SPECTRUM ANALYSIS OF MACHINERY VIBRATION PROBLEMS

Diagnosing an impending failure in rotating machinery by spectrum analysis is the process of measuring and extracting vital information from vibrational forces exhibited by the machine, identifying the applicable excitation source and making an accurate assessment of the problem. Proper analysis provides a sound basis for decisions such as when to shut the machine down and when to schedule the repair. Correcting the problem is usually quite straightforward once the cause has been identified.

The use of narrowband spectrum (signature) analysis has proven to be a very powerful and expedient tool for solving machinery vibration problems. Amplitude vs. frequency spectra provides a clear presentation of vibrational forcing frequencies and the energy level (amplitude) of each. To make an efficient and accurate interpretation of mechanical signatures from operational systems, it is imperative that the processing and analysis of this spectral data make use of all information influencing these signatures. Two primary influences, separate from dynamic forcing frequencies generated within the machine are the characteristics of the surrounding structural system, and operating speeds and conditions.

The key to success in diagnosing vibration problems associated with rotating machinery is to thoroughly understand the machine dynamics and to record accurate and meaningful vibration test data. The following sequential procedure outlines a practical approach for performing a spectrum analysis on an operating machine.

1) Verify that the vibration problem is caused by mechanical conditions of the machine as opposed to abnormal operational (process) conditions. If operational conditions are normal, investigate the circumstances leading up to the problem:

a) Review all available vibration test data recorded prior to the occurrence of the problem for indications as to whether the vibration levels increased gradually or suddenly.

b) Inquire about any changes in performance, including operating speed, that may have taken place just prior to, during or after the problem occurred.

c) Review machine maintenance records for signs of recent repairs and frequent or repeating failures.

2) Visually inspect the machine for obvious faults and solicit from operating personnel information that might lead to a better understanding of the problem. Become acquainted with design parameters of the machine; such as, type of bearings (rolling element or film), gear types, speeds and number of teeth on each, number of vanes, blades and volutes, etc. Knowing the rotational speed and design characteristics of the machine will provide insight as to what forcing functions to expect and at what frequencies they should occur.

3) View the spectrum on the display in the instantaneous mode of operation. This will assure that sufficient frequency range and resolution were selected to satisfy testing requirements. Initial test considerations include:

f) Averaging type and number (generally, summation frequency averaging is used. But, if synchronous events are of importance, or if your attempting to isolate synchronous events from non-synchronous synchronous events, time averaging or synchronous time averaging should be used);

g) Are phase relationships important to your diagnosis? If so, you may want to process both frequency and synchronous time averaged data.

h) Are there any frequency or amplitude modulations, which might require zoom processing. If so, it might be beneficial to pre-set cursor locations for easy processing. Alternatively, maybe "zoom averaging" is a more effective means since re-running data can be a problem.

4) Select the appropriate average display and process your data accordingly. In cases where higher frequency information is important, several ranges may be processed to improve frequency resolution.

5) Record meaningful spectra by appropriately mounting your transducer. Avoid weak, flimsy mounting surfaces, as they tend to resonate and distort the actual measurement.

6) Record (hard copy or mass storage) "preliminary " unbiased data at all available measurement locations. Never assume the diagnosis is " a piece of cake" after recording just a few spectra. Many faults exhibit vibration characteristics that can easily be misconstrued for other phenomena. Remember your isolation tests.

7) Review your data carefully, develop a list of probable causes for the spectral characteristics displayed. In some cases, it may be easier to develop a list of problems that "do not " match your preliminary test results. Work with which ever is easiest for you. Then isolate your suspicions with appropriate isolation testing techniques. Document your isolation test results even if they do not corroborate our initial diagnosis or "gut" feelings.

8) Believe your test data. If you have skillfully executed the test, then the data will lead you to further isolation tests and/or ultimately to the proper diagnosis. Do not be swayed by others outside your test group if you feel confident with your results. If you have reservations, implement additional confirmation tests to resolve the discrepancies.

9) Once you feel confident with the diagnosis, prepare a list of recommended corrective actions. Always prioritize them in the event that management steps in and elects to only make critical repairs. If you are unsure of the exact cause of the problem, say so. But also, prepare a recommended plan of action to pin down the source.

10) After recommendations are implemented, re-evaluate the machine to confirm that the corrections have been successful. This data can then be considered "new baseline" data for future comparisons.

