Traditional vs. operational modal analysis

[Vibac]

Hello

I wondered if any of you have had experience comparing traditional (experimental) modal analysis with operational modal analysis? To my understanding, from reading the B&K website etc. it seems that the latter output-only method provides a very versatile and practical alternative to the traditional method where you need to control and measure the input force. Let's assume the test subjects cover both machinery and civil structures such as bridges.

The shortcomings I've noticed so far are:
- Modal shapes are generally unscaled
- Harmonic excitation from the test object itself (e.g. rotating machinery) may show up as natural frequencies in the analysis.

Do any of you experienced people have practical experience regarding the limitations of Operational Modal Analysis? How about accuracy in terms of modal parameters compared to the traditional method?

Regards
René

 

[SomptingGuy]

I've always though it unfortunately named, because it is simply a measurement of the response of a structure in its operating condition. It doesn't really tell you anything about the modes of the system, although you can usually infer them from the response. In layman's terms you get a scaled up animation of the motion of a structure, frequency-by-frequency.

We've used the technique primarily for problem diagnosis (e.g. finding the causes of interior booms). You would normally expect see some component going absolutely AWOL at the problem frequency/speed (e.g. an alternator or power steering pump). If you do find a single component like this, you win - change the mounting and retest.

"Operational modal analysis" is not a lab technique! You get burned, cut and bruised. Then you make your computer dirty.

 

[SomptingGuy]

Oops, I was thinking of Running Mode Analysis (operation deflection shapes, etc). The likes of B&K/LMS tend to make up so many overlapping and confusing names it's easy to mix them up.

 

[GregLocock]

I think SG is on the right track, OMA is the same as ODS, so far as I can see.

On the one hand it is almost useless for correlating FEA models, which is the only way you'll ever get the screenhuggers to produce useful results, but on the other hand it is a handy tool for the development engineer.

If you think about it it is just a high tech version of running a screwdriver down the driveline, or using a strobe.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[SomptingGuy]

I still dispute my first post. These two pages from the LMS website are distinct:

This is what I've done many times before and described:

http://www.lmsintl.com/test/24-channel-stationary-operational-deflection-shape-analysis

Whereas this looks more like what Vibac was describing:

http://www.lmsintl.com/test/24-channel-operational-modal-analysis

I'm obviously not endorsing these products and have no connection with LMS. It's just that their descriptions seem clearer than those on B&K's site.

 

[GregLocock]

The data is the same. So far as I can see the ODS is the exact equivalent of a strobe, OMA is taking the same data and throwing signal processing horsepower at it, in the hope that some underlying properties can be extracted.

OK, I agree there is a difference. If you can make safe assumptions about your forcing functions then you may be OK.





Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Vibac]

Dear SomptingGuy and GregLocock

Thanks for your comments, and yes, I agree that the names ODS, OMA etc. create some confusion, they are too similar.
Here is a link for you:

http://www.iomac.dk

A conference was held quite recently concerning operational modal analysis. I'll check the proceedings and see if something can be found about accuracy etc.

Regards
René

 

[GregLocock]

The problem isn't accuracy, the problem is the fundamental assumptions.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[cincibcats]

I've directly compared the results from OMA and traditional modal analysis for a few simple test cases and for some theoretical derived lumped mass systems. My overall feel is that OMA, when properly applied, is very good at determing natural frequencies. However, depending on the structure, excitation, sensor distribution, and other factors the mode shapes and damping are less accurate than traditional methods.
I agree with Greg that traditional modal will be far superior for correlating an FEA model.

 

[jungi]

Hi,
For correlating with FEA there is no alternative than EMA (experimental modal analysis). OMA only shows how the system responds to the current excitation (like in rotating machinery) which doesn’t have to mean that all modes, from the frequency band of interest, will be seen. Using OMD, it can just happen that will miss some natural frequencies because they just weren’t excited. EMA, if properly done, gives you all the natural frequencies you need. Of course OMA is very useful if you are looking for vibration problems in your machinery, but it is not secure to compare it to natural modes extracted by FEA.

 

[GregLocock]

As an advocate of hammer testing, a cheap, simple way to check that your OMA is reasonable would be to give your instrumented structure a thump with an appropriately sized hammer, in an appropriate place.

Even though the magnitude of the force is unknown, as is its bandwidth, at least you can compare the low frequency behaviour and check that you aren't missing any modes.

Of course for the price of one force gauge you could be doing a proper modal anyway, in that scenario.

Incidentally one of the most interesting ODS tests I did was the first one I ever did, which was the backing plate of a squealing disc brake pad - proving that the steel plate was bending at the same wavelength as the brake disc. That was fun.


Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[WCFoiles]

Some of the structures tested with OMA have been large like bridges. It takes a big hammer for a bridge.

Regards,

Bill

 

[Vibac]

Hi Bill

One way an impact test has sometimes been performed in the case of bridges or ships is to drop a very heavy load onto the structure. If you apply some sort of transducers in between to measure the force you might gather input data similar to the hammer test.

Regards,
René

 

[GregLocock]

The classic way to excite a bridge is to release a tensioned cable.

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Rob45]

One point about OMA and ODS I'd question:
Why worry about correlating your FEA model?

Once you've got something that's "operating," playing with the model is just something you do for fun; what you should be doing is testing the actual real-live machine.

 

[GregLocock]

I hope you are being funny.

Reasons why I like correlated models

1) I can make changes very quickly

2)My new design will start from a known baseline before any hardware is available

3) there are details I can interrogate in a model that I cannot measure in a real system (eg, suppose a stress concentrator exists that will only show up in 1000 hours of testing)

Note that I am not saying analytical models are better than real test, but a /correlated/ analytical model is a very fast and powerful tool. An uncorrelated model is just an opinion.



Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Rob45]

Well yes, I suppose I should have used the /humor html tag at the end...

Unfortunately, my organization never seems particularly interested in correlating the FEA models: management seem to think that once the model is complete and the analysis results are available, what else do you need?

And that's when they bother to budget for FEA models at all.

Regards,
- R

 

[jungi]

Similar attitude at my company they are so impressed by moving, colorful pictures. “If something is so beautiful it has to be right, so what do we need model updating for?”