I am trying to solve the flow over a circulation controlled aerofoil using Ansys FLUENT, (using SA turbulence model) and I have noticed that if I want to reduce y+ to a lower value, the continuity residual will decrease up to an order of 10^-4 after which it will slightly oscilate forever. After a few hundred iterations the CL,CD and CM monitors converge and don't change at all anymore. I was wondering if this is a false solution or if it is acceptable.
Hi, Checking the convergence of only residuals that is provided in Fluent is not good ideas, especially in complex problems. In most cases, it is impossible to reach 10^-4 or less, especially in the continuity. The best way to check the convergence is monitoring the convergence of some variable (such as velocity, temperature, pressure, force, and so on ) in a desired point, surface, ….. By monitoring this value and the residuals as well, you can make sure that the problem is converged! Siamak
Most standard practice is to accept it up to 10^-6. But it is seen that in some complex studies some people publishing journals with the 10^-4 error. Also, in FLUENT it is scaled residual, means an error is lower than what u think actual is(http://aerojet.engr.ucdavis.edu/fluenthelp/html/ug/node1047.htm). Your case can be seen with convergence 10^-4.
You can also check your mesh orthogonal quality in FLUENT, which is influencing parameter for convergence. goto-mesh quality check
Thank you for your answer Prafulla. I was trying to maintain a good orthogonality of my mesh but I find it difficult with a C grid as my geometry consists of an aerofoil with a round trailing edge which has a Jet inlet as well. any ideas how I could solve this issue?
Hi, Checking the convergence of only residuals that is provided in Fluent is not good ideas, especially in complex problems. In most cases, it is impossible to reach 10^-4 or less, especially in the continuity. The best way to check the convergence is monitoring the convergence of some variable (such as velocity, temperature, pressure, force, and so on ) in a desired point, surface, ….. By monitoring this value and the residuals as well, you can make sure that the problem is converged! Siamak
Thank you Siamak! I was suspecting the same thing as my CL, CD and CM monitors do actually converge eventually even though the continuity keeps oscillating around 10^4. I was actually considering adding a surface monitor as well but I need to get familiar to how to use them properly.
It's very importat to view the mesh and the model. Geometries with high changes of flow direction (cyclons) works fine with k-epsilon Realizable, for example. For others, k-epsilon Standard is enough as long as the mesh is very fine.
Please note that 10^-4 is a dimensional value. Its value is not important before the time and length scales are defined. For example, if you increase your length scale by a factor of 1000 (by scale your mesh and velocity by 1000), you can get identical results even if you choose the residual to be 10^-1.
I suggest you to test different residual like 10^-3, 10^-4 and 10^-5. Indications of well-converged results are: 1) lowering the residual does not change the results 2) residual cannot get smaller than a certain value and oscillates around a constant.
hello sir,im working in reacting flow and i had the same problem of you..the
Continuity convergence it not stable,if you slove your problem please share with us the solution...im so tired...regards
- Badges
- Science topic