Vibration problems with an electrical generator 1

[AlbertFM]

Dear all,

I am dealing with a 35 kW and 90.000 rpm electrical generator to work coupled to a micro gas turbine.

I face some problems when I try to ramp-up the generator. When it pass through 550 seems to stay in a critical speed and later on at 750 Hz an other one. We can pass the first one but the second one destroys the system.

Trying to find out what is the source of the problem I have simulated some parts and I've used the hammer test to find the natural frequences of the system and all the components. What annoys me is that all the natural frequencies are at least 3 times bigger and I am not able to find the source of the problem.

Does anybody have any usefull comment that can helps me to solve the problem?

I will be really pleased with your help.

Thanks in advance

 

[BobM3]

If they are not being excited then there would not be a problem.

 

[AlbertFM]

BobM3,

The harmonics can excite them but I'm not sure both together. In any case now I'm trying to reduce mecanical looseness to minimize harmonics.

 

[btrueblood]

Is this a shaker-table test, or (as it seems to be) are you rotating the generator through a range of rotational speeds?

If the latter, what is the driver that spins your rotor, and have you looked at possible torsional vibration modes between driver and rotor (as Greg mentioned 'way back on 27th).

 

[AlbertFM]

Btrueblood,

It is a BLDC motor with permanent magnets. All testing is done by spining the generator through a range of speeds. The driver is electrical (PWM Control) so we are injecting current to the stator of the motor and controlling it with halls and the control electronics. There is no physical coupling to any element (yet).

As I understand the forces that actuate in the system are:

- Magnetic forces
- Contact forces (bearings)
- Inertia and mass of the shaft
- Residual Imbalance
- Possible misaligment
- Possible axial end-play

Please if you think I can face torsional vibration I would like to learn more because I am unaware of it.

Thanks

 

[Tmoose]

Where were your correction planes for the low speed balance?
What is the arrangement of the ball bearings?

The change in response to axial restraint is most interesting.
I had a pesky fat 15,000 rpm motorized spindle with fairly straightforward pairs of precision spindle angular contact ball bearings at each end. Ensuring the floating bearings could actually float was a real challenge. maintaining preload was challenging too

 

[btrueblood]

Nope, sorry, the self-driving element was not made clear to me.

 

[AlbertFM]

Tmoose,

The first time we balance a systme we first balance the rotor by itself. From then we balance all the system mounted. Our correction planes are the two nuts of each end of the rotor. This two nuts maintain bearings, cases, smollies and some more parts fixed into the rotor(see drawing attached). We give more or less preload to the system by controling the torque of this nuts (apart from smollies and other components)

We mount angular ball bearings in each end of the shaft.


See the drawing attached to understand better the way the system is mounted.

 

[TPL]

The generator is home-made
What is the purpose of building it – is it a commercial undertaking or is it some sort of educational project?

The natural frequency of the shaft is around 2250 Hz at one end, 3000 Hz at the other end and 7000 Hz at the middle (thicker) part with the magnetic
This doesn’t make sense – a natural frequency is a natural frequency. Different components might have their own natural frequencies and the assembled rotor might have several different natural frequencies (1st , 2nd …etc). What you need to identify is the rotor critical speeds – these are characteristics of the entire rotor.

To find the fundamental vibration mode shapes and corresponding frequencies I've done FEA, calculations and also hammer testing. In all this calculations I find the natural frequencies 2 or 3 times higher than the problem frequency
How are you measuring the response of the hammer test? What transducer is used? Have you considered that the weight of the transducer might influence your results?How do you know that the response from the hammer impact is from the rotor and not from the support structure. Are you sure that the hammer will excite the rotor natural frequencies?

My hand calculations were just an approximation
So just how good are the assumptions/approximations?

In FEA simulations, Ansys added weak springs to solve the case (at least it was said in an information message)
Who did the modelling? How confident are you that the model accurately represents the behaviour of your assembly?

To think about a reasonable residual imbalance I've used ISO 1940
ISO 1940 applies only to rigid rotors and is for low speed balancing of rotors in a workshop– at the speeds you describe, you should consider your rotor to be flexible. You need to consider modal balancing which takes into account distribution of the mass unbalance and rotor mode shapes. Until you get this right, the most likely cause of all your problems is rotor unbalance.

