Chevron’s vibration-monitoring crew was called to assess the situation. The company has a sophisticated predictive-maintenance program known as Integrated Machinery Inspection (IMI) and a group to administer it. On its way to greatly reducing maintenance cost, the group has amassed considerable experience in diagnosing noise and vibration problems.
According to Chevron’s IMI Supervisor John Lizarraga, many times the solution is found by elimination. “The complex dynamics that result from the interaction of all the moving parts in the machinery make finding the answer difficult,” he says.
Still, Mr. Lizarraga and his crew checked a number of places made obvious by experience. First they wanted to eliminate the possibility of bearing damage that could have occurred during motor shipment or installation. Mr. Lizarraga explains that sometimes bearings can be damaged in transit if the motor isn’t packed properly. The damage is called brinelling, produced by the balls striking the bearing raceway repeatedly in the same spots. Brinelling can also occur during installation if a bearing is hammered onto a shaft.
Mr. Lizarraga gathered time and frequency data and analyzed it with a Scientific Atlanta SD 385 analyzer. Tests were inconclusive, but the frequency of the troubling noise pointed to one of two things: a bearing problem or an antirotation ratchet. (Some pump motors include a mechanism to prevent impeller reversal. It has been known to cause vibrations in some instances.) At this point, Chevron contacted pump manufacturer Gould. Unfortunately, it too, was at a loss to explain the noise. The refinery then turned to the motor manufacturer, GE Motors.
GE Sales Engineer Frank Judy inspected the pump and motor installation and noticed a shroud installed to protect the motor from moisture. His first recommendation: Eliminate the possibility of sympathetic resonance in the shroud. “Sometimes the inherent vibration of the machinery will excite a natural frequency in the shroud,” Mr. Judy says, “so we had our service shop weld a gusset to the shroud. The idea is to change the stiffness of the shroud so its natural frequency is altered.” When this procedure was done, though, the sound persisted.
Mr. Judy decided to call on GE Motors’ sound laboratory, headquartered in Ft. Wayne, Ind. I responded. The central problem in these matters is to determine if the noise is due to a defect or if it is simply the machinery’s response to a harmless force. We had eliminated the latter with the gusset experiment.
Tool price is right
Our team had been solving noise and vibration problems for decades, using an array of test equipment, including an anechoic chamber. The difficulty: The troublesome noise was 2,000 miles from the lab. Previous experience with these situations, though, gave us a quick and inexpensive tool — a $70 Radio Shack tape recorder. The analog data from a little tape recorder can tell everything we needed about the problem — no need to equip our sales engineers with expensive analyzers.
The analog recording captures all the data, including the time domain and frequency domain of the noise or vibration in question. This is an advantage over readings of frequencies alone because the analysis can be done in three dimensions: frequencies and their amplitudes plotted over time. Viewing the noise over time was especially important in this situation because the noise persisted even as the motor coasted down when de-energized. This is a good way to determine if the noise is due to a mechanical source like a bearing, or a sympathetic response like the shroud.
I directed Mr. Judy to tape the pump and motor in several operating modes. First, he recorded the motor running detached from the pump, which allowed evaluation of the motor sound separately. Next Mr. Judy recorded the motor and pump running at full load to get a perspective on how the two interacted. And finally, readings were taken while the motor was slowing when de-energized. He made all three measurements by walking the microphone a full 360 deg around the motor.
Continue on page 2
Mr. Judy sent the recordings overnight to the lab in Ft. Wayne, where we fed them into an SD385 analyzer. Preliminary results pointed to the thrust bearing of the vertical-drive motor as the culprit. The spectral readings revealed a high amplitude at 620 Hz, three times the ballpass frequency of the outer race (3BPFO) of the thrust bearing. Ball-pass frequency is the number of balls passing a given point (specifically, a defect) per unit time. The level, especially the harmonic content, dropped substantially with the motor uncoupled from the pump, which removed the pump load from the thrust bearing, Figure 1. This also explains how the bearing, and later the complete motor, could slip through manufacturing tests.
Typically, these circumstances would dictate thrust-bearing replacement. However, if we changed out the bearing and the noise went away, we’d never be certain of the cause. It is important to learn from field issues, thus preventing problems and reducing service costs.
The fact that this reading occurred at the third harmonic — and not at the fundamental frequency or first or second harmonic of the ball pass — indicated one of two possibilities:
• A sympathetic resonance at 620 Hz with the fundamental present at the source but not amplified by the structure.
• The fundamental is really at 3BPFO.
Here is where the taped data are valuable. Using a log-amplitude scale, we tracked the 3BPFO signal and its harmonics, dropping in frequency as the motor coasted down unenergized and disconnected from the pump, Figure 2. This confirms the presence of high bearing force, including some impulsiveness at 3BPFO. (Impulsiveness is a term given to an errant frequency and attendant harmonics. This vibration results from a sudden, repetitive force acting on an irregular surface. It is similar to “washboarding” of a road.) The trouble was, we still couldn’t account for the cause of the noise. We consulted with Test Engineer Tim Dickens at the bearing supplier, SKF USA Inc., King of Prussia, Pa.
Mr. Dickens analyzed the time and frequency plots we sent. Various bearing operating characteristics can be extracted from the vibration signal, although most of the time other noises in the applications mask bearing vibrations. Hundreds of possible frequencies, specific to bearing design and shaft speed, were calculated to determine the source of the 620- Hz signal, including frequencies that denote audible noises and do not influence bearing life like component waviness or grease noises. From these calculations, Mr. Dickens predicted that 47 or 49 slightly sawtooth (harmonic of 3BPFO) waves about the outer bearing race would account for the fundamental 620-Hz tone.
No machined part is free of waviness; it is a matter of degree. Amplitudes of a few nanometers can cause audible noises, depending on localized resonances (amplifiers of specific frequencies) within the application. Mr. Dickens stated that bearing waviness is imperceptible to the eye and does not normally shorten bearing life. In this application, though, he believed that, because of the severity of the noise, this bearing was abnormal and should be replaced if only to eliminate the annoying noise.
To confirm that waviness was causing the noticeable harmonics, the motor was taken to the local GE service shop. There, it was tested by installing a new thrust bearing and comparing bearing performance to that of the bearing installed at the factory. No-load vibration readings were taken again with the tape recorder and sent to Ft. Wayne for analysis.
The coast-down test, before the bearing change, confirmed again the significant amplitude at 3BPFO, along with some noticeable harmonics. After bearing replacement, this tone content and its attendant harmonics disappeared, Figure 3. The motor was subsequently returned to Chevron and reinstalled. Chevron reports the motor is running quietly.
Small waves, big ripples
With the customer satisfied, GE Motors and SKF took an in-depth look at the audible noise. To begin their investigation, we sent our test results and the two angular-contact bearings that made up the upper thrust bearing to SKF. There, Mr. Dickens disassembled the bearings and, upon examining them under a microscope, found no sign of abnormal bearing wear. The SKF Altoona plant did tracings on the two bearing rings. On one of the two, the trace indicted 47 waves on the outer ring as Mr. Dickens had predicted. The waves measured 5 microns peak to valley.
Continue on page 3
“We didn’t feel that the impulsiveness in these waves would significantly affect the life of the bearings,” says Mr. Dickens. “Still, we’re always greatly concerned when our customer has a problem with a bearing, especially a high-frequency whine like Chevron experienced.” According to Mr. Dickens, SKF discovered that the combination of its tooling setup and the number of balls in the bearing had caused the noise problem. The chance wave pattern of 16 balls and near multiple 47 (almost 16 × 3) produced an amplification that made the surface sound worse than it was. A modified finishing step has been installed on SKF’s line to alleviate the problem.
SKF monitors component waviness for every bearing produced as well as assembled vibration quality of bearings for electric-motor applications. “But,” cautions Mr. Dickens, “each application amplifies different combinations of component waviness depending on load, speed, and shaft and housing fits as well as the general structure of the application in which the bearing is assembled. The benefit of cases like this is that we can improve our processes to avoid audible noise.”
GE Motors is also taking steps to improve its processes from this and other experiences. It has accumulated a library of information over the years from problems like Chevron’s, compiling a database on vibration analysis of motors and motor applications. That database will be incorporated into a diagnostics program to be used in factory production testing. GE will test each motor coming down the line. This vibration-related production process joins another program GE Motors initiated last year to lessen inherent motor vibration. Each rotor in the Energy $aver X$D line is balanced to within 0.055 ips, well below the NEMA standard.
Vibration is an important issue, especially in the process industries. With the eventual diagnostic database, the motor manufacturer can test motors according to the specifications of a customer in a particular industry such as automotive, chemical, and pulp and paper industries.
Waviness in the bearing of the Chevron motor would have been discovered and diagnosed with on-line testing. Thus, not only will potential problems be caught before they leave the factory, the factory will also know why and can share that information with suppliers like SKF.
Kerry Shelton is Senior Sound Engineer, GE Motors, Ft. Wayne, Ind.