Equipment: Joy Air Compressor (TA-85) Horsepower: 2250
Fault: High Vibration Trip, HS Rotor, Coupling assembly error
Speeds: Motor 3600 RPM, LS Rotor ~22,960 RPM, HS Rotor ~31,990 RPM
When one of these machines trips out due to high vibration, it can be very intimidating as the plant is immediately faced with, dare we try to run it again at the risk of more damage, or can we put together an action plan from the information we can pull together right now? This case study is a review of how a high vibration trip was handled successfully in this instance.
#15 Air Compressor was taken down for planned maintenance. The work to be done was replacement of a leaking seal on the first stage to intercooler piping, changing oil in the motor bearings, some intercooler drain piping work, and other minor repairs. The couplings were not scheduled to be opened. All planned work was completed, but when the compressor was restarted, a higher than normal vibration was noted by the operator on the compressors prox probes. When the operator tried to load the compressor it tripped out on high vibration.
The plant had the good fortune of the condition monitoring group collecting data the morning this failure occurred, prior to the maintenance work beginning. Figure #1 is that data. This data was taken with the plants walk-around box with a standard 100 mV/g accelerometer on a magnet. The cursor is placed at turning speed of M2H (motor, coupling end, horizontal). The data showed more energy in it than normal and it was obvious that something was very different in the machine. This is a sleeve bearing motor and both ends appear to show sidebands, but the 2 ends appear to show different d-frequencies, (which should be the same if they are true sidebands). However, this was found to just be a resolution issue.
A sleeve bearing motor should have very clean spectrums as they usually show only a small amount of 1X the RPM and a few harmonics but nothing else. Sidebands normally are not seen in sleeve bearing motors so the plant questioned if they really were sidebands, and if they were, what could cause them in this motor?
It is important to understand where sidebands come from and what conditions cause them to appear. There isn't some single event as a shaft rotates that causes the sideband; they are a function of the FFT calculation as it can not handle frequencies cleanly that are not periodic or repetitive in the time signal. A typical example of a machine that we expect to see sidebands on would be; a shaft with rolling element bearings that has a bearing fault. The fault will cause a modulation in the frequencies of the time domain signal as it rotates. Sidebands then appear in the spectrum when the FFT is calculated from this modulating time signal.
As the inlet piping to the second stage had been opened to replace the seal, the immediate concern was the possibility of impeller damage due to debris or some type of maintenance material passing through the compressor. If this were the case, it would manifest itself as impeller unbalance at the damaged wheels turning speed. This plant experienced such a failure many years ago on a similar but smaller compressor when an inspection mirror passed through a 2nd stage impeller. That high $ memory still burns bright, and was of great concern.
The prox probes only monitor the High and Low Speed Rotor shafts of the compressor (not the bullgear shaft), and indicate and trip on overall vibration only. Bullgear vibration must transfer to a rotor shaft to be picked up by the prox probes. To determine the source of the high vibration trip, the offending shaft had to be identified.
All prox probes were monitored simultaneously with an SD-390 multi-channel instrument in real-time during a compressor run up. This was done to identify the dominant amplitude and frequency at the time of trip, and thus the offending shaft. At startup, amplitudes were noted to be slightly above normal and the spectrums contained some trash, but were not thought significant. After it was warmed up and stable, the operator attempted to load the compressor, which again tripped on high vibration. But, it was noted that the dominant amplitude at the time of trip was on the High Speed Rotor prox, and was at 3600 RPM or motor turning speed. Shafts turning at this speed are; oil pump (direct coupled on the motor free end), motor, spool piece/couplings, and bull gear. The spectrum at the time of the trip was not captured due to an error with the real-time analyzer.
The obvious place to start the inspection was the spool piece and so it was pulled. It was noted that an abnormally small amount of grease was found on the teeth but they were definitely lubricated. A normal small load or contact patch was observed in the tooth center of the male and female gear teeth on the compressor end as expected. However, on the motor end less grease was found and the contact patch was found to be on both the load and clearance side of the teeth. Clearance measurements were taken and found to be .0025" on the compressor end and .0055" on the motor end. Though the motor end is near the manufacturer's spec for maximum clearance, it was felt to still be serviceable. Couplings were reassembled and the gear teeth packed with the normal quantity of grease lubricant. Concentricity of the coupling halves was minimized with a dial indicator, as the plants standard procedure requires. TIR obtained were .004" on the compressor, and .003" on the motor end, or approximately what they were prior to disassembly. These values are a function of the coupling halves machining tolerances, shoulder concentricity, etc.
Again, the compressor was test run while observing all prox probes in real-time. All was found to be normal. The compressor was loaded up and was put back into service.
Figure #2 shows data recorded on the HS Rotor prox probe, with the plants walk around box, during the trouble shooting process. The top spectrum shows the third stage with the problem on June 9. (Note the random energy present). The bottom spectrum is after corrected, and loaded up on the 11th. It was determined that the minimal grease originally present in the coupling could not carry the 2,250 HP load on the meshing coupling teeth and didn't provide enough dampening. This caused a torsional motion, (causing frequency modulation), and allowed the coupling teeth to in effect chatter producing the unusual contact patch and sidebands.
This machine uses all sleeve type bearings. The bullgear is located axially by a fixed bearing. The motor floats and seeks magnetic center. They are connected by a ~42" spool piece/couplings that has both radial and axial clearance. The spool piece axial clearance or float is set by machined buttons in the center of the couplings that limit the motor axial movement.
In rare circumstances, all the axial clearance will end up on one end of the compressor. This is normally due to a sudden change in operating conditions or compressor load. This can hold the motor off magnetic center and cause high machine vibration. To correct this condition the couplings need to be floated with a pry-bar. With the machine stopped, the spool/piece is moved to its axial limits of travel, first one direction and then the other. It is then positioned in the approximate center of the axial clearance. If excess grease has been packed in these couplings, a hydraulic lock can form and prevent this centering of the spool/piece.
It was determined that during the previous maintenance period, the couplings had been repacked with less than the normal amount of grease. This was done in an attempt to insure that this hydraulic lock could not happen. It proved to be enough for the initial run period but centrifugal force and load had caused the grease to spread out within the coupling until it was too thin at the contact patch. This is why the data collected on the 9th had the abnormal energy in it. There was not enough present on the teeth to support the loads when it was restarted after the current maintenance period's work was completed.
Figure #3 is the same measurement points shown in Figure #1 but after the spool piece/couplings were corrected. No other changes were made to the machine than what is described here. The compressor was loaded and on line when this data was taken. Note: The cursor is marking 3600 RPM and the Y full-scale is the same as in Figure #1, also note that the spectrums are clean of the extra energy of the 9th and have returned to normal condition.
Article submitted by, Dave Williamson, Industrial Reliability2, firstname.lastname@example.org (816) 673-6209