Pre and Post random signature sweep discrepancy

[NaikD]

All,

We are subjecting an instrument to random vibration along 3 axis. Low level (0.25 g) signature sweeps are performed before and after the random vibration.

Random vibration along first two axis went well. For third axis pre random signature sweep had a distinct peak at 135 Hz. After full level random vibration, the post signature sweep showed a small blip at 135 Hz, but the amplitude was 1/10 of pre random sweep. About half of the higher order modes had similar pre and post sweeps indicating an interface might have slipped.

The resonance at 135 Hz was that of an antenna (spherical/cylindrical berylium part 0.015" thick attached to drive shaft supported by two duplex angular contact bearings). After carrying out FRB (failure review board) meeting we continued testing of sine and sine burst tests that were satisfactory including limited performance test on the instrument.

I do not think that the discrepancy in pre and post radom vibration signature sweeps after last axis random vibration test were due to bearings settling in.

Any ideas what this could be due to? Have you seen something like this before?

Thanks in advance.

- Dipak

 

[CESSNA1]

NAIKD:

With the information it is difficult to figure out what is happening. Here is some stuff you may want to consider.

Apparantly you are doing a sine sweep resonance search and then doing random testing. There is no need to do a sine sweep. The resonance search can be accomplished with low level random.

The part you describe seems very small. If you amount an accelerometer on it you may be mass loading the piece.

Do you remove the accelerometers after the sine sweep for the random? If so are they reinstalled them in the correct location.

Are the accelerometers calibrated and functioning properly?

Are the duplex bearings rolling element or plain bearings? If rolling element with grease the vibration could cause the grease to move adding damping to the rolling elements.

What do the calculations show for the expected natural frequency?

How are the accelerometers mounted? beeswax, cyanoacrylate adhesive or mechancial attachment are preferred. The attachment must be firm. Thick beeswax or thick CA can behave as a spring.

Regards
Dave

 

[NaikD]

CESSNA1,

Thanks for responding. The random vibration test that we do is to find out workmanship problems (solder joints, assembly etc.).

We compare the low-level sine sweeps pre and post random vibration to see whether they match. Any deviation (frequency shift) indicates that the assembly is settling or something has changed. Normally we allow 10% frequency shift for settling bearings, screw joints etc. But even then normally so far the pre and post random vibration low level sine sweeps matched exactly as if they were zeroxed copies of the same chart.

This time the mode at 135 Hz completely disappeared. A close investigation showed that the original 135 Hz mode split into 127 and 135 Hz indicating a structural interface may have slipped.

We re-ran the random vibration test and there were no more changes in the pre and post low-level sine sweeps in the repeat test.

Somehow since we cannot disassemble the assembly to investigate the root cause, I wanted to hear from the vibration community here if they experienced something similar to this.

To answer your questions:

1. We do not think that the accelerometer is mass loading the piece. Since the accelerometers were always there for pre and post low-level sine sweep as well as for the random vibration it would not make a difference in the low-level sine signature charts. Also, the discrepancy happened only in one (last) axis. Other axes results were fine.

2. The accelerometer were calibrated and function properly. We checked the functioning of the accelerometers again after the discrepancy, and they found to be perfect.

3. The duplex bearings are rolling element with grease lubrication. Since the random vibration results were good for first two axes we do not think that the grease is influencing the results for the last axis.

4. The calculated (from FEM models) mode at 135 Hz showed in pre random low-level sine sweep for the last axis where the discrepancy occurred.

5. The accelerometers were mounted using adhesive.

Thanks again for your suggestions.

Regards,

- Dipak

 

[Higgler]

Being an antenna engineer, I suspect the antenna thickness was just the thickness, and it had some length to it, like 1 to 4 inches.
Antennas need attachments to outside hardware. If the antenna was suspect and it was a printed circuit 15 mils thick, either the connection to the outside was the cause, or possible adhesive in the bonding (or antenna part to antenna part, or antenna to outside housing) was letting go or maybe the copper cladding is letting loose. The copper on circuit boards isn't as strong as you may think. And the peel strength isn't even over the full circuit board especially if there were high temperatures involved, and maybe it partially let go.
kch