[oj.bulmer]

Hi all,

I was curious about how much I can push the simplification in CFD modelling.
Please have a look at following image:



My goal is to find the optimum angle for a casting, which obviously is 3D, so that I have least pressure drop. So I may need to create a series of 3D models with different to get a trend and then when I know the minima of a curve of pressure drop versus , I have found my optimum value.

Now, I was thinking, if i only create a 2D sketch of this model with different and obtain the optimum value, I can have some guidance about optimum angle for 3D. Agreed, the actual casting is cylingrical and I am representing it as 2D, and the physics is entirely different. But what I want only, is to get closer to optimum value through the series of 2D simulations. Suppose if I find that the optimum angle for 2D section is say 50 deg, I can then take series of angles close to 50 and run 3D simulations.

It is not about exact result of 2D, but about zeroing in on the range of of angles I need to select for 3D simulations that I can focus on. It is far better to run with range of angles at 45, 47, 50, 53 and 55 deg instead of running with range of 30, 40, 50, 60, 70 deg, to get the accurate inflection point of the curve, to find optimum angle.

Does this approach sound reasonable? Or am I missing something?

Thanks!

[oj.bulmer]

I tried to find any such crude studies done but couldn't find .Any thoughts on this?

[JuPa]

Originally Posted by oj.bulmer

I tried to find any such crude studies done but couldn't find .Any thoughts on this?

Hi Oj,

Your reasoning sounds plausible. Once you found the optimum angle in 2D (let's say 50 deg) it's worth doing this in 3D but with opposite but extreme angles (i.e. 3D at 35 deg and 3D at 65 deg) just as a sanity check.

If you are also modelling buoyancy you can forget about symmetry - from personal experience buoyancy is never symmetrical (granted you stated in your opening post 2D results may not reflect 3D behavior).

Do this in Fluent: 2D Fluent is much faster than 2.5D CFX.

From your picture the location of your BCs may be an issue. The inlets/outlets look too close to where the action is. In my opinion you should move them further upstream/downstream.

Lastly when modelling in 2D beware of the coanda effect.

Edit: I think you should some time developing a Bernoulli type resistance model to see where the maximum pressure drop is with respect to the opening in the middle (by opening I mean the size of the bump, which will change with angle).

[ghorrocks]

I don't think a 2D model is going to tell you much which is useful. This looks like a reasonably straight forward model to do in full 3D. So I would just go straight to 3D.

This looks like an ideal case for a parametric sweep. This is easy to set up in Workbench where you link a parameter to define the angle to an output pressure loss.

[oj.bulmer]

2D (let's say 50 deg) it's worth doing this in 3D but with opposite but extreme angles (i.e. 3D at 35 deg and 3D at 65 deg) just as a sanity check.

I don't think a 2D model is going to tell you much which is useful. This looks like a reasonably straight forward model to do in full 3D. So I would just go straight to 3D.

The whole point of doing a 2D model is to get just close to the optimum angle for 3D so I don't have to run for a wider range. Assuming that for 2D if it is 50 deg, for 3D it won't be 20 deg or 85 deg, but close to 50 deg. Doing the extreme cases actually defies the purpose of this hassle, I might as well just run all 3D as Glenn suggests.

Of course the actual models will be setup with inlet and outlets considering the lengths required for fully developed flow. Sorry I am so used to doing this that I didn't think of mentioning it and focused only on the casting.

If you are also modelling buoyancy you can forget about symmetry

This will be typically installed in pipelines with line pressure in excess of 10 bar and mean velocities ranging from 5-30 m/s. So buoyancy/gravity forces are ignored while modelling this.

I was just curious about this approach. I intend to optimise all aspects of the casting and just thought if I could use 2D-3D analogy to some extent to find the range for 3D. Will update if I get anything meaningful from this exercise...

Cheers