Shown is a typical modern day vibration acquisition system. There are several good ones on the market.
It has certainly changed since I bought one of the first ADRE systems in 1981. That was back in the days of HP-9000 (300) used as controllers for real time analyzers.
Fast Fourier Transform (FFT) is modern technology applying Jean Baptiste Fourier's (1768-1830 AD) formula into electronic format displaying time wave form data into frequency spectra.
This really makes time wave form data from machinery vibrations meaningful and somewhat easy to interpret.
Mr. Ray Dodd is held in esteem as the guy that fathered the modern day predictive maintenance (PdM) programs. In the 1970's we were basically carrying a portable instrument and collecting vibration data via a velocity transducer and recording that data onto a chart with pen in hand. If overall level vibrations above a certain magnitude were encountered, then we would pick out the predominate frequencies and log them as well. A long process, but one that proved its worth.
This writing is to serve as a help and guide to assist in diagnostics of vibration problems. First we'll make sure that certain terms are defined to avoid confusion.
FUNDAMENTAL: This is running speed or 1X; referred to as the machine's fundamental or fundamental component. Also component fundamental is sometimes used; i.e. the bearing's ball pass frequency fundamental is expressed as component fundamental to distinguish it from harmonics in a ball pass problem within the spectra. Always make sure it is defined if used for anything other than running speed.
FIRST HARMONIC: Some confuse this with the symbols 2X and 1X; the first harmonic (primarily an acoustical term) is also referred to as two times RPM (2X). So, be careful. One should visualize this as a bell. The bell is the center of concentric circles - circles or harmonics are numbered from the center. The bell is the center and the first circle encountered is the first harmonic. But in machinery terms, that would be the 2X component. [The bell, itself being the fundamental or 1X component].
BEARING COMPONENT: BPFI (Ball Pass Frequency Inner Race)
BPFO (outer race frequency)
BSF (Ball Spin Frequency)
FTF (Fundamental Train Frequency) cage problem
Here's the so-called classic bearing equation
And then there's the easy way: The rule of thumb method.
There's no substitute for experience; again use acceleration, you can't analyze what you can't see - you can't trend what you can't see or your instrumentation doesn't see. Integration can often drive signals into the dirt or noise floor where they'll be hidden or camouflaged to death. The noise floor masks or covers them up, and a small acceleration signal 0.05 g's 0-P (the beginning of trending) at 600 Hz wouldn't be noticed.
Always check the limitations of your system. Transducer selection for a paper machine should be of very high resolution with as low a frequency response, as temperature will allow. There are a number of good transducers out there. Always make sure your acquisition system is within limits of your interest.
I've seen so-called consultants gather vibration data when the machine of interest was outside of the limitations of their instrumentation. Naturally, the sub sequential report that follows is a real gem. This writing is only a field book and to inform you of certain things that should be of concern to you in your environment. Question your transducer supplier but make certain they have the product you need or find one who does.
Bearing's Natural Frequency: Lack of lubrication excites this frequency and creates a broad vibration pattern around this frequency region casing this phenomenon to display in the vibration spectra.
Special Note: A broadened spike or spectral peak (from the peak on top, the shape of the peak will broaden toward the noise floor) will denote mechanical deterioration. If you see mechanical deterioration and the bearings natural frequency you may want to inject grease lubricant very slowly as not to shock the bearing. If there's only a small magnitude of vibration then you may want to only note it in the report. If it's severe and a determination must be made the addition of lube may give insight as to time of failure or assist in analysis.
LINE FREQUENCY: 60 Hertz (Hz) USA Power Turbine RPM (3600)/60=60. In Europe the power is 50 Hertz so the generators are larger and operate at 3000 RPM.
LINE SYNC FREQ: 120 Hz (120 / # poles X 60 = RPM of a given machine) Look for especially in DC motors. In AC & DC this frequency phenomenon is generally found at upper harmonics of high frequency problems associated with electrical problems or the beginning of potential problems.
The electrical frequency is also displayed in fluorescence light bulbs and is used to calibrate a photo tachometer I've also used it for a trigger.
Field balancing using a tachometer for a trigger device to sync at once per rev.
Overall or Overall level (OA): A term I use often, but not really sure if all are versed in its meaning. Normally an analyzer has 200, 400, 800 and 1600 lines of resolution. Imagine a spreadsheet of 400 columns, as that's normal for most PdM programs. The height is normally 1 Volt with a resolution of 0.005 and automatic scaling. Time waveform data are broken down into frequency spectra. So, each of the 400 windows has amplitude: that amplitude for each window is squared. Overall level (OA) is the square root of the sum of the squares. It is the sixth band in my PdM software. Or in a 200 line analyzer it would be line 201.