11) Having gone full circle now, it is up to you to maintain empirical data on your success and failures to enhance future diagnostic endeavors.

d) Review all diagnostic results, recommendations and the outcome of all problems with others in your group. This will insure a "pooling" of the minds as well as provide cross training on other equipment, which ultimately will lead to an intimate understanding of that particular machine.

Mechanical signatures of unbalanced rotors will exhibit a predominant spike at operating speed. Unbalance is classified as static or dynamic.

STATIC unbalance is affected only by gravitational forces. If an unbalanced rotor is placed in frictionless rollers, it will turn until its heavy spot is at the bottom. If the rotor is statically balanced it can be set at any angle and it will remain at that position, since there is no heavy spot.

DYNAMIC FORCE unbalance is similar to static unbalance, but appears during rotation. Typically, force unbalance produces similar in-phase forces on two opposite support bearings, when measured in a common direction. In force unbalance the axis of rotation is displaced but remains parallel to the axis of the center of mass of the rotor.

Pure COUPLE unbalance is caused by two equal but out-of phase forces on the two bearings. Couple unbalance produces a kind of "rocking" motion. In couple unbalance the axis of rotation is not parallel to the axis of the center of mass of the rotor, but intersects it at the midpoint. In a combination of force and couple unbalance, the axis is not parallel to the axis of the center of mass and intersects it at some point, which is not the midpoint. The force/couple balancing solution assumes that the dynamic unbalance consists of either a force (in-phase) or couple (out-of-phase) unbalance or a combination of the two. It is assumed furthermore that each can be solved independently, since one does not affect the other. Other problems, which occur at operating speed, include bent shafts, eccentric rotors (electrical – 3600cpm) loose rotor parts, resonance, etc. These problems have been listed below along with isolation techniques for each.

LOOSE ROTOR PARTS - starting and stopping the machine will normally result in different amplitudes and/or phases. A peak-averaged signature on run-down may drop in amplitude suddenly due to the loose part re-seating itself.

BENT SHAFT - phase data will indicate if the shaft is bent. A bend between the bearings will exhibit 180 out-of-phase axial readings. A bent shafting at the coupling will appear as misalignment. Always confirm your suspicions of a bent shaft with a dial indicator.

ECCENTRIC MOTOR ROTOR - a 2-pole synchronous motor will exhibit energy at 60hz or 3600 cpm, (1x-line frequency). Because the rotor is not concentric, it cuts the electrical field unevenly and therefore manifests itself as mechanical vibrations characteristic of imbalance. Use of "zoom" will usually provide sufficient resolution to separate the mechanical and electrical constituents. Synchronous time averaging or discontinuing power to the motor will also confirm this type of problem.

RESONANCE - any vibration due to unbalance can increase if the rotor is operating at or near (+ 20%) a critical frequency. Slight changes in rpm can yield drastic changes (increase or decrease) in the vibration level. A “peak” average or waterfall display is excellent for obtaining a permanent record of the resonance responses with speed changes.

LOADING - the process can cause drastic changes in the vibration level at operating frequency. A temporary change in the loading will usually confirm this type of problem. The quickest method for isolating unbalance on machinery is to perform a balance shot. It will either work or not work. If it doesn't, then one of the above causes should be investigated.

Misalignment is characterized by a predominant second harmonic (2x rotational frequency).

Vibration due to misalignment remains unchanged in amplitude with an increase in operating speed. The vibration amplitude will not fluctuate with the unit operating at a constant speed.

The spectral relationships of the vibration amplitudes at running speed, and the second harmonic, are directly related to the length of the coupling span and the stiffness of the inboard bearing supports (i.e. a long narrow coupling span will exhibit a higher first order vibration amplitude than the second harmonic). Vibration amplitudes at the second harmonic cannot be treated equally with the first order amplitudes. As a rule of thumb, the velocity level of the second harmonic is multiplied by a factor of 5. The resultant should be within acceptable vibration limits of 5. The resultant should be within acceptable vibration limits for that particular frequency or, a misalignment is indicated. Relative phase relationships across the coupling in either axis will be significantly out-of-phase in the presence of misalignment. The predominant axis, which indicates misalignment, cannot be stated as due to the fact that cross-axis transmissibility is different from machine to machine. Different types of misalignment do cause different signatures and phase relationships. Listed below, the types of misalignment with the predominant axis and phase listed.

