[RobR_CFD]

Most Learned Colleagues,

I wish to examine drag coefficients for structural shapes and will be validating via wind tunnel tests once COVID restrictions allow.

In the meantime, I am setting up some benchmarks via some CFD simulations.

Firstly I have setup a domain which represents a bare wind tunnel analogue - equivalent to the real version I will use. This will be my boundary condition calibration case.

I have a reasonably fine mesh, have looked at the various guideline parameters for the runs and these are proving within normal bounds.

Using Fluent, I am starting out with a KE RNG model using steady conditions which I will then use as a primer for a SAS transient run.

My question is - in principle - what is a reasonable level of convergence?

The reason for asking is that:
1. my velocity monitors setup through the domain are reaching steady values relatively quickly in the KE steady run at low iterations.
2. All of the convergence residuals are reducing down to between 1e-6 to 1e-8 very quickly.
3. My y+ values at the walls appear reasonable, but I do have fairly large aspect ratios, albeit othogonality values are good.

But, the continuity does do improve much beyond 1e-2. It does flat-line and have small oscillations, it just hits a limit and stall regardless of iterations.

Given all other parameters appear fine, is this acceptable, or should I be refining further?

Thanks

[mazhar16823]

Originally Posted by RobR_CFD

Most Learned Colleagues,

I wish to examine drag coefficients for structural shapes and will be validating via wind tunnel tests once COVID restrictions allow.

In the meantime, I am setting up some benchmarks via some CFD simulations.

Firstly I have setup a domain which represents a bare wind tunnel analogue - equivalent to the real version I will use. This will be my boundary condition calibration case.

I have a reasonably fine mesh, have looked at the various guideline parameters for the runs and these are proving within normal bounds.

Using Fluent, I am starting out with a KE RNG model using steady conditions which I will then use as a primer for a SAS transient run.

My question is - in principle - what is a reasonable level of convergence?

The reason for asking is that:
1. my velocity monitors setup through the domain are reaching steady values relatively quickly in the KE steady run at low iterations.
2. All of the convergence residuals are reducing down to between 1e-6 to 1e-8 very quickly.
3. My y+ values at the walls appear reasonable, but I do have fairly large aspect ratios, albeit othogonality values are good.

But, the continuity does do improve much beyond 1e-2. It does flat-line and have small oscillations, it just hits a limit and stall regardless of iterations.

Given all other parameters appear fine, is this acceptable, or should I be refining further?

Thanks

As from what you told in 1,2,3 leads to deduce that the solution has converged but if I understood correctly you said continuity residuals don't drop below 1E-2 so you should not trust this solution - until "continuity" reaches the same order of convergence i.e. up to 1E-3 or below this threshold. It happened in my case as well, so I reduced the cell size on the boundary surfaces except the main flow acting part, and the "continuity" dropped down to 1E-3.


P.S: Usually, for the convergence residuals have to drop down to the value 1E-4 and the quantity under consideration must become steady.