[eSKa]

Hello guys,

I´m simulating a tunnel with 2 different jet fan types.
I´m looking for the average velocity at the end of the tunnel to compare the results to the experimental values (meassured in a tunnel in england).

1. fan simulation: 5.1 m/s (experimental: 3.3)
2. fan simulation: 5.6 m/s (experimental: 3.8)

I don´t know if the diffrences between this values are meaningful, but it looks nice, for the first.

The tunnel is over 5 km long, and in the simulation I´m looking only at 200m. So I expected that the tunnel velocity will be higher than in the reality. Now I searched for some boundary options on the opening tunnel inlet and outlet:

loss coefficient.

I tried a few values on the opening boundaries, I hit with the value 1.5 my desirable velocity on the first fan. I took this value on both boundaries. Now I took this loss coefficient to the second fan:

1. fan simulation: 3.3 m/s
2. fan simulation: 3.4 m/s

At least I expected to get the same difference of ~ 0.5 m/s between these simulations. But my second fan is now a little bit too weak.


Now to my question:
Am I wrong to use the loss coefficient to costumize the tunnel before and after my tunnel-part, or is it just the physics, and my expectations are wrong. Or is there another option to improve my calculations.



I´m very thankful for every answer or idea thanks


eska

[ghorrocks]

"I hit the value 1.5..." - what do you mean? 1.5 what? How did you apply it?

It appears you are trying to model 0.2km of tunnel to represent 5km of tunnel, which means you need boundary conditions which represent the effect of 4.8km of tunnel. You will want to do this by defining something with the same system characteristic as the actual 4.8km of tunnel - which means something where the pressure loss is the correct function of flow rate. You might be able to estimate this from empirical 1D flow stuff, pressure loss versus flow in a long duct.

[eSKa]

Hi. Thanks for the reply.

In the first fan simulation I need the speed of 3.3 m/s. So I tried a few loss coefficient values…1…..1.5….2….2.5…..3. and With the value of 1.5 I got my desirable speed.

Yeah thats right. But I dont know what in the tunnel behind and after is working. There are driving cars and maybe another fan systems. The Wind?
So I have a little bit to guess. I have only this average velocities 140m after the fan and the tunnel/ fan- geometries and the fan setup. I m working with rough walls in the fan (silencer) and in the tunnel. It helps not enough.

[ghorrocks]

You are trying model the system characteristic of the unmodelled tunnel. So if your loss coefficient approach is the wrong system characteristic then it is never going to work - you can tune it to work at one condition, but as soon as you change the condition it it wrong. This appears to be what you are seeing with your loss coefficient=1.5.

So you have to model the unmodelled tunnel system characteristic. So you definitely need to evaluate what loss characteristic you would expect if it was just a long duct and some loss points (entry, exit, major cross section changes etc). You should be able to guess a reasonably good system characteristic from this.

But if the unmodelled tunnel contains lots of active elements (fans or dampers which respond to the conditions with their own system characteristic) then the only way you are going to model this accurately is to know how they work and get an accurate system characteristic from them. If your model is significantly affected by these things then this is the ONLY way you can model this accurately. You cannot guess - unless you know your task is impossible.

In short: you need enough detail of the unmodelled section to be able to specify its effects in a boundary condition. If you cannot do this then your task is impossible.

[eSKa]

Hi.

Thank you very for this long answer!

If I understood it right: the loss coefficient is too main for my case ….it will be wrong? Or maybe wrong? I think there is definitly more then the pressure loss of a duct (=simple pressure losses). Without the loss coefficient my calculations are looking great …in relation to the real values. So the loss coefficient must be wrong…

What other options do I have in der CFX Pre Setup? —> I think I have to work with a roughly calculated pressure loss on the openings. But this won t be possible. Because of all the unknown tunnel details.
So I will work with this higher velocities. At least I must only change the height of the fan and compare it to the thrust. To find the best fan height.

[ghorrocks]

To explain a bit more: The loss coefficient for an opening is dp = 0.5*f*rho*U^2. So for an incompressible flow the pressure loss is proportional to velocity squared. So that is your boundary condition's modelled system characteristic: proportional to velocity squared.

So if the actual system has the pressure drop proportional to velocity^2 then your model matches reality and will be accurate. But if the actual system has a system characteristic different to velocity^2 then your model will not accurately model the system. But note that you will always be able to adjust the constant term so the wrong model fits the actual results at ONE point. But it will never match the actual results at more than ONE point.

This is why you can get it to work at a single operating point, but when you try other operating points the are increasing error.

This is why if you do not know the system characteristic of the actual system then the system is not specified enough to model it accurately.

[eSKa]

very helpfull this explanation!

all my questions are answered. thanks!