Natural frequency in Vibration sweep test 4

[elogesh]

Hai,
I have a problem regarding natural frequency measurement in vibration sweep test. The natural frequencies of the component in three different orientations (perpendicular to each other) is evaluated from FEM and found to be 50Hz(bending mode), 120Hz(bending mode) and 360 Hz (axial mode) in horizontal orientation, vertical orientation and axial orientation respectively. Then the component is mounted on a fixture and fixture is mounted on vibration shaker. The vibration shaker is from Ling dynamic systems/Dactron. Then Vibration sweep is carried out from 10 Hz to 2000 Hz in order to find out the natural frequency. The component is mounted in horizontal position, so that horizontal mode gets excited. Using accelerometer the displacement values monitored against frequency. Surprisingly the horizontal frequency is found to be 30 Hz (FEM predicted = 50 Hz). The procedure is repeated for vertical orientation is 35 Hz (FEM predicted = 120 Hz).
I checked the FEM model and found to be ok. Even for modal correction, I can expect + or –10% variation. Now, I am doubtful about experimental measurement. One of possible reasons for difference between FEM and experiment is, thin fixture used for mounting the component. They used thin fixture, since they have weight restrictions that can be tested on shaker.

How can I find experimentally that the natural frequency is of component or component with the fixture?

While writing this thread, I got a strike that during experiment; we measured displacement-using accelerometer on component only. I should have done that in fixture also. I will do it?

Apart from that is there any other way of checking the ambiguity in experimental testing? This is because we have limited knowledge about experimental testing.

Regards,
Logesh.E

 

[madal]

Hi Logesh.
In my opinion You're right when you think that fixture has the most probable responsability. I had a lot of experience in these problems and in every case I found that the stiffer the fixture, the most accurate results I had.
A couple of question regarding test on shaker:
1. where did you put the accelerometer used to control the bench?
2. did you have a look at mode shapes at natural frequencies with stroboscopic flash lamp?

If I were you, I'll try 2 ways: the first, why not modelling on FEM also the fixture to see its influence also from a simulation point of view?
The second: modify your fixture to make it more stiff, eventually using low density materials (aluminum?)
Of course I don't know how complex your component is and what performance your shaker has...

Regarding your last doubt: for sure you must verify also fixture behavior. The best way is to start with a first sweep in which you put the control accelerometer on shaker table directly and an analysis accelerometer on a representative part of the fixture, for example nearby the component connecting points. In this way you'll find fixture resonances. After that you should switch shaker control to the 2nd accel. and use the first one as analysis, together with other accels. in various parts of your component.

Last answer: you may try to discriminate nat frequencies of fixture and components unloosing a component fixing screw. If natural frequency lowers, it's a component one.

Good luck and keep us informed with your work.
Ale

 

[GregLocock]

Another alternative is to reanalyse the thing in FEA, free-free, and to do a hammer test on the physical part, again free-free.

Although hammer tests are not good for a full modal analysis, they are unparalleled for investigation work.

Cheers

Greg Locock

 

[elogesh]

Hai,

Thanks to Ale and Greg.

Ale

1. where did you put the accelerometer used to control the bench?

Do you mean the accelerometer used for sending feed back signal to shaker and control the vibration level "g".Then the accelrometer is mounted at corner of vibration shaker. In the plan view the vibration shaker appears as circle and acclerometer is placed near to extreme end.
The vibration signal from the component is measured separately through universal vibration meter.
When vibration sweep is carried out from 10 to 2000 Hz, the instantaneous frequency appears in the computer monitor attached vibration shaker/dactron unit.At every 5 Hz intervals the vibration level(in displacment) is noted down.
Multiple channel analzers should have reduced the human effort.
2)Did you have a look at mode shapes at natural frequencies with stroboscopic flash lamp?
We have don't have stroboscope.

3) the first, why not modelling on FEM also the fixture to see its influence also from a simulation point of view?
I have plans to do it.

4)The second: modify your fixture to make it more stiff, eventually using low density materials (aluminum?)
In our company, testing group uses mild steel for making fixtures.It may be odd to convince the testing groups, if there is only a marginal benfit.But I will discuss with our testing group.

Definitely, I will keep informing about the progress.

Greg Locock

Unfortunately, we donot hammer testing setup.

Once again Thanks for your useful replies.

Regards,
Logesh.E

 

[madal]

In spite of your little lack of testing equipment (a bench controller which is able to manage directly up to at least 8 channel would be really useful...) you could try anyway a different approach.
To "filter" fixture resonances, try and put the control accelerometer (Yes, it's the feedback one, used to control in closed-loop the imposed acceleration) near to your component fixing point instead of on shaker "corner".
In this way you will be quite sure that your component will be tested with a real flat acceleration profile of your choosen amplitude (say 2g 0-peak, for example, which is my standard).
Another general suggestion is to use a sweeping speed of 1 octave/min at maximum.

Bye
Ale

 

[IRstuff]

What is the aspect ratio of your UUT?
A top-heavy UUT can cause a conventional slip table to rock during vibration.

What kind of vibration table is it?
The standard slip table is prone to UUT induced resonances in the tabletop/oil film, due to the extra degrees of freedom in the vertical axis of the table.

We've have better performance with a hydrostatic bearing vibration table.

TTFN

 

[GregLocock]

You /do/ have a hammer testing setup.

