CFD model validation with experimental data
 

Hello

I am trying to validate my CFD model using the experimental data published in the literature. However, I am not very sure on the CFD model validation procedure since the geometry of my model is different from the geometry used for the experimental setup... does this mean that I should simulate the geometry of the experimental and compare my CFD results with the results obtained in literature??? Then if the two match, I can change the geometry to my geometry??? Will that make my model valid ?
Or should I simulate my model using their fluid and then use dimensionless variables to compare between their experimental results of their geometry and my CFD results of my geometry ??

I thought that the results produced by CFD also depend on the geometry on the model. I am sorry if my question isn’t very clear, I feel a bit confused on the validation process. Thank you !!
Any help will be greatly appreciated!

I would make a CFD calculation on the geometry in literature. If that is ok, then my cfd setup with the models used is validated. Secondly I would use my own geometry, provided the models in the CFD calculation are still valid in my own geometry. Meaning, the scale and physics should not differ too much.

Hello Gert-Jan,

Thank you so much for your reply.
The published experimental results are in a graphical form so I used Digitizer to extract the data (Pressure drop vs. Flow rate) from the graph.
I then simulated the geometry corresponding to the experimental data in CFX at each velocity. However, when I compare the pressure drop produced by cfx and that of the experimental data, I get a huge percentage error (I used single precision)... Then I run two simulations each with a different velocity (using double precision) and I got a percentage error of 17% and 38% (I am not sure why I am getting different percentage errors at each velocity if the model is exactly the same except the velocity value at the inlet boundary condition is different) ... Also, is there a way that I can reduce the percentage error ?

My model is of the flow through an eccentric annulus .. The number of nodes are 495990 for the mesh.
The properties of the fluid are shown in the picture below.

https://ibb.co/1mTSdB6

On CFX-Pre, I chose "Pure substance", "Constant Property liquids" for the material ... and a residual target of 10^-6 and a time scale factor of 1s. My advection scheme is High Resolution.

What could be the problem ? My fluid?

Thank you!!
Any help would be greatly appreciated.

It is very difficult to comment on this. I don't know the geometry, fluids, convergence, imbalance, etc. I would suggest to take a look here: https://www.cfd-online.com/Wiki/Ansys_FAQ

One thing: pressure drop, is that static or total pressure? In principle it should be total........

The geometry is:
OD: 0.726", ID: 0.46" , Eccentricity: 0.43


Its a high polymer aqueous solution of HEC. The fluid properties are:
https://ibb.co/1mTSdB6

The RSM is 10^-6. Convergence graph is:
https://ibb.co/GJVjfzP

Imbalance:
https://ibb.co/qNf4jWg

Thank you for the link, will have a look at it now.

It was static pressure, I changed it to Total pressure but it didnt make any difference (I got the same values).

Thanks again for your reply and for the link!

See, it is very difficult to comment without knowing information beforehand.
Looks like you have a complicated system. Is it laminar?
Also, number of elements doesn't say anything. It all depends on the number of elements close to the wall, since you have a shear thining liquid. What is your viscosity at the outlet? Did you include temperature?

Two models are similar if they are dynamically(forces), kinematically(velocities) and geometrically (dimensions) similar, u can check if they are similar using dimensionaless variables or what we call similarity analysis .... i will advice u to watch these videos to understand better
https://youtu.be/i2a6CDd-Yko