New subject: In the US we use g's 0-P but in Europe they use mm rms or g's rms which is totally foreign to us. I did not mention it at all. In addition, CSI boxes used a notation of rms on their box but the output was really 0-P - not rms. It was a flaw in their box, but better not go there.
In this world, people in vibration should realize that integration from g's to IPS (in/sec) isn't absolute but only retaliative. The universal language is IPS but the analytical world is better if kept in tune with the type of transducer one is using. Acceleration - work in g's. Velocity transducer -work in IPS. Eddy current probes - work in mils. But, translate to common terms when reporting.
MEASUREMENT PARAMETERS: Acceleration, Velocity, Displacement and Acoustics (dB).
There are contact type and non-contact type. Which is best? That may be impossible to determine. Large turbo machinery - both. The contact type are cheaper and easier as a rule but are affected by the machines dampening effect caused by bearing type and mass. Mass determines resonance so the sensor's mass should be so that the frequencies of interest are below its natural frequency. The non-contact sensor (eddy current probe) provides rotor dynamics and good down to DC. We are not getting into all systems, laser, all non-contacting as this is only a field reference book.
Bearing cap data are now taken almost exclusively in acceleration. I've used that approach since 1980 exclusively. A velocity transducer is sometimes useful for balancing, apart from that I won't use a velocity transducer and don't integrate an acceleration spectrum into velocity or double integrate to a displacement spectrum. Those integrations are only relative, not absolute. I will and do present digitized data in all three parameters for some. I do not believe in producing a velocity spectrum because it is so limited as compared to an acceleration spectrum. Velocity (IPS) is a good language to translate into words and is a common language as it is irrespective of frequency. So, you can iterate an amplitude and one can understand its magnitude or severity.
A picture is worth a thousand words. You can't analyze what you can't see. Why waste your time with a velocity spectrum when it isn't even a true integration from acceleration to velocity. Integration is relative, not absolute. Work in acceleration and get used to it; it's the right thing to do now and the way of doing in the future. It took some transition period to get from displacement to velocity. Machines fitted with anti-friction bearings should always be monitored in acceleration, and use acceleration spectra for diagnostics: 3600 RPM to 5 kHz. Likewise, if I'm using a velocity transducer or an eddy current probe I monitor IPS and mils respectively - never integrating the spectra only displaying digitized data in all three parameters.
Both displacement and acceleration are respective of frequency, therefore velocity is the language to speak in for clarity and ease of understanding. But in the analysis stage when gathering acceleration data use acceleration for analysis. Discipline yourself and you'll find over time your PdM program will evolve into planned scheduled maintenance and you'll get maximum life from the machine and eliminate overtime. I have successfully implemented such programs. You'll tear into a machine and realize that you have gotten maximum life: no need to pull the magnifying glass out and look for scratches. I've seen people do that - it stands out that it's an excuse for inability or a poor program.
Relationships of sinusoidal velocity, acceleration, & displacement
Microphones (acoustics) are not usually associated with machines except for OSHA sound level measurements.
But, if you have a machine high and away and don't have transducers mounted, then a highly directional microphone can assist for frequency analysis.
This will sometimes produces good results and always is better than no approach or program at all.
Sound isn't our topic although sound is vibration and vibration is sound. We won't be getting into media that has so many variables and dampening factors and background noise etc. But, here's a chart nonetheless.
In a PdM program we normally are concerned with bearing cap data and sometimes with permanently installed transducers. However, one must know that the transducer is in calibration. An outside consultant may choose to use known transducers such as his or her own accelerometers for bearing cap data. Normally, (for example) I will go into a control room and acquire data directly from the panel by plugging in to an existing system such as a Bentley
Nevada 3300; this gives me rotor dynamics. But, I also acquire bearing cap data.
NOTE: A normally good machinery analyses instrument system will cover 20 kHz to 0 Hz or DC and have transducers covering 0.05 Hz to 20 kHz.
They may be phased matched having phase measuring capabilities down 30 RPM within 2%. Phase gathering triggering devices either photo tachs, laser, eddy current probes or strobes. Set transducers side by side to confirm amplitude/phase integrity as a field check.
High quality tachometer may be necessary when phase readings aren't stable using conventional phase gathering instruments.
