Hello,

I am trying to understand the concept of phase measurements better in condition monitoring, as I have read a lot of stuff saying for example 'for unbalance, check if the phase is 180 degrees out of phase between the bearings,' but there is rarely papers telling you how to do it (I have to do this using my own software).

There is also the phase measurement shown on a bode plot during start up and coast down, so I am getting a bit confused.

Is the phase in the first example measured by taking the time waveform from accelerometer vertical data for example, filtering the data to 1X and then comparing the phase from different bearings? or am I on the wrong track?

Can it be done with both proximity probes and accelerometers?

Any help is appreciated.

Regards,

Matthew

Original Post

Phase Measurements

The first method requires at least one sensor (accelerometer or proximity probe) and a tachometer signal connected to vibration analyzer. Most of the high-end vibration analyzers easily perform this measurement. For this application the vibration phase is referenced to shaft position at a frequency corresponding to 1X shaft speed. The measurement of amplitude and phase is used for condition monitoring (detect unbalance change), rotor balancing, variable speed test, and relative motion of structure (ODS test) that is limited to vibration frequency at 1X shaft speed.

The second method requires two or more vibration sensors (typically accelerometers) with one serving as a fixed measurement point reference, and the other is a rover that is moved to other measurement points (locations-directions) on structure. Typically, the reference accelerometer is placed at the location-direction with the highest vibration. A vibration analyzer with two or more channels is used to measure two or more accelerometers simultaneously. The Transfer Function is computed with the rover accelerometer output divided by the reference accelerometer output. The computed result is the vibration amplitude ratio and the phase angle of the rover accelerometer to the reference accelerometer. Vibration measurements at several measurement points can identify relative motion along a straight line or in 3D space. Measurements can be plotted manually using Microsoft Excel spreadsheet or with special software using a 3D dimensional wireframe model.

My CSI 2120 vibration analyzer can make both types of measurements. I have also programmed a four channel USB data acquisition module to make these types of measurements.

Walt

()

Thanks Walt.

So the first is basically referencing the 1X to the tachometer signal, and the second is referencing two signals to each other by dividing them and thus giving the difference in their phase if I have understood correctly? So this one is what would be used to check if there was a 180 degree difference between opposite bearings for example?

Matthew

()

When you deal with 180° phase difference between two bearings, usually it's not because of unbalance. I said usually, it depends from rotor width, overhung rotor etc... Every case has it's own story behind.

()

The signal are not actually divided, at least in the time domain as real valued signal.  In effect, a transfer function in the frequency domain would divide as a complex number valued process, or divide the amplitudes and subtract the phases.

()

Matthew,

Yes, and Yes provided you correct phase for different orientation of Rover to the Reference. For example, if Reference is on motor drive-end bearing in axial direction, and Rover in on pump/fan drive-end bearing in axial direction, then subtract 180-degrees to account for accelerometer orientation difference. Most ICP accelerometers produce a positive-going signal outward from the mounting surface.

"for unbalance, check if the phase is 180 degrees out of phase between the bearings" This is not  a general "rule" to detect rotor unbalance. It can indicate a 2-plane rotor unbalance and/or resonant structure.

Yes Bill, I was referring to Transfer Function in the Frequency Domain which, by definition, gives a complex number result from division. I made no mention of Time Domain.

Walt

()

Likes (0)
×
×
×
×