The unit for displacement is a mil, which is the same as a thousandth of an inch right? It seems that displacement is a favored unit of measurement for vibration analysis. Why is this?
Does displacement directly coincide with the actual movement in thousandths of the transducer?
Would that movement be considered peak to peak, or 0-peak?

Sorry for the elementary questions