Even if you don't have a suitable force gauge, just whacking the structure with a screwdriver or a brick or your fingernail (depending on its size and robustness) and looking at the peak hold spectrum of the accelerometer signals will tell you whether the frequencies are right.

Cheers

Greg Locock

 

[Rob45]

One more point:
If there were in fact a 50 Hz mode of the part, and the fixture mode was 30 Hz, then an FFT should show you BOTH modes, with peaks at 30 and 50 Hz.

But I like the hammer test, which we at Ford, in the days before the force-hammer, called the "K-Bong (or ka-bong) Test."

When we retired the brass mallet, I put it to use as a knock-off hammer for my E-Type.

 

[elogesh]

Hai,

Back to the thread with some more findings.

1) Confirmed it is fixture resonance in testing as well as FEM.

Following are the readings taken from Experiment on fixture & in component.

Fixture

Hz Amplitude in microns

10 300
15 80
20 50
25 64
30 160
35 32

Component

Hz amplitude in microns
10 1530
15 968
20 870
25 133
30 1548
35 372

Hence at 30Hz amplitude is more in Fixture as well as in component. The readings are taken with sine sweep and hence the trasient effecfts weren't died down.But I don't think that is going to affect natural frequencies very significantly.

Second, I ran FEM analysis of Fixture with component(somewhat crude model). The natural frequency is 38 Hz. The 8 Hz may be corrected using exact fixity conditions and refining the model morethan doing modal correlation for this particular case. Since, I have taken very rough model of fixture.

Thanks for all your response.With all your help, I could able to resolve the problem in fixture.
We are planning to go for rigid fixture.

Next, I have to pass the product at 67 Hz, 25 'g' according to JIS specfications, which is a tough task.

Regards,
Logesh.E

 

[kaiserman]

Elogesh:

This thread describes an isuue we face every day in our validation lab. I would like to comment. We have found that virtually all of our problems come down to the boundary conditions of the FEA model. These are tough to duplicate.

As you are finding out, it is extremely difficult to duplicate a test fixture to match the field impedance. One might use a small section of the field structure itself. This assumes that the field structure has been designed and the test item in the field structure is know. A fixture can then be modeled around the sectional piece. The field inputs must then be known at the sectional mounts.

In our lab we either make use of sectional pieces, or we use fixtures with infinite impedance. In the later approach, no attempt is made to match the structural impedance that the test item would see in its field installation. In fact the fixture used to connect trhe speciman to the shaker is as rigid as possible. This of course leads to heavy, stiff fxtures. Using this approach requires that each mounting point is held uniaxial and uniform. Again, this is hardly realistic as far as the interface is concerned. Aside from the convienence for testing there are other reasons for employing the infinite impedence approach. The main reason we ahve adopted this approach involves various legal aspects. The buyer and the seller know when the conditions of the contract have been satisfied. Secondly the test must be exacting and repeatable to offer any tiype of comparison.

Virtually all of our test fixtures are resonant free out to preferably 2x's the highest frequency of the bandwidth being tested. Thus if you are testing a product from 10 to 1000 Hz - the the first reosnant frequency of the fixture should be ideally 2000 Hz. It can be done! We have virtually 750 test fixtures that have resonant frequencies at or over 2x's the highest frequency being evaluated.

You might considering encorporating the fixture as part of the overall FEA model.

Over the years I have seen many, many FEA models for our particular application be wrong, and require tremendous modifications. Granted, some are good out of the shoot. I have only seen a few modal or lab tests have errors. I get tired of hearing my FEA is right and your modal is wrong, only to confirm the lab and modal results. I am very partial to the modal test being superior to the FEA model out of the shoot. I would trust modal lab data way before I would trust a FEA model. I can see, hear, feel or measure something in a lab. I have yet to see anyone driving a finite elemental car!

My guess would be to modify or tweak the fixture and then the FEA model. Good luck.

Regards,
Kaiserman

 

[Rob45]

What I find a little curious is Elogesh's mistrust of the experimental results since they conflicted with the FEA...

 

[IRstuff]

Experimental results are often influenced by uncontrollables and things that were not modeled. As indicated above, we had a situation where our UUT was found to have a 35 Hz resonance, but our FEA claimed >50 Hz.

As it turned out, the FEA was correct and the problem was the test equipment. Going to another type of tester showed that the FEA was correct.

TTFN

 

[elogesh]

Hai,

Kaiserman and Rob45: I concur with your statement. I trust on experimental results.
I don't know how it was inferred that I have mistrust on experimental results?

In my case, I felt the experimental natural frequency of 30 Hz is the actual. The predicated value of 38 Hz from FEM is due to poor FEM representation of actual physical model.I felt, If I improve the FEM model I will get frequencies closer to 30Hz. Then modal correlation can be done to get very close to 30 Hz.Since the natural frequencies are low, I felt usually FEM predicted value will be close to experimental value.

Regards,
Loganathn.E

 

[CESSNA1]

DEAR ELOGOSH: Sweept sine is not necessary for a resonance search, It can be, and usually is, performed by just using the 5 - 2000 hz random excitation. The resonance of structures mounted on a shaker is different from the resonance tbat is calculated. Also material properties and constraintes vary . Mounting a piece vertically, particularly if it long and slender should not be done. Cross-talk in the shaker head can affect the results. I suggest taking some good vibration/fixture design coures.

regards
Dave