I haven't found any natural frequency of any element of the system (hammer test) that can cause it.
See above – just because you cannot find it, doesn’t mean its not there. You measuring system is suspect to say the least.

Looking at the sketch that you provided, I again suggest that you consider adding a pair of proximity probes at each end of the rotor, (mounted radially to the shaft and at 90 degrees to each other), going through the casing (just drill and tap a couple of holes or get hold of small button probes) and looking at the shaft just about the positions numbered 4. Depending on the nature of the alternator construction, it would be really helpful if you could get an additional pair of probes to look at the alternator to help identify mode shapes. You also need to obtain vibration data in polar and bode formats.

 

[AlbertFM]

TPL,


What is the purpose of building it – is it a commercial undertaking or is it some sort of educational project?
It is not really home-made. An external company has been developing the project but we have been required to solve the mechanical problems. They will work in the elecrics and we will finish the mechanical part.

This doesn't make sense – a natural frequency is a natural frequency. Different components might have their own natural frequencies and the assembled rotor might have several different natural frequencies (1st , 2nd ...etc). What you need to identify is the rotor critical speeds – these are characteristics of the entire rotor.
As I showed in pictures I attached, natural frequencies of the rotor seems to be 2250, 3000 and 7000 Hz. These are the natural frequencies of the rotor itself because are the ones we have found by hitting the shaft with a “hammer”. I have said that the 2250 Hz seems to be the natural frequency of one leg because when I hit with the hammer there this component is the one that is most amplified. When I do the same at the other leg the component amplified is 3000 and when I hit in the middle the component amplified is 7000 Hz. I think that if the problem was from the rotor I should see some peak around 750 Hs. Please correct me if my assumptions are wrong.

How are you measuring the response of the hammer test? What transducer is used? Have you considered that the weight of the transducer might influence your results?How do you know that the response from the hammer impact is from the rotor and not from the support structure. Are you sure that the hammer will excite the rotor natural frequencies?
The weight of the accelerometer is 1.3 grams. I mount it in a small aluminium base. I have done the test in several ways of mounting it. The results are nearly the same. The picture I sent is measuring the rotor when it is hung up with thin linen so there is no support structure response. I’ve hit with a brass hammer because the rotor is magnetic and I cannot use steel. I’ve hit also with a pen and the results are the same.

So just how good are the assumptions/approximations?
They were just to see if the value was similar to simulations. I’ve done calculation using the distance between bearings, the geometry and mass of the rotor.

Who did the modelling? How confident are you that the model accurately represents the behaviour of your assembly?
I did the modelling and it is almost impossible to create an exact model. If I would need to do an exact simulation I would need to simulate carbon fibber and sinteritzed neodymium. The model respects geometry, constrains and so on

ISO 1940 applies only to rigid rotors and is for low speed balancing of rotors in a workshop– at the speeds you describe, you should consider your rotor to be flexible. You need to consider modal balancing which takes into account distribution of the mass unbalance and rotor mode shapes. Until you get this right, the most likely cause of all your problems is rotor unbalance.
This is something interesting I will have a look at ISO 11342. Just one question, If I were ok with hammer test and there was no natural frequency of the rotor within 0 and 90.000 rpm, How would the rotor pass through critical speeds?. Then the rotor will be rigid or I will need to study deflection due to rotation forces? If the shaft deflects, is it possible to experience a change in natural frequencies due to redistribution of mass?. Please, I would appreciate some help in this point because I think that I am misunderstanding something…

 

[GregLocock]

Bear in mind that the bending mode frequency of the rotor on its shaft will be different when you mount it in the bearings. This is often the troublesome mode for whirl.

Cheers

Greg Locock

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

 

[AlbertFM]

GregLocock,

Does it make sense to do a hammer test with thes shaft mounted in the bearings? I did it and although we can see a slight difference it cannot explain the fault.

Don't you think that if the natural frequencies are above the problems we are facing, means that the problem is mode shaft? I want to test a stiffer shaft and see how the spectrum varies. I will say something after testing is done.

Thanks!

 

[GregLocock]

"Does it make sense to do a hammer test with thes shaft mounted in the bearings?" Yes, absolutely.

I'm still suspicious that you have torsionals.






Cheers

Greg Locock

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

 

[AlbertFM]

Greg,

I will check again the frequencies of the system mounted, just in case.

How could I check if we have torsional vibrations? Which could be the origin of this vibrations?

Thanks

 

[TPL]

An external company has been developing the project but we have been required to solve the mechanical problems. They will work in the elecrics and we will finish the mechanical part

Albert,
At the risk of sounding rude (I would prefer that you regard my comments as blunt and to the point), you are simply not qualified to do this work. You don't appear to have a clue what you are doing and you certainly don't have the tools for this job.

I suggest you turn down this job and recommend that the external company goes soemwhere else for their problems to be solved.

 

[AlbertFM]

TPL,

You're right that I am not qualifyied to solve the problem but since the external company has not been able to finish our generator, we are looking for another company to finish or rebuilt it. At the same time I have to learn something about mechanical vibrations (3 months ago I didn't know what does it mean) to understand our problem and not lose all time.

I didn't expect to explain all this and I would appreciate information that can helps me to understand the vibration problems of my product instead of suggerstions about changing my job (I also don't want to sound rude).

All information you can provide me will be relly relly usefull to me.

Thanks,

 

[TPL]

At the same time I have to learn something about mechanical vibrations (3 months ago I didn't know what does it mean) to understand our problem and not lose all time.

Albert,
Your determination to see this through is admirable but I just don't think you will get there by yourself.

This forum is just not going to provide the huge amount of detailed info that you need.

You have to get better instrumentation and learn how to use it - I suggest you call in an outside consultancy that specialises in machinery vibration problems to assist. They shoud have no problem in allowing you to work with them to solve this problem. You can learn a lot in a very short time - in the short term, this will be quite costly, but you need to weigh up this cost against the cost of the time you will take to get to your own solution.

 

[AlbertFM]

TPL,

The external consultancy is begining in 2 weeks and this is the time I have to learn and test by myself to give them information and be ready to learn with them.
It is clear that this forum will not give me the solution but I am learning important things from people that for sure knows much more than me.

 

[Tmoose]

I think you need -
a- bearings interference fitted to the shafts
b-Bearing OD fitted "loose" at one end to allow axial thermal growth and spring preload effects
c- more balance correction planes, including some on the rotor faces or even along the rotor body

If the bearing preload can be changed by tightening nuts on the shaft, it sounds like the bearings are a slip fit on the shaft. For the conventional motor/generator setup (stationary outer bearing ring, rotating shaft inner ring) this is a very bad arrangement, practically guaranteeing 2 separate bad things-
1 - the shaft will creep inside the inner ring, polishing, wearing, fretting and generally tearing things up
2 - allowing variable random centering changes to the full extent of the shaft-bearing clearance. Multiply rotor weight X diametral shaft/bearing clearance for the size of the gram-inch balance variation devil

Depending on the bearing size, 90,000 rpm can make the inner ring grow significantly, creating or adding to the loose fit and related problems mentioned above.

I'd rate a low speed balance with outboard correction planes as useless, or even counter productive. A low speed balance with outboard correction planes as shown will aggravate first mode bending at higher speeds. Depending on geometry and initial unbalance the rotor can behave "flexibly" (whip) well below first bending resonant speed.

 

[AlbertFM]

Tmoose,

First of all thanks for this valuable information.

Although bearing preload is given by a smalley (kind of flat spring) it is true that the bearing is fixed by the axial force given by the nuts. It can be a source of error, specially for possible radial clearance, but where I think that is the true solution of the problem is at the balancing method.

As I've learned, I have to consider my rotor as a flexible rotor so a low speed balancing doesn't help. What I will try to do (following ISO 11342) is balance the rotor at different speeds; low speed, first flexural mode, second flexural mode, and operational speed. I know that our current equipment is not valid to do this but I will contact the balacing equipment supplier to see what we can do. Referred to the number of planes it could be possible for us to find a third plane at the begining at the thicker part of the shaft but not in the middle.

Thanks,

 

[Tmoose]

what brand and part number are your bearings?
What are the details of the arrangement?
Did the bearing mfr approve this arrangement?

I'm having a hard time imagining how the nuts and a wave spring can co-exist influencing preload.

Again, if the shaft nuts can influence preload then I believe the shaft fit is wrong.