OUT OF BALANCE CONDITION: Imbalance occurs at running speed and produces a clean sharp spike at 1X. Running speed vibration is very often misunderstood and gives many technicians problems. I've seen technicians attempt to balance rotors with high 1X and almost nothing else - no 2X or 3X and relatively smooth the rest of the base band. Misalignment (gross) will totally manifest itself in 1X and you'll play havoc attempting to balance that problem. Bearing problems and faulty machine parts will manifest themselves into running speed vibration at some percentage as the machine deteriorates. That percentage may be very small but a bearing flaw (for example) will produce drag and contribute some magnitude into running speed though its source may not be determined. Analyze the spectra, different data points, and different positions with phase data. Correct faults then balance. All too often I see technicians balance and amplitude at running speed will become acceptable and it'll be called good. However, that balance job didn't magically cure the machine component that also has a fault.
Never just balance a machine to an acceptable vibration level and walk off with out some analysis to determine machine condition. Imbalance in vertical machines is usually highest transverse to the flow or in the most flexible direction. Bottom mounted machines usually reveal higher 1X vibration in the horizontal direction.
Once imbalance has been determined to be the fault, balance the number of planes necessary to achieve smooth machine operation.
When field balancing one may want to use eddy current probes or existing transducers. Most large turbo-machinery will have a permanent monitoring system installed that you can plug into in a control room or local panel.
On paper machine dryer cans where you balance two planes and still have excessive amplitude: does the phase match - yes, then whip is the problem and requires third plane balance or further testing. The third plane is the center of the roll.
1X vibration problems. These are imbalance, resonance, bent shaft, external mechanical run-outs (coupling hub, pulley, etc., machine part, electrical, eccentrics) internal run-outs.
Phase measurements (see photos below) are very useful in determining misalignment and bent shaft but one of the easiest methods of detective work is to simply shut down the machine when you can and if practical. This enables you to monitor power off cascading to a complete stop. Use a good analyzer with capturing capabilities or a tape recorder when you can.
Often the vibration disappears upon power off if electrical in nature. By monitoring coast down one may record via instrument tape recorder, or internal capture in a real time analyzer, information that can identify resonance. Once the machine is shut down, physical measurements can be taken; such as shaft run outs, coupling run outs, alignment, etc. On electric motors you may find a chipped blade on its fan. A problem can be simple too.
Strobe for phase and/or motion study.
Photo tachometer for once per sync or trigger.
OK, how do we check and solve for these problems?
Let's go to misalignment; it is usually associated with a 2X component along with 3X in the axial direction. However, gross radial misalignment will manifest itself in 1X and not even show 2X or 3X. Angular misalignment is usually found at 2X in the axial direction. Using phase as a tool measure both ends and across the coupling. When searching for an unknown source, don't form an opinion before you've gathered sufficient data. I've seen people try fruitlessly to balance a misaligned machine. It happens too often. Of course a good maintenance program will have alignment records. And those alignment records should be truthful and exact so they can be used for an analysis tool. If physical measurements must be taken one may do so if the situation permits.
Use good field proven techniques and procedures to gather alignment data and/or correct alignment.
The procedures that follow are for all types of alignment. These are field proven and work. They will enable you to align in the shortest period of time. The types of alignment are: Indicator Reverse, Long Spacer Coupling and Rim & Face. The methods for these types are: Laser (single and dual beam), dial indicators, parallels and micrometers.
Of course this one with the big picture is my personal kit (shims are extra): PdMK-2.
SAFETY; lockout and tag. Rack out fuses and chain valves when necessary
Ensure driven machine has no stress from piping or soft foot and that the anchor bolts are tight
Inspect the feet and base of the machine to be aligned for burrs, roughness, dirt
Evaluate if a temporary spacer is need or useful for this alignment process; note
Inspect coupling and check for run outs
Coupling halves must if possible rotate together for dynamic alignment
Evaluate method of alignment to be used
Which is best suited for this job
Is axial thrusting a factor
Collect SAG data and record
Install brackets on machines and record machine dimensions
Check for soft foot; correct as necessary
Rough align within 1/16 inch
Record alignment readings
Are they valid?
Mathematically correct and repeat?s thrust a consideration?
Calculate shim and lateral shift change
Via computer? Print and file.
Calculator? Record, plot and file.
Make changes and verify & PLOT
If spacer was used, remove it at this time
PLOT THE FINAL ALIGNMENT DATA
UNLOCK AND REMOVE TAGS; NOTIFY OPERATIONS
PLACE RECORD AND PLOTS IN MAINTENANCE FILE
2X running speed - mechanical looseness is a prime factor when this type of vibration occurs. It is radial in nature and usually predominate in the vertical direction. It sometimes pays to acquire data at the foot or split of a bearing cap. Reciprocating forces; are you analyzing a compressor?
