CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)

 jane luo March 25, 2004 13:23

Hi, I have a simple general question about unsteady pressure and acoustics. If I have unsteady pressure on the surface body, say p(x,t), (it's not p'(x,t)), how can I calculate the sound field radiation? Is that by Fourier transform? or you need to integrate the unsteady pressure as a sound source to calculate the radiated sound? many thanks.

Jane

 vasanth March 25, 2004 16:23

hi I have a similar problem.I am validating a linearised euler equation for the prediction of jet noise.I used fft for finding the growth rate of instability in the jet.I think first of all are u solvinga linear or a nonlinear problem.If u are solving a linear problem u will have a maean velocity.You can easily calculate p'(x,t) from p(x,t).Then use Lighthill's theory to find out far field noise from the presure disturbances.Tell me if you know more about it vasanth

 jane luo March 25, 2004 16:53

I am pretty new to this field. Do you mean that if for linear problem, I have p(x,t), u(x,t), which is the coefficient in the linearized euler equation, I can solve for the p'(x,t), which is the sound source and then using Lighthill's acoustic analogy to solve the sound field? Do you know more about how to solve a nonlinear problem?

 john March 26, 2004 00:33

Vasanth, How you did FFT on all cells? I think it is huge task and generate very big files. It is a very challenging job. But sometims doing FFT on selected cells give ressonable results. Long back we tried this but failed to match with experimental data. 1. Solve for p in transient mode using time step based on guess frequency. 2. get p vs t data for some selected cells(guess locations) 3. Do FFT. 4. Find out db level.

This procedure may be wrong since we were not able to match with experimental data.

comment.........

 vasanth March 26, 2004 17:02

The code on which I am working is first a finite difference. There are two phases of the code run 1)transient 2) steady.The total no of points in the grid is 300 X 700. p'(x,y,t)(disturbance)represented by complex no.at every point in the code. For doing fft on every point the code dumps the output(every point's p'(x,y,t))to a file after regular intervals (appropriate interval for fft,should be very careful while choosing the timestep while collecting data for fft).I use 16 files (equally spaced in time,NOTE:Appropriate time). Then I use a postprocessor to do fft to give me complex output.The results I get are good.

May be the timestep you chose for doing fft was wrong.

Vasanth

 Dimitri Nicolopoulos March 27, 2004 11:29

A couple of comments. 1)If you use a 2D grid as the 300x700 suggests, then there will likely be a problem since sounds propagate in a 3D medium and the 2D confinement will yield a large overestimation of levels. 2)The 300x700 grid sounds quite small. In our experience, at least 10 elements per smallest wavelength should be used in order to avoid attenuation and dispersion of SPL while the pressure waves travel on the mesh. In the noise generation zone, a much more dense mesh should be used 3)Do you use acoustic boundaries? If not, the outgoing acoustic waves will bounce back in the domain. For instance, a fixed vel inlet has an infinite acoustic impedance (like a rigid piston moving).

Regards,

Dimitri / www.mcube.fr

 vasanth March 27, 2004 12:10