Resonance of mechanical structure

[nkb]

Hi

I am designing an electrical machine which has 3 cubical of size (30”x20”x10”) connected together side by side. Eaxh cubical has a fan for cooling.

I was told that I should find out the natural frequency of the structure so that it does not create resonance with the FAN frequency.

My question is:
Can any one tell me how the natural frequency of this type of structure is calculated and also if possible tell me material to read for my knowledge in this area.

Thanks in advance,

nkb

 

[blackeye]

Well you could predict the structural modes using finite element analysis or you could test the hardware using modal analysis. Actually, the best thing is to do both so that you end up with a correlated finite element model. Then you have the confidence to predict changes to the model before cutting metal. This way you might create a revolutionary design, rather than evolutionary.

You might also want to predict the acoustic modes inside the enclosure.

If you're looking for software then EDS/SDRC/MTS I-DEAS will do all of the above.

 

[nkb]

Thanks Blackeye for your reply

Can u please tell me how to test the hardware using modal analysis.

Thanks once again.

Nkb

 

[electricpete]

A simple approach used by vibration folks in the field (not designers) is a bump test. Whack the structure and monitor the vibration time waveform and spectra to determine resonance.

A tricky part is determining where to whack and where to monitor. You will want to try several combinations that you think will excite modes in the frequency range you are concerned about.

 

[blackeye]

Step 1 - Support the structure, grounded or free-free.

Step 2 - Excite the structure with an instrumented hammer or exciter at the minimum number of locations which excite all modes of interest.

Step 3 - Measure response, usually with an accelerometer at enough locations to describe the modeshapes over the frequency range of interest.

Step 4 - Measure at extra response locations to make your test model look like the test article.

Step 5 - Curve fit the frequency response funtions (frfs) between all excitation and response locations to provide a modal model of natural frequencies, modeshapes, damping, modal mass.

Step 6 - Synthesise frfs from modal model and compare with measured data to check accuracy of model.

Step 7 - If appropriate, perform correlation studies (frequency difference, modal assurance criteria, orthogonality) with FE model.

 

[electricpete]

In the absence of having any detailed model of the system, it all boils down to a bump test (at intelligently selected points), right?

 

[nkb]

Thanks a Lot for guidance. The info will certainly help me.

Regards,
NKB