If you are fortunate enough to have a good vibration lab, you can duplicate most vibration situations or create experiments for further study. Bentley Nevada markets a good lab kit complete with eddy current probes and oil whirl attachment.
Measurement techniques are important as well as applying common sense. Approach the machine like a crime scene. Notice everything; does the foundation appear correct and properly isolated? Are there cracks on the Floor; stand on each side of the crack - is there a difference - measure? Can you feel or hear a beat? Use a stopwatch, count the beats over a five minute period. I've uncover the source of vibration using this technique. Peck on a sole plate, is there thud? Are there splits in the case or bearing housings - measure both sides. Eddy current probes give rotor dynamics, but also take cap readings.
Cap readings make up machine and rotor movements. Is there a difference from machine to base? What about phase end to end, radically? Phase axially: measure every ninety degrees.
In low frequency problems, resonance is usually a consideration. Can you alter speed or other factors? A coast down provides good information. While the unit is down, perform ring testing - this will give you natural frequency. One problem can excite another.
Another consideration in taking low frequency measurements; low pass filters. They're available in a variety of ranges. Low pass filters can let you blow up low frequency signals. This works well especially with an analyzer having a 90 dB range or better. Also real time zooming is a good tool. This can give a resolution of 0.0156 Hz or better; very good to separate frequency components. Sixty Hertz from running speed; side bands around 60 Hz such as around 1.2 Hz delta freq and so on.
While we're in low frequency stuff, let's consider half running speed. A rub in a sleeve bearing will show up at ½ and may be very severe at 0.03 IPS (inches per second).
What is a machine's mass, dampening factor? A rub at 0.05 IPS is trouble. Also look for sub-harmonics. In reality; you never want to see a rub.
Sleeve bearings: rubs, excessive clearance, misalignment, lubrication and looseness are main concerns. Since these bearings ride on a film of oil and the hydraulic pressure provides certain lift, clearance is an important function. Excessive clearance will cause harmonic content. Oil whirl is - a riding up and slipping back on the journal; when it becomes violent it is called whip. Oil viscosity is an important factor and one must consider the type of oil system.
We have oil bath, oil circulation, ring oiled, mist lubed; etc. Lubrication usually isn't a big problem anymore. I once encountered an oil whirl in a machine with five tilting pad bearings. If a sleeve bearing is loose in it's housing it can produce a 4X component; make sure the 4X isn't being produced from another source. Again, always consider a phase study.
The fifth and sixth harmonic (5X & 6X); machine parts and looseness are main contributors but, the 6X can be generated by a loose rotor and is usually accompanied by 4X. You may have never seen this in print before, but this comes from practical experience. Just like the misalignment condition manifesting itself exclusive in 1X. Physically setup that condition in a machine and let it prove itself.
I made this statement in 1983 about acceleration; within ten years everyone will work in acceleration. Velocity will only be used as the universal language and rightly so. However, you can't argue with success; if you're using velocity successfully and almost only worked in velocity: give acceleration a try. Back in 1980 velocity was the new kid on the block for most. I think you'll be glad you switched. I'd rather fight than switch back. There was also a transition period between displacement and velocity. Who works in mils today for a PdM program? Remember, acceleration is the right thing to do: integrating to velocity is not absolute only relative. There is no pure integration and you can't analyze what you can't see.
Anti-friction bearings: use both time waveform and frequency spectra. These bearings transmit forces more readily than sleeve bearings and the stiffest direction usually reveals a higher magnitude of vibration. Banding analysis and overall level diagnostics work well.
On 3600 RPM machines use a base band of 5 kHz. For purposes of resolution I normally view 2 kHz as my primary base band. But, never limit anything when troubleshooting. Defective anti-friction bearings are a product of mechanical deterioration. When this occurs, a sharp spike will broaden: study spectral shape and use shape as a diagnostic tool. Overall analysis is also useful but always with the frequency spectra. Windowing as in different base bands is also a tool.
There's more than just the Hanning window too. Explore! Technical Associates of Charlotte have a good section on A-F bearings; R-0792-3, page 2. A good analyzer is a very valuable too. Just any ole black box won't do. There's more than shown.
