# using the profile of laboratory scale TKE as the inlet of full scale ABL

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

 July 6, 2021, 20:32 using the profile of laboratory scale TKE as the inlet of full scale ABL #1 New Member   Lakshman R Join Date: Apr 2017 Posts: 16 Rep Power: 9 I want to know how we can use the turbulent profile obtained as the output of a small wind tunnel experimentation as the inlet of full-scale simulation of ABL. The laboratory model is equivalent to the scale of 1 :370.

 July 7, 2021, 01:13 #2 Senior Member   Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,683 Rep Power: 66 You can calculate the turbulence intensity since that takes care of the velocity scale. Then use the same turbulence intensity on the new simulation. If you have only the tke scalar then this is straightforward. If you have detailed u',v',w' then you can still use the same idea, but you have a turbulence intensity (variance-covariance) matrix instead of a single scalar profile. lachuktr likes this.

 July 7, 2021, 01:26 #3 New Member   Lakshman R Join Date: Apr 2017 Posts: 16 Rep Power: 9 thanks for the early response.. that means I can take the turbulence intensity of the laboratory experiment and can use it as the initial boundary condition for full scale? and what about other parameters like U epsilon etc for the full scale initial boundary condition.?

 July 7, 2021, 14:11 #4 Senior Member   Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,683 Rep Power: 66 So what exactly did you measure and what do you want to scale? Only statistics scale. Mean velocity scales like Reynolds number. Mean dissipation rate you can get from k and the turbulent length scale. k is straightforward to measure. You still need to calculate the length scale in the new problem. But no one ever remembers to measure the turbulent length scale in the experiment... lachuktr likes this.

July 7, 2021, 20:08
#5
New Member

Lakshman R
Join Date: Apr 2017
Posts: 16
Rep Power: 9
Quote:
 Originally Posted by LuckyTran So what exactly did you measure and what do you want to scale? Only statistics scale. Mean velocity scales like Reynolds number. Mean dissipation rate you can get from k and the turbulent length scale. k is straightforward to measure. You still need to calculate the length scale in the new problem. But no one ever remembers to measure the turbulent length scale in the experiment...

Let me make it more clear.... I have the outlet profiles of U, k, epsilon etc.. of a wind tunnel experiment. The domain of this experimentation is having a height of 2m. Now I want to use these output profiles as inlet profiles of full-scale doman having height around 1000m. I want to know how to use the profiles obtained in smaller domain (10m x2mx 2m) into a big size domain (10km x10kmx 1km)

 July 8, 2021, 04:10 #6 Senior Member   Lucky Join Date: Apr 2011 Location: Orlando, FL USA Posts: 5,683 Rep Power: 66 Sorry, that did not make it any more clear. No one here knows what you even plan to simulate to even guess what BC's you even need for your ABL. What is not clear is whether you are talking about instantaneous or time-averaged velocity. What models you even plan to run on your ABL to even say what BC's you need. And so on. Anyway. Mean velocity scales with Reynolds number. k scales with turbulence intensity. epsilon can be re-scaled if you know the turbulent length scale of the experiment because the ratio of the turbulent length scale to geometric length scale is also a Reynolds number. Or even if you don't know the turbulent length scale, just knowing the ratio of k/epsilon is enough. That should be enough if you plan to run RANS with a k-epsilon model. If you have questions about how to scale a time-resolved velocity BC for a DNS, then ask. Also, please don't ever say experiment alone when you mean a surrogate numerical simulation, i.e. a numerical experiment. People hate it. Or wind tunnel when you mean a numerical wind tunnel (it's just another CFD). I'm here trying to figure out how to feed raw experiment data into a CFD when you're just trying to take outlet values from one CFD to use as an inlet to another CFD.... FMDenaro and lachuktr like this.

July 8, 2021, 09:25
#7
New Member

Lakshman R
Join Date: Apr 2017
Posts: 16
Rep Power: 9
Quote:
 Originally Posted by LuckyTran Sorry, that did not make it any more clear. No one here knows what you even plan to simulate to even guess what BC's you even need for your ABL. What is not clear is whether you are talking about instantaneous or time-averaged velocity. What models you even plan to run on your ABL to even say what BC's you need. And so on. Anyway. Mean velocity scales with Reynolds number. k scales with turbulence intensity. epsilon can be re-scaled if you know the turbulent length scale of the experiment because the ratio of the turbulent length scale to geometric length scale is also a Reynolds number. Or even if you don't know the turbulent length scale, just knowing the ratio of k/epsilon is enough. That should be enough if you plan to run RANS with a k-epsilon model. If you have questions about how to scale a time-resolved velocity BC for a DNS, then ask. Also, please don't ever say experiment alone when you mean a surrogate numerical simulation, i.e. a numerical experiment. People hate it. Or wind tunnel when you mean a numerical wind tunnel (it's just another CFD). I'm here trying to figure out how to feed raw experiment data into a CFD when you're just trying to take outlet values from one CFD to use as an inlet to another CFD....

Sorry for not giving all the details Luckytran. and thanks a lot for the response in detail.

I am trying to simulate flow over an isolated hill using the k epsilon turbulence model. I have replicated a wind tunnel experiment using OpenFOAM and I am having those outlet data (including U, k, epsilon, Turbulence intensity etc.). The results of CFD simulation, i am having is obtained by using spires and roughness elements inside the test section.
Now I want to use this data for the simulation of full-scale flow over the isolated hill (by using the data as the inlet boundary condition for U,k, and epsilon). The domain height of the first simulation was 2m and the domain height for the second simulation 1000m.