CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   System Identification and cfx (

Al Mazdeh May 19, 2008 12:41

System Identification and cfx
I like to get your opinion on something that has been a challenge for me in a past couple of months-

The question is somehow related to cfd analysis-

I have a simple cylindrical tube-

I want to study the acoustic behavior of this tube containing fluid such as air-

Let's say the tube is open-open on both end

ANSYS classic have an acoustic tool that lets u calculate all the acoustical natural frequencies and its related modes shapes-

But the problem with ANSYS classic is that it only simulates a non-moving media with its acoustic tool-

But what I am really interested to know is that to compare the natural acoustic frequency of a moving flow versus a nonmoving flow in the tube-

Or in other word I want to know how much the natural frequency will change when there is a flow in the tube-

The conventional way to do this kind of analysis is to perform a frequency response analysis on the model-I have researched some resources and I am convinced that CFX can only perform in time domain-

Then I started to think may be it is better to perform a time history analysis and then convert the data into a frequency domain-But I am not sure if this is a right way or if there is a more established method for these kind of work that I am not aware of

Do u have any info on the subject that u can direct me? I truly appreciate u.

Dr. toner May 20, 2008 05:59

Re: System Identification and cfx

not sure how cfx can help you with your problem. cfx does not calculate any frequencies as far as i know. the only way how it might work, is to calculate your desired frequencies out of the velocities you get from the cfd calculation. (not sure if this is easy to do, but most likely not) ;-)

Al Mazdeh May 20, 2008 08:31

Re: System Identification and cfx
can u elaborate more on this

Glenn Horrocks May 20, 2008 18:54

Re: System Identification and cfx

CFX does not calculate in the frequency domain so the only way of getting frequency responses out of it is to run a transient model (possibly multiple models each with different excitation frequencies) and extract the frequency information from that. Extracting the frequency information is easy, it is just FFT. Deciding exactly what is an input and what is an output and recasting it in a form suitable for FFT such that the results are meaningful is the challenge.

Glenn Horrocks

al mazdeh May 20, 2008 19:26

Re: System Identification and cfx
by multiple models u mean

if we say q is the input quqntity-

I apply q*sinw1t then I measure the output let'scall it y1

then we run this model with different w such as w1,w2,..wn then we measure the associate output quantities such as y1,y2,,yn

then finally we plot y versus w

this is my understanding- why do we need FFT TRANSFORMATION ANY MORE

PLEASE get back with me since this topic is becoming a pain for me- my dissertation is about due


Glenn Horrocks May 21, 2008 01:10

Re: System Identification and cfx

That's right. Try different input parameters and see what output they give. Graph one against the other and you can find the resonance point. Slow and tedious but it should work. You might be able to speed it up by ramping w slowly and then you look at the response sweep.

If you use this method you are right, no need for FFT.


Rogerio Fernandes Brito May 21, 2008 07:52

Re: System Identification and cfx
yOU CAN USE Origin Pro 7.5 software to extract these frequencies, using FFT!

All times are GMT -4. The time now is 04:09.