Learning Forum

[Jason Sum]

Dear all,

Firstly thanks for Rob for helping me regarding the UDF issue on the last post, the floating error settled.

As my aim is to capture the negative pressure induced near roof region due to wake effect, I set up a scenario where 5m/s wind velocity flowing through a building in steady case. However I encountered following questions:

1. I would like to seek some advice - should my case require atmospheric wind profile / simply uniform wind velocity inlet for such building wake analysis?
2. In steady simulation (using uniform velocity inlet profile), the result struggled to converge esp for continuity even with mesh refinement / a high order upwind scheme as shown as follows, could you please deduce how can I improve this situation:
   
   Turbulence model: k-epsilon RNG Enhanced Wall Treatment
   Solver: SIMPLEC with 2nd order and QUICK schemes

turbulent viscosity limited to viscosity ratio of 1.000000e+05 in 21463 cells at the last iteration
Mass imbalance recorded

pressure monitoring at a point near building's roof

BOI refinement to capture wake and leeward side
face refinement for building and ground surface
Maximum Aspect Ratio = 3.00354e+01 (high aspect ratio mostly happen in inflation layers of building edges)
Minimum Orthogonal Quality of: 0.30

    3. I have tried using transient case, it surprisingly converged soon after few timesteps but the contour plots along time are not showing sth reasonable or similar in steady case (no obvious drag seen in transient)

Thank you for addressing my concerns!

Best regards,
Jason

[Rob]

You're welcome, and thanks for starting a new topic for a new question. 

1) Up to you - how high is the building relative to the ABL? What is upstream to decide if the ABL has formed or been chewed up by the scenery? 

2) Your monitor shows a roughly repeated variation every 50-ish iterations. Plot velocity contrours around the building every 5 iterations. Also pressure on the building. What do they tell you?

3) How long a transient did you run? Time step was probably well under a second, but what frequency do you expect the flow to change at?

The extra data from 2 & 3 will help you decide if the flow changes enough to need to worry about it. A transient flow may be a very small fluctuation that can be ignored, or a very big change that can't. 

 

[Jason Sum]

Grateful if you could address my problems down below ????

[Jason Sum]

Hi Rob, based on your guidance, I have included the ABL (by UDF) for my building aerodynamics case to account for the roughness of urban terrain at the upstream.

Through a steady-state simulation with ABL inlet, the continuity residual still fluctuated between 1e0 ~ 1e-1 without converging.

 
As mentioned, BOI refinement made to capture wake and leeward side, Face refinement for building and ground surface
Maximum Aspect Ratio = ~39 (high aspect ratio mostly happen in inflation layers of building edges)
Minimum Orthogonal Quality of: 0.30 

Mass imbalance and pressure monitored at a point near rooftop (per 5 iter) also did not obtain consistent values as simulation going on. And turbulent viscosity limited to viscosity ratio of 1.000000e+05 without decreasing.

I am confused, in this case should I keep working on improving mesh quality for steady state continuity convergence, OR look for a transient process as the flow phenomenon in this case may be dynamically changing and hardly reaching a steady outcome? Is it possible that no steady state convergence because the building flow and wake oscillate (e.g. shedding)? 

Kindly hope for your experienced and comprehensive guide on this, such issue struggled me a lot!! Mush appreciate!

[Rob]

OK, first off the turbulence limiter is there to stop the solver running away, but with urban models whilst the ratio is high the turbulence values aren't, so increasing the limit by 4-5 0s is fine, and something I tend to do before running the model. You'll not find that in the text books! 

Now, the flow result. Your domain is big, if you look at the mass imbalance and/or fluctuations and work out how big a velocity variance is required to see that on the outlet how big a change in velocity is there? For a local monitor, wait for a windy day and go and stand outside. How stable is the wind? Ie are you seeing a (non scientific) steady value or are you being buffeted by gusts? Now compare that to what you're seeing in the Fluent model, and think about shedding. But... can you afford to run a full transient, or is the steady solution good enough for what you need? When thinking about that, how confident are you that the building shape is exactly correct (do the windows have frames etc or balconies), did you include trees & cars in the model?