# High residuals

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

 November 25, 2008, 14:26 High residuals #1 ecram Guest   Posts: n/a Hello, I have a new question for You, in my simulation of wing in free flux, the residual are high but the result in terms of Cl and Cp is good which it is the reason? TankU

 November 25, 2008, 23:17 Re: High residuals #2 no one Guest   Posts: n/a The magnitude of the residuals is not important, they are pretty arbitrary. I think Star-CD and Star-CCM+ normalize your residuals by the largest change in any cell from iteration 0 to iteration 1. If the initial conditions were very close to the solution, then your residuals will be normalized by a very small number, making the residuals high. The thing to look out for is increasing or oscillating residuals that could indicated mesh or setup problems. Monitoring values like Cp and Cl as you have done is a much better gauge of convergence. Residuals are a better indicator of mesh quality than convergence.

 November 26, 2008, 04:43 Re: High residuals #3 ecram Guest   Posts: n/a Thank you very much hour I am most calmer

 November 26, 2008, 09:19 Re: High residuals #4 James Guest   Posts: n/a At least in 3.26 you can turn ANORM on. This turns the residual methodology back to the former physics based system, used before they decided that they needed to be like Fluent. For most of the variables this ends up being the sum of all fluxes through the cell walls normalized by the inlet flux. The absolute value is related to the inlet flux to volume of the model, but at least it means something physical. I believe they changed because Fluent was 'easier' and 'faster' to converge. Since you have no real idea what these numbers mean if the number was low then you can pretend that it is converged if it asymptotes. It is somewhat strange that if you have a bad initial guess you can converge faster (fewer iterations) than if you mapped a full solution. In the first case the initial residuals are very poor and can be reduced fast, in the second case you may be very near convergence and thus they should not change much.

 April 20, 2020, 12:14 #5 New Member   Roger Fernandes Join Date: Jun 2019 Posts: 5 Rep Power: 6 Besides simulating a lot and learning Is there any literature that can be recommended on this?