[fruitkiwi]

Sometimes when i cannot work, i will think back to origin of CFD.
Today i meet a question,

What is the origin of CFD?
I draw a tube, and set the material properties, then mesh, and set the inlet velocity and outlet pressure =zero, the model converge.

But What is inside the tube before i draw the fluid into the tube? is it empty or not.
if it is empty, then why i set the material properties to fluid?

i know this question make me looks stupid, but i cannot explain this one to a normal guy.so i hope seek some help here.

[fruitkiwi]

just realize it is actually the grid in, i set the material proprties into the domain, then by calculate element one by one, it seems like the fluid flow into domain..

haha, ask a innocent question, i do not how to close the thread, so just leave it la.

[julien.decharentenay]

Hi Kiwi,

Interesting question. The answer is simpler: what is inside the tube before is defined by your initial conditions...

Julien

[fruitkiwi]

Hi, Julien

Thanks for your support, haha, i get caught in a thinking box.
Correct me if i am wrong. Initially we define the material properties, on the grid point, which identify the domain as fluid domain, but inside only has static point, every point has material properties, but without velocity and pressure term.

After we define a inlet boundary condition, then the velocity calculate from the inlet element , pass from one element to another element. that's why we see the fluid flow into the domain.

so initially, there is not fluid in the domain, just a material properties for the later computation method to be applied.

am i right?

[julien.decharentenay]

Initially there is fluid in the domain - its properties are defined by your initial conditions.

If the velocity is zero and pressure is equal to atmospheric, the fluid is at rest. You set the velocity to be 1 m/s in the x direction, and then you would have a fluid moving at 1m/s in the x direction.

Hope it helps.

[arjun]

Originally Posted by julien.decharentenay

Initially there is fluid in the domain - its properties are defined by your initial conditions.

If the velocity is zero and pressure is equal to atmospheric, the fluid is at rest. You set the velocity to be 1 m/s in the x direction, and then you would have a fluid moving at 1m/s in the x direction.

Hope it helps.

Initially there is a fluid in the domain that does not necessarily follow Naiver Stokes (no convergence) but this fluid has material properties of fluid you want to simulate.

After iterations and convergence, this fluid starts to look like a fluid that follows Navier Stokes. The amount of error when compared to Navier Stokes will determine how off this fluid is from Navier Stokes.

CFD is approximation to real thing.

[fruitkiwi]

Hi, Julien,

Thank you for your proper explanation.
However, it means that is i apply inlet pressure around 1000Pa, and i close the outlet set velocity=0, i am apply pressure on the static fluid, so the internal fluid pressure increase, am i right?

Hi, arjun,
if i am not mistaken, that's why we need to do proper initialization to make sure the fluid can be solved by navier stoke equation, am i right?
so the only way i can make half of fluid in the domain, if i separate domain to two material properties, one is gas, another one is fluid, right?


i just wish to learn more so i can pass knowledge to others.

[fruitkiwi]

Hi, any one has other opinion, you are free to share the thought.

[arjun]

Originally Posted by fruitkiwi

Hi, Julien,

Thank you for your proper explanation.
However, it means that is i apply inlet pressure around 1000Pa, and i close the outlet set velocity=0, i am apply pressure on the static fluid, so the internal fluid pressure increase, am i right?

Hi, arjun,
if i am not mistaken, that's why we need to do proper initialization to make sure the fluid can be solved by navier stoke equation, am i right?
so the only way i can make half of fluid in the domain, if i separate domain to two material properties, one is gas, another one is fluid, right?


i just wish to learn more so i can pass knowledge to others.

You can solve the system even without proper initialization too. If the solver is stable it could work with utterly ridiculous initializations too.
Initialization only makes convergence to be easily achieved. Thats the only role it plays.

Originally Posted by fruitkiwi

if i am not mistaken, that's why we need to do proper initialization to make sure the fluid can be solved by navier stoke equation, am i right?

Further fluid is not solved by Navier Stokes equations. Most of the fluids we know of could be very well approximated by Navier Stokes equations.

Imagine that you were in an universe where flow of fluid is determined by some Not-Navier-Stokes equations. And if in this universe the absurd initiation that you have given to solver in our universe are the solutions to Not-Navier-Stokes equations then the solver in that universe will say that solution is already converged. But in this universe the solver would take lots of time to converge the solution.

[fruitkiwi]

Hi, arjun,

thank you for your in-depth explanation.
I will get back to you soon, it take some time to understand it.

Hi, Julien,
any idea on my Q?

[julien.decharentenay]

Hi Kiwi,

Your reasoning seems correct. But I would be cautious when doing this kind of "toy" case/scenarios as you may get counter-intuitive results that are correct within the limits of the modelling. For instance, in your example, the results would likely differ (in my opinion) between a variable density and a constant density solver. Temperature may also have an effect.

Feel free to run the test and report on the results, if you want.

Julien