All phase readings must be carefully recorded in regards to pick-up reversal. Machinery with natural frequencies near the first harmonic may indicate misalignment problems even if the unit is properly aligned. A precision alignment may reduce this problem. If a resonance is suspected, perform an impact frequency response test.

Most electrically induced problems will only appear with the unit loaded. Obviously, the increase current draw will cause greater excitation of electrical problems. Fluctuating amplitude of a harmonic at 3600cpm and/or 7200cpm is caused by an electrical and mechanical frequency going in and out of phase and subsequently beating in amplitude. Greater frequency resolution may be necessary to isolate the electrical frequency from the mechanical component. Synchronous time averaging will also discount electrically related frequencies from synchronous shaft vibrations.

All electrical problems can be isolated by discontinuing power to the unit and observing the 3600 and 7200 peaks. If they are electrically induced, the amplitude will immediately drop off to a level attributable to any residual mechanical problems. Electrical problems usually do not cause unpredictable failures, so equipment can continue in operation until scheduling permits repair. Such problems must, however, be evaluated from a standpoint of vibration severity, current draw, bearing and winding temperatures, etc.

Eccentric rotor armature

Loose status laminations

Broken rotor bars

Unbalanced coil or phase resistance

Heating

Shorts

Loose iron

Bearings can be divided into two major groups for diagnostic purposes… anti-friction and oil film.

Many articles have been written on the calculation of frequencies generated by anti-friction bearings. These frequencies have been faithfully reproduced under laboratory conditions and also on machinery with low process related vibrations (motors, generators, etc.).

Our experience has proven that most process machinery (pumps, compressors, grinders) produce such high levels of baseline energy that these bearing-related frequencies are masked. In addition, the availability of bearing geometry from manufacturers is limited. Therefore, the calculations are rarely of immediate help.

A bearing, which has no factory defects and is installed properly, can be successfully trended by recording signatures periodically. The frequency range will vary with bearing size and speed; however, an envelope should develop attributable to excitation of the bearings components' natural frequencies, which will normally appear in the 30k to 60k cpm frequency range. Close inspection will show multiples of the calculated bearing fault frequencies superimposed.

As a bearing approaches failure, this energy envelope will reduce in amplitude but broaden in frequency. Extreme bearing looseness will be indicated by a high running speed component followed by multiple harmonics.

Diagnostics of oil film bearings can be accomplished from bearing cap readings, however, proximity probes reading relative shaft vibration will be more accurate, due to the attenuation of the vibration signal through the oil film.

When a sleeve bearing starts to wear, the oil film characteristics can change and may cause rotor instability.

Unmonitored sleeve bearings are not usually detected as faulty until high bearing cap vibration is noticed. At this point, the signature will typically indicate a high rotational frequency with harmonics superimposed on an elevated baseline.

A new sleeve bearing should always be analyzed to insure proper alignment and clearances. The above mentioned instability problems may already exist, and may be manifested as oil whirl, whip or rotor rub by sub-rotational frequencies.

Mechanical looseness between machinery and its supporting structure, looseness of impellers, bearing housings, bearing inserts and couplings can create unwanted vibration which at times can be quite difficult to analyze. The reason for this difficulty is that the looseness characteristics can easily be mistaken for imbalance and alignment problems. However, through the use of case histories presented, we will establish a format for evaluating looseness problems, which distinguish them from other sources of vibration.

General characteristics of any type looseness condition are manifested as broad-band energy spanning a frequency range from one to ten times rotational speed. In addition, the rotational frequency and its harmonics will be superimposed upon the envelope. During instantaneous viewing, the baseline will be very unsteady and constantly modulating in both frequency and amplitude.

Mechanical looseness between bearings and their pedestals, or between pedestals and its supporting structure (floor, etc.) will be manifested primarily by a prominent 1x and occasionally 2x and 3x rotational speed. The latter usually occurs when the loose component has sufficient energy and free play to allow it to strike its restraining tie downs, at both extremes of excursion. For example, a working fan pedestal may strike first the supporting floor and then at its maximum excursion, striking the loose bolt head restraining it.

The loose component can be isolated by measuring relative levels between the bearing housing and its support, the support to the base of the foundation, etc. Any difference in readings that are greater than 75% indicates a loose condition. Water or oil oozing between a metal base and the concrete foundation is a sure indication of looseness.

