Discrete Event PSD Creation

I have a data file that is 10 minutes long. There is a 30 second section of driving over bumps. I would like to create a power spectral density plot of the 30 second event and export it into Excel so that I can import it into Ansys. Is there a way to do this?

Hi,

firstly sorry for the late reply, unfortunately DEWESoft doesn't have a feature to calculate the spectral density of a signal.

An option to calculate this would be by exporting your signal to matlab format or any other format and then calculating the PSD in 3rd party software.

Thank you for the reply. I was able to create what I mention above, but now after your comment I question if it is doing the math correctly; here is what I did.

I export the 30 second clip of interest as a DEWESoft file, then opened it in it's own application. In "Setup", I added "FFT Analyzer", selected the accelerometer channels of interest, then selected "Amplitude", "Overall RMS", "Overall (Averaged)", and "PSD" from "Amplitude Type".

I assumed this would create an averaged PSD from the entire 30 second time trace, am I incorrect?

Hi Jeff,

I would like to apologize, I didn't know that the FFT analyser, Fourier transform math and FFT graph had the option to calculate the PSD amplitude of a given signal. I haven't used this feature so I overlooked it.

The PSD in DEWESoft works correctly. The options that you selected calculate the whole (30s) average of the datafile.

I compared the data I get from DEWESoft to the one in matlab and they are the same as you can see in the picture.

In Matlab I used the SPD found here: http://www.mathworks.com/help/signal/ref/dspdata.p...

In the attachment you can also find the matlab example. In DEWESoft you must select the rectangular window to get the same results.

I hope this information helps you and that you can successfully use the PSD graph.

Best regards

Anže

I'm doing something very similar to Jeff. I'm running a random vibration analysis on a component attached to a Vehicle.

The data was collected while the vehicle was driven through a highway, streets, and then off-road, in a single go.

In the past, I've only used standard PSD profiles like Mil-std, etc. However, this time we've collected the actual vibration data by mounting the accelerometer to the vehicle.

Now, I wanted to generate a PSD profile based on the collected vibration data. The issue I'm having is that there are so many options to select before generating a PSD profile which I haven't used before, and I'm wondering if I'm selecting the correct options to generate an accurate PSD profile that would suit our application.

(I've already read the manual but I couldn't find all the answers I was looking for)

Here's what I'm doing:

1. Offline Math --> Add Math --> FFT .

2. In FFT Setup :

Selected the required channel.

"Spectra Tab" : (Following are the options I choose)

Amplitude, PSD, Peak, Lin(Z)

Blackman Window

16k Lines

Overall Averaging, 0% overlap

"Scalars Tab" : Full range (Everything left to default)

3. "Review Tab" --> Widgets --> 2D graph -->(Then display the FFT function in the 2D graph)--> Auto-Recalculate

4. "Export tab" --> (Then I export the data of FFT function to an xlsx file)

Please let me know if I'm selecting the correct options, and if not, then which ones should I select.

Secondly, I've observed that if I generate a PSD profile based on the complete set of data, the Peak points of the PSD profile become smaller than If I generate a PSD profile based on the cropped data of the dirt track only (I suppose this is due to averaging, unless I'm doing something wrong). Hence, I concluded that generating a PSD profile from sets of data collected for each terrain would be better than using a single set of data to generate a PSD profile for all terrains combined. However, in this case, I'll have to run the FEA analysis for each profile which would be computationally more expensive. Please let me know if my conclusion is wrong.

Your help would be highly appreciated.

Thanks.

Dear Syed,

We would recommend that you split the datafile into sections regarding the terrain. Then you could calculate and export these sections separately.

Or you can make a keyboard trigger in order to ''mark'' where the vehicle was on a different terrain. Then the post-analysis would be easier and you would not need to make multiple tests.

Best Regards.