[C72]

Hi folks. I'm attempting to model a standard low-rise building in a kind of urban environment with a wall and roof thickness of 1m. I've created my first mesh and refined this appropriately to capture most of the wake phenomena etc.

Since then, to begin with, i've selected the following physics models that i've seen used before:

Implicit Unsteady
Gas
Segregated Flow
Constant Density
Turbulent - K-Omega
Turbulence Suppression with all-y+ auto-selected

My inlet velocity magnitude is a constant 40m/s (to begin with as I will later apply a power law field function with height) and i've specified turbulence intensity and length scale as 0.01 and 0.02 respectfully.

Time step is 0.01s with temporal discretisation 1st order and max. inner iterations is set to 10.

When I run my first simulation. My residuals all seem to oscillate forever, meaning that I can't get any accurate readings from pressure etc. I plan to make the building walls and roof a porous media in future so i dont know if i should have maybe done this before simulating or if that would make a difference to this.

I'm more or less a beginner with Star anyway and would just like to what is causing this and how to solve it, or if anyone could just point me in the right direction so i can learn it would be much appreciated! I've attached some images to make this easier to understand.

Thanks!

[LuckyTran]

Residuals do that in unsteady simulations. At the start of each time-step, the residuals will spike if there's any change in the solution; and there should be unless you found the steady state solution.


I'm not saying there isn't a million thing you could be doing better, but this behavior in residuals is expected.

[C72]

Hi thanks for the response!

I've since gone and changed my approach, switching to the following physics models:

• Three Dimensional;
• Steady;
• Gas;
• Segregated Flow;
• Constant Density;
• Turbulent;
– Reynolds-Averaged Navier-Stokes (RANS);
– K-Epsilon Turbulence / Standard K-Epsilon / High y
+ Wall Treatment.

This has improved my output however my residuals still don't converge to flat lines. Is this simply due to the fact that i have turbulence involved in the physics and will this always be the case?

Thanks and i appreciate the help.

[LuckyTran]

Although a turbulence model might be involved, there isn't a rule that residual plots will look like this if there is a turbulence model activated. There's tons of reasons why residuals might act that way and almost none of them can be determined simply by looking at the residuals.


You could model a simple box or pipe with the same turbulence model and get a very different looking residual plot.

[SWalter_RSI]

In addition to the residuals, I'd look at what the outflow pane looks like. Are there gradients across the outflow? If so, while modern CFD codes are pretty robust with respect to that, you may want to increase the distance from your area of interest to let the flow settle out. If you still see the behavior in your residuals, it's likely an issue with how you set up your simulation.



Also, check your outlet type. Different outlets are better suited for different problems. The user manual/help file should be able to give you some guidance on which is best suited for your simulation type.

[Nikpap]

It is likely that your simulation does not have a steady solution and that's why your residuals look like this for your steady results.

If that is not the case, other reasons include: discretisation errors, modelling errors, ill-conditioned physics etc.

Just by briefly looking at your residuals for the unsteady case, they are what I would expect to see.

Your results will not be what you want if you're looking at the instantaneous values of whatever you're looking for, so you need to average them by yourself using a monitor for each value of interest.

The results at the beginning of the unsteady simulation are considered "noise", that is why you generally want ~4 flow through times through your volume to start collecting averages. You can have a tiny bit of confidence when your average is not changing anymore.

It would be better if you had a look at the book by Ferziger, Peric and Street: Computational Methods for Fluid Dynamics to get yourself familiar with what each option does.