Looseness of any rotating component (impeller, etc.) will be manifested at the rotational speed and appear as imbalance. This occurs because under operation, centrifugal force will hold the impeller in one position, which will be offset from its centerline by the amount of the looseness. Therefore, the resulting vibration can easily be misdiagnosed as an imbalance.

A simple test would be to stop and start the unit several times and record both the phase and amplitude with each start-up. If variation of either parameter is noticed, you can assume looseness does exist. Also, increasing vibration levels with warm-up is indicative of looseness being affected by thermal growth changes or loading.

Operational problems, or more accurately, vibrational effects of operational problems can vary widely in spectral make-up dependent upon the process and its supporting machinery. A system' s operating condition can account for some basic vibrational phenomena such as vane pass frequencies, pump cavities and on fans, starvation. They also can be directly related to many more complex problems such as super harmonic instabilities. All of the above phenomena are capable of exhibiting extremely high energy levels and must be avoided by proper operation of the equipment.

The most commonly seen operational problems are attributed to "vane pass" frequencies (known as " blade-pass" in fans). This phenomenon occurs in hydraulic (pumping) or aerodynamic (fans) systems. It is characterized by a prominent multiple of rotational speed equivalent to the number of impeller vanes times rotational frequency. E.G. Vane pass = # vanes x rpm

Often times many harmonics of rotational speed are also encountered, coupled with broad energy along the baseline. This broadband energy will appear as a modulating baseline in instantaneous viewing. This modulation is indicative of turbulence and in some cases may appear as "white noise."

Basically, "vane pass frequency" is just as its name implies. It is simply a pressure variation or pulse due to the sudden unloading or loading of the vane as it passes the discharge volute.

Quite often both vane pass frequencies and the resulting turbulent responses, are attributable to improper discharge restrictions, such as valve settings.

Cavitation is generally described as the sudden implosion of air bubbles suspended within a liquid. This implosion occurs due to compression of the liquid during the pumping cycle.

A common source of cavitation is when an insufficient pumping media is available to meet established pumping capacities. The resulting spectral characteristics consist of very broad-band (modulating) energy often as high as 2000 Hz in frequency. This energy can either appear as "white noise" with no discernible frequency content, or with vane-pass frequencies superimposed. Visual inspection of impellers exposed to cavitation reveal a pitted surface often times on both sides of the vane due to recirculation. As a result of this impeller deterioration, imbalance also sets in as a major vibratory contributor.

The aerodynamic counterpart to cavitation is simply termed starvation. Similar to cavitation, insufficient airflow relative to fan capacity is the primary source of this type problem. Again, damper settings and in some cases, miss-application of the equipment are major contributors. Spectral characteristics are identical to those of cavitation. Broadband energy (and in some cases violent vibration) will be present in suction and discharge ducts due to the extremely turbulent flow.

Variation of operating conditions from the norm can produce drastic vibration in the form of super harmonic vibration. This phenomena, although somewhat rare, manifests itself as an extremely violent vibration that can appear suddenly and apparently without good cause. Our experience has found this problem to occur primarily with direct drive overhung rotors operating near resonance.

As might be expected, the basic problem is the excitation of a system's instability, however this instability is excited via operating parameters that can alter loading characteristics between the rotor and its supporting bearings. The resulting vibration is indicative of a "whirling" rotor with the fundamental vibration occurring at the rotational speed of the rotor. This problem can temporarily be alleviated by reducing the rotor speed slightly (until the vibration reduces). Once this has been accomplished, the speed can be re-established usually without incident.

Belt driven machines can exhibit frequencies inconsistent with harmonics or sub-harmonics of fundamental machine orders. The spectral make-up of drive belt problems is characterized by several harmonics of the belt rotational speed. The belt frequency can easily be calculated using the formula below:

*Use the same sheave (either driver or driven) for pitch diameter and speed inputs for the formula.

Belt frequency energies will be modulating in amplitude. The predominant axis of vibration will be radically in the direction of belt tension. Common sources of drive belt vibration are an unmatched set of belts (or stretched set of belts), adjustable sheave, eccentric and/or unbalanced sheaves, belt misalignment and drive belt resonance. An unmatched set of belts in addition to belt frequencies will exhibit a large 1x rotational speed of either the drive or driven units. The largest vibration amplitude will generally be measured on the driver.