This has been the common analytical approach since the late 70's. A word of warning; spike energy is not a good evaluation parameter, never rely on it for a tool. And, it isn't good at all on low frequency machines; i.e.. paper machine dryer cans. On high RPM machines note the harmonics of bearing fundamentals. Another factor of detrimental high frequency vibrations related to A-F bearings - shaft currents. This can be measured (shaft currents) with a good DMM. You can also have a buildup of static current that discharges through the bearing. Again, use 5 kHz base band on 3600 & 1800 RPM machines in conjunction with your normal observance frequencies. If magnitudes of vibration are observed near the end of the spectra (high frequency region), then you may want to look at the next higher base band on your analyzer.
With the absence of machine defects, bearing defects having a magnitude of 4-7 g's 0-P over 5 kHz are nearing failure. This is a broad statement, mass & dampening must be taken into account. Example: 5 g's 0-P over 5 kHz @ 3600 RPM usually will fail within a month; maybe a little less, maybe a little more.
But, the same machine situation @ 1800 RPM may take two months to fail plus or minus a little. Please, this is an illustration, so don't hold me to exact numbers although the scenario is close. In PdM program mode use trending, if the program is old enough. Continuous monitoring systems provide good trending information much better than a monthly or quarterly route program. Monthly or quarterly monitoring programs require a good chief analysis. Trending is valuable on continuous monitoring programs but has short comings on anything less.
Lastly, lubrication or lack of it. Check schedules, quantity per time interval, type, etc. A bearings natural frequency will be excited in the absence of lubrication. Lubrication makes the wheels go around. Alignment makes them run longer, PdM makes them cost efficient.
Got a lube program? Your grease guns should be labeled - one shot equals so many grams! Each grease lubed bearing should be given and exact quantity over a given time interval (i.e. 28 grams per month; 14 grams per two weeks). Classify machines as to speed and environment to determine frequency of lubrication. Lubrication is your single most important program; not alignment or balancing or PdM. I have papers I've written on grease lubed bearings and quantities per given time interval. I like to calculate that time/time interval as accurately as possible. If I have a good PdM program it will fine tune those numbers. Therefore, when a good PdM program is in place one can know when the lube is lacking. With this in mind; calculations and estimations can be ‘backed off' slightly and fined tuned over time through frequency analysis.
Gearboxes: These can be very complicated; I don't consider any to be categorizes as simple. Correct gearbox set up is essential. It sometimes becomes necessary to stress a gearbox to achieve internal alignment. I was call on a very high H.P. unit operating at 12.0 g's 0-P @ gear mesh. Upon shut down I inspected and measured clearances. There was originally internal misalignment. But, it had already worn a pattern. I recommended to continue operation as distorting the case to correct internal misalignment would only cause an increase in vibration level. They opted to distort the case. Vibration increased. I again recommended no action as it would now wear a new pattern and reduce in magnitude. It was left there and monitored regularly. At 18.0 g' it was inspected and then put back in service to 20.0 g's. This is unusual but not odd. Always use time waveform as an analytical tool. The defective gear will be modulated or side-banded by the defective gear's shaft speed. Also, look for a small `side spike' on the side of defect peak in time waveform if a gear is cracked or chipped.
A voith coupling (high H.P.) may give you a bit of diagnostic problem as they can have an amplitude/frequency display in an axial position that appears excessive. But, It's a characteristic of that box.
Some of these things are only obtained from experience or being around someone with experience. In troubleshooting always draw on all natural resources available to you.
I believe that's a good philosophy in most all aspects of everything you do. Gear mesh frequency in most cases is in trouble at or around 10.0 g's (high frequency). Also, look for half gear mesh. How is the fundamental affected? Use phase.
Sometimes you are put in a position of doing diagnostics above your head. If you gotta, you gotta: get your feet wet. But, a critical, costly machine?
Maybe it's time to call the outside consultant, who? Again, these are guidelines to help you troubleshoot and solve your problem. Know when it's time to call in help. I have developed these standards with and through others: The chief analyst must have at least ten years experience as an analyst with five years spent as an outside consultant. It shouldn't insult him if you ask for a resume' and sample report as well as references. Is it warranty work; then the work must be traceable to the NIST, formerly The National Bureau of Standards, now, National Institute of Standards and Technology.
I have developed standards for acceptance of pumps & motors, plus specifications for motor rebuild shops.
On a last note; don't over look the simple. Calibrate - using a level, check calibration. Alignment -check calibration before every job: Indicator Reverse, Long Spacer Coupling, Rim & Face and especially Laser. Using a micrometer - check against a standard before using.