Measuring axial natural frequency of rotor in its bearing

[jeyaselvan]

I was trying to measure the axial natural frequency of a male rotor on its bearing(antifriction) in rest ( load equiavlent to its normal operating condition). I have prediction value (just the first mode, which is very similar to SDOF system) based on bearing stiffness and this is 386Hz. The setup is that for a screw compressor and the tests are carried out with gears & female rotor removed. The axial bearing stiffness is proportional to the load and this has been considered in the prediction.

In the experiment, the rotor is loaded to equivalent load while in operation, through hydraulic jack.

I am not able to get anywhere closer to my prediction. Suspect something is wrong with my experimental layout. Kindly for your suggestions / feedback on the attached experimental layout.

 

[Strong]

It is not surprising why the test data is bad. The load cell, jack and floor have been added to your spring-mass system. A static load should not be necessary to conduct a dynamic test. Just utilize gravity to be sure that the bearings have a slight preload and do not bounce during the impact test.

Walt

 

[electricpete]

Some thoughts fwiw:

In the experiment, the rotor is loaded to equivalent load while in operation, through hydraulic jack.

In your test the load is at the end. If I'm understanding right, in operation, the load is distributed and also has some mass attached (the other rotor). Those are differences.

The axial bearing stiffness is proportional to the load

So it is a non-linear model of a bearing....force prop to displacement squared? I guess that would be close for angle contact bearing. Knowing the preload would be important.



=====================================
(2B)+(2B)' ?

 

[GregLocock]

Walt's point is the problem. You need to simulate the load in such a way that it does not modify the dynamics of the system in an unrepresentative fashion.

That can be easier said than done. One option might be to do a modal analysis on the running machine. The data quality will be awful, so you'll need to test for a long time. Synchronous sampling may help.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[electricpete]

Recreating the running system was my point as well. I don't think gravity would do it.

=====================================
(2B)+(2B)' ?

 

[jeyaselvan]

Thanks for all of your replies.
Walt : I was trying to simulate the operational load as static, since the bearing stiffness is influenced by the load. In fact I am trying to create a scenario of bearing stiffness at operating stiffness by applying an equivalent static load. Hence apart from the gravity load I need to apply additional force.

Electripete: Yes, the stiffness of the bearing is non linear...the load is applied to the entire shaft face, instaed of rotor face, owing to the experimental layout..u r right, during operating condition, the load is fairly distributed on the rotor face / lobes...but I am not sure how to do it real time testing

Greg : You are very right..to my boss it is just a simple free response test on a rotor.., but doing it real time without altering system dynamics is pretty challenging.. the problem with doing this operating condition is that under assembled condition, I will be left out with no access to the rotors in its bearings (have access only of the housing).

Also tried mounting the system horizontally (photograph attached), which is more close to the normal orientation.. but am still not successful..

 

[Strong]

If you can mount a proximity probe (eddy current or capacitance type) through a case cover in the axial direction, then you could impact the compressor housing in axial direction during normal operation. You would use the "negative averaging" method to measure the natural frequency. If you want to apply the load on the test stand, I would consider using a dead weight (mass load) so that you could correct your test results easier.

Walt

 

[GregLocock]

You only need data from one location to vaildate your frequency estimate, I would have thought the casing would respond at resonance even if you can't get to the rotor itself in the running condition.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?