The major contributor to belt vibration is the application involving adjustable sheaves. These applications, common with new installations, create undue vibration and premature drive belt and sheave deterioration. Adjustable sheaves are generally installed for air-balancing requirements however, once the airflow requirements are met and the proper speeds established, they should be replaced with a fixed sheave of equal pitch diameter.

The problem with adjustable sheaves is that the sheave faces are not parallel with one another. This allows the belt to ride up and down in the groove with each revolution. This, in turn, creates a constant variation in belt tension and provides the necessary forces to be manifested as vibration.

Eccentric and/or unbalanced sheaves manifest themselves as a simple imbalance. The fundamental will occur at the rotational speed of the shaft with the eccentric or unbalanced sheave. Field correction using small washers (under the mounting bolts) or putty can often alleviate this condition, however eccentricity, although not creating excessive vibration, will still propagate belt frequency energy due to belt length variations and result in wear of belts and/or sheaves.

Drive belt (or sheave face) misalignment can also create unwanted vibration. This is more commonly seen in short length high-speed belt applications. Predominant vibration occurs at rotational speeds of the driver and occasionally the driven units, and is manifested as an axial vibration. Sheave face alignment can be checked with a straight edge or piece of string.

Drive belts resonance manifests itself as a "flapping" belt (especially on the tension side) at a frequency, which correlates with its (drive belt) resonant frequency. It should be noted that for a resonant excitation to exist, a forcing frequency such as a belt rotational frequency (1x, 2x, 3x, etc.) must be within the resonant frequency response range. So be on the lookout for this resonance to coincide with and (and maybe misconstrued) as a belt frequency. To determine the drive belt resonance, just pull a belt down and quickly release it. This will allow the belt to vibrate at its natural frequency. While the belt is vibrating, a commercial strobe light can be used in the oscillate mode to "freeze" the vibrating belt. The frequency at which you freeze the belt is the belt's resonant frequency. Several excitations of the belts may be necessary to complete the "freezing" of the belts. If you determine that drive belt resonance is a problem, adjusting either the belt tension or changing the belt length will be necessitated.

Mechanical resonance is a condition that exists when a frequency of an excitation force coincides with a natural frequency of a mechanical system. The elements, which describe this natural frequency, are mass, stiffness and damping, any combination of which has a unique natural (or resonant) frequency. A component or system resonant frequency is primarily dependent upon the combined mass stiffness ratio. Increasing mass lowers, the natural frequency while increasing stiffness will increase will increase the natural frequency. Damping has a minimal effect in determining a natural frequency but provides resistance to motion. The quality of damping is determined by the ratio of peak amplitude at resonance to the amplitude away from resonance and is defined as the amplification factor or Q factor. A low Q factor indicates a large amount of damping. Rotating machines possess several natural frequencies that may or may not be excited, but operate satisfactorily because the response due to resonance is restrained by damping.

Excitation of a component or a group of components as a system at its natural frequency can often be a problem with rotating machinery. When a component such as a blade, impeller or bearing support structure is excited at its resonant frequency, fatigue cracking can lead to catastrophic failure. A response due to resonance will depend upon how well it coincides with the exception force in frequency and in the amount of damping in the structure. Resonance can be excited by a number of forces such as imbalance in the rotating mass, two times rotational speed due to misalignment or bent shaft, any harmonic or sub harmonic of running speed, gear mesh, blade passing, load surges and sometimes through the machinery foundation by a remote source. Torsional resonance in shafting is especially susceptible to excitation by forces due to gear mesh, blade passing and fluctuating axial product loading.

When a system or component resonance is affected (excited) and is an apparent contributor to the vibration problem, it will be readily detectable by viewing the CRT while monitoring the vibration with a spectrum analyzer in the instantaneous mode. Adjust the frequency range on the analyzer to a minimum range that will include all predominant energy for optimum resolution. The energy envelope affected by a resonance, will be somewhat broader in bandwidth at its base than, for an example, a clean energy spike such as at running speed due strictly to residual imbalance in the rotor. This broadband energy at the base of the peak will exhibit modulating amplitude resembling the characteristics of mechanical flexing.

