation should converge. Then it looks, meshing & solver set up might be the reason for this convergence issue. There are many questions you have to answer and think...Is mesh fine enough to capture the flow accurately? Hex or Tetramesh? What is the Y+ Value and mesh growth rate? Have you added boundary layers in the mesh? How you decided the flow is turbulent flow? What is the time step size you have chosen? Have you monitored the flow in any particular points? I recommend to read some journal papers about Rayleigh-Benard convection to understand the phenomenon and underlying physics of this simulation. Also read Ansys documentation about natural convection simulations. This might help you to get the converged solution quickly. ns.cskn likes this. __________________ With regards, JSM