Accelerometer Signal to Detect Shaft Imbalance

[Moh]

Good day every one,

I've been trying to find a way to detect imbalancing in a shaft fixed to a motor in a simple educational package, the problem is that vibration is not my field. I wonder if there is any one out there who can help me, i guess i dont need much but some hints. I do appreciate your efforts in advance.

I've installed an accelerometer on one of the bearings and am facing the following problems,

1- I read that I probably have DC oftset and I have to get rid of it. I am not able to do that, I viewed NI example about that but couldnt do it , also i couldnt deal with the FIR filter. any help about that please?

2- My FFT plot is not fine !! .. I am getting peaks everywhere .. and those peaks dont move with the speed of the shaft .. ( I knew they should )

3- I cant understand the frequency resolution thing and how it affects my measurements .

4- what's the frequency at which i would note the imbalance ?!

I am getting imbalance myself please help

Moh

[preston_johnson]

Thanks for the questions. 

1) is the accelerometer powered?  Most accels, IEPE or ICP, require a 24V power supply with 2 to 4ma of current.  If the accelerometer is not powered, then you will get a crazy FFT similar to what you describe. 

2) the DC offset comes from the power used by the accel.  Most vibration acquisition boards (aka - usb-9233) provide an AC coupling circuit that removes the DC.  THese boards also provide the power.  If you are powering externally, your DAQ board needs to have +/- 24V range to see the AC signal (vibration) riding on top of the power (DC offset = power source for accel).  You can use a simple voltage divider to reduce the voltage if you need.  Once you have the signal digitized, the easiet method of removing DC, is to capture enough data in one block of acquisition that you have several rotations (10 or more).  You can use the statistical median then to find your DC value, and simply subtract this from your original array of data.  This is less sophisticated than the high pass filter, but it works well if you have enough cycles (rotations) in your data block. 

3) frequency resolution is the number of points in your FFT plot.  It is also know as delta-f, or the distance in frequency of each data point in your FFT.  Better resolution means smaller delta-f.  You can reduce the delta-F by collecting more data (longer data acquisition time).

4) unbalance is a one times rotation vibration, it will appear in the FFT at running speed.  You will need to know the speed of your machine to find the unbalance peak.  Do you have a a tachometer?

Consider reviewing the following materials:

http://zone.ni.com/wv/app/doc/p/id/wv-102

http://www.ni.com/industrial/machine_vibration.htm

Message Edited by Support on 01-10-2008 11:16 AM

Message Edited by Support on 01-10-2008 11:16 AM

[johnsold]

In addition to Preston's suggestions you may find it necessary to experiment with the location of the accelerometer. Just because it is mounted to the bearing does not mean that you will get good information about the balance. A ball or roller bearing may have a stronger signal at (Number of balls in bearing) * (rotational frequency) than at the rotational frequency. If one or more of the balls is worn, the signal may be even more complex. Since you described this as an educational setting, is the unbalance something which can be removed and reconnected easily? If so, look at the signatures with and without the unbalanced load. The vibrations with the balanced load may include bearings, torque cogging, and other things which will not change too much (you hope) when the load becomes unbalanced.

Also, double check your frequency calculations. It is easy to make a mistake when converting from rotational units to conventional frequency units.

FFTs like long data series, where long is relative to the period of the lowest frequency of interest. As Preston mentioned 10 or more rotations of the shaft is a good starting point.

Lynn

[Moh]

Thanks gentlemen ,

The accelerometer is not powered. It is connected to the NI 4472 that supplies the enough current to it, i guess.

i'll try the statistical median method for the DC offset.

about balancing, this is educational kit , i can connect and disconnect the weights easily to get it balanced , I also have a speed sensor , but what i am seeing here, is that the peak is not changing with the speed ( 1x rpm ) , or actually there is no specific peak! .. the FFT am getting looks really depressing! i've been studying vibration since a month and theory is far differnet than what am getting here.

Thanks one more time.

[johnsold]

Attach a simplified copy of your VI with some measured data. Someone will probably be able to offer some suggsetions. By simplified I mean that you should remove the data acquistion parts since no one will have the hardware setup to duplicate your system. Here is how to do it:

Run your system with an indicator connected to the output of the data acquisition parts. Select the indicator and Make Current Value Default. Then remove all the data acquisition VIs and change the indicator to a Control. At this point you should have a VI with test data and no hardware dependency. Save it and post it to the list. The Choose File button for Attachment is below the box you type yourt posting into. Also specify your LabVIEW version.

Lynn

[preston_johnson]

If you are using the 4472, insure you have IEPE turned on - this turns on the power.  Also, make sure you have the 4472 in AC coupling mode - this removes the DC offset.

How fast is your rotor spinning?

[Moh]

Thanks Lynn for the tip. Attached the file.

Thanks Preston, but i couldnt find the IEPE and coupling options , there are available at the test panels only. It seems that power is on, i mean that the signal changes when unbalanced or speed changes.

the rotor spins in the range of 1000-8000 rpm.

I use LabVIEW 7.1

Moh

[Moh]

here's another VI which shows all signals .

Please advice.

Moh

[preston_johnson]

What version of DAQ are you using?  Are you using DAQmx?

[Moh]

Yes I do.