A broadband spike due to an excited resonance can appear at any frequency within an averaged spectra depending upon structural characteristics of the machine. The worst case occurs when the resonant frequency coincides exactly with a predominant forcing frequency. The energy amplitude will be the product of the forcing function at this frequency times the magnification factor due to resonance. The energy spike will usually appear clean (sharp or narrow band) except near the baseline where the bandwidth broadens abruptly due to the damping effects. When the forcing frequency is slightly lower or higher than the resonant frequency, the broadening of the energy bandwidth near the baseline will shift accordingly.

Frequently, the predominant vibratory energy at a machine running speed is the result of some imbalance in the rotating mass that is being magnified due to a resonance at or near running speed. Attempting to trim balance the rotor when it is aggravating a resonance is usually very difficult and sometimes unsuccessful depending upon the Q factor and amount of rotor imbalance. Shaft phase readings may be erratic and unstable because there is always a 180-phase shift across the resonance response frequency and usually small amounts of corrective weights will affect large amplitude changes. In most cases, some reduction in vibration amplitude can be achieved. If balancing the rotor fails to reduce the vibration level sufficiently, isolate and modify the resonant component to shift the resonant frequency away from the running speed.

One method for verifying whether a predominant forcing energy is exciting a resonance is to vary the running speed of the machine. A change in machine speed will vary the forcing function frequencies but will not affect the self-exciting or resonant frequencies. In the case of blade passing energies causing excitation of a resonance within the system, verification can usually be made by changing operating conditions to reduce the excitation forces.

A technique utilizing a spectrum analyzer that has proved quite successful is to record a peak hold (continuous peak averaging or a stack plot presentation of a 10k point input buffer) spectra during machine run up or coast down. Usually it is best to perform this test during coast down because most machines change specs less rapidly therefore, any resonance is more readily excited. If forcing frequencies pass through a resonant frequency too fast, the resonance will not respond. When possible, perform the test by slowly increasing machine speed from start to maximum. As the fundamental forcing frequency and harmonics pass through the resonant frequency, sudden increases in energy amplitude will be viewed on the spectrum analyzer CRT. Then the amplitude will decrease after the forcing frequency has passed through the resonant frequency.

Another method for confirming the presence of resonance and determine its exact frequency is to attach a small transducer such as an accelerometer (must be light enough that it will not affect the component's response) and strike the component with a soft object such as a block of wood, plastic hammer or lead mallet. The resulting response when viewed by a real time spectrum analyzer will occur at the natural frequencies of the component. Extreme care must be taken to make sure that the shock impulse (forcing function) is adequate, both in frequency range and amplitude to excite the measured component at its natural frequency and that the measured response is not simply energy transmitted directly from the shock impulse.

A great deal of confidence in the test results can be achieved by averaging several transient responses while making sure there is repeatability in both the shock pulse and frequency response. The bandwidth of the resonant response envelope indicates the amount of damping present in the component – the wider the bandwidth, the more the damping. This method of shock testing has been used successfully on a wide variety of structures and components for locating critical resonance. This procedure is normally utilized when a machine is shut down; however, in many instances where it was not feasible to stop a machine, this technique has been used while the machine was in normal operating conditions and satisfactory results were achieved.

In some instances, it may be necessary, or at least desirable to perform a resonant search on a machine's structure prior to start-up. This resonance search is performed by attaching to the structure a controlled excitation force, normally a sinusoidal vibration exciter that is tunable over the entire applicable frequency range. The response of the system can readily be mapped by attaching one or more accelerometers, and monitoring the response vs. the input excitation force. Performing this structural response or resonance search on all equipment prior to start-up is highly recommended because normal operating conditions of all machinery is related directly to the forcing functions and system response. Calculation of the normal forcing function frequencies and relating them to known system resonant response frequencies will provide the desired knowledge of how a machine will perform under normal operating conditions.

Rotor instability occurs when a bearing is unable to exert sufficient preload to hold a rotating shaft in a stable position. When a rotor bearing system is prone to becoming unstable, any outside force, which acts to upset the bearing load, may provide the condition necessary to cause instability. Instability is a condition of rotor operation in which various elements in the system combine to induce self-excited vibrations that can remain even after the original stimulus has been removed. Only a significant reduction in the rotational frequency can stabilize instability. Once stabilized, the rotor can usually be returned to its normal operating speed and will remain stable at least until some disturbing force upsets the system again. Rotor instability is usually associated with oil film bearings; however, it should be emphasized that even a rolling element machine can be unstable especially due to gyroscopic effects on overhung rotors. Oil whirl in a hydrodynamic (oil film) bearing is probably the most common cause of sub-harmonic instability and it often excites the rotors first critical.

Oil whirl is easily recognized in vibration spectra as a single peak at approximately 40 to 46% of rotor running speed. Sub-rotational instability may also be caused by a metal-to-metal contact (rub) often due to un-lubricated bearing surfaces. Rubs exhibit a half frequency peak with half frequency harmonics as well.

The onset of instability usually appears as irregular amplitude fluctuations at the sub-rotational frequency at 20 to 50% of the amplitude at running frequency. These sub-harmonic energies may appear and diminish suddenly in a sporadic manner or they may stabilize due to restraining forces within the system. Should the system restraining forces become lost or over-powered energies at sub-rotational frequencies will increase significantly and can result in destruction especially in the case of a rub.

Preventing instability in a rotating machine is a basic design criterion that is as important as designing a machine to operate away from its critical (resonant) speed. If a machine is prone to instability while operating under designed conditions, indications are an inadequate bearing and it may be necessary to replace a sleeve bearing with a tilt pad type, for example.

Instability is often experienced with a machine at start-up following a repair such as a sleeve bearing replacement even though the machine has operated satisfactorily in the past. Probable cause may be a shaft/sleeve misalignment preventing the bearing from being properly loaded, bearing clearance or eccentricity in the bearing which will prevent the oil film to form a proper a proper wedge. Verify by plastic gauging both top and bottom halves of the sleeve to make sure that both shaft position and clearance are within specifications.

Rotating machines that include gear drives can become subjected to vibrational forces that are unique to gearing in addition to those normally experienced in other rotating machinery. Vibratory energies generated by gears are usually periodic, relative to rotating elements in the gearbox, except for resonance, which may be excited and can be identified through narrow band spectrum analysis.

Gears usually generate a complex broad vibration spectrum beginning with frequencies below the shaft rotational speed and extending to several multiples of gear mesh frequency (the number of gear teeth X the gear shaft rpm). A typical vibration spectrum generated by gears will include energy at the low end of the spectrum at shaft running speed and running speed harmonics, energy at approximately mid-way between the running speed and the gear mesh frequency (frequently related to system component resonance) and at gear mesh frequency and occasionally its harmonics.

Gear mesh frequency and harmonics are almost always predominant in the vibration spectrum. The energy at gear mesh frequency may be a single peak, but more likely, it will be surrounded by sidebands spaced at intervals of one of the shaft rotational frequencies. Amplitude at the mesh frequency varies significantly from one case to the next depending upon loading and transmission errors from the quality of gear design and manufacturing.

Gear related vibration that occurs at shaft running speed cab be caused by gear mass imbalance, pitch line run out due to manufacturing or assembly errors, shaft run out, variations in tooth support stiffness around the circumference of the gear due to a cracked web, severe shaft coupling misalignment to some gear drives or a serious defect such as a broken or pitted gear tooth. These faults also modulate the gear mesh frequency to generate side bands.

Because mechanical signatures will vary significantly from one gear drive to the next, recording baseline spectra initially on a good gear drive under normal operating conditions will aid immensely in diagnosing an impending failure in the future. These baseline spectra should include plots of expanded frequency ranges, centered about predominate energy frequencies including shaft rotational speed of all gears, gear mesh and those mid-way between. All discreet frequencies should be identified including any sidebands that might be present. Any deviations in subsequently recorded spectra will indicate changes in operational loading and/or degradation in mechanical conditions and will help isolate the cause depending upon where changes take place in the spectra.

Fault isolation requires identification of vibrational energies at frequencies that are usually very close to one another, especially when several gears are involved and because these frequencies are high, they are rapidly dissipated in the structure. This means that it is necessary to use a vibration transducer that has a frequency response range covering all gear mesh frequencies and harmonics. Attach the transducer as close as possible to the bearing supporting each gear and mount the transducer rigidly for high frequency transmissibility.

Condition	Type of Misalignment
Radial 180° out of phase; Axial in phase	Parallel Misalignment
Radial in phase; Axial 180° out of phase	Angular Misalignment
Radial 180° out of phase; Axial 180° out of phase	Combination

Condition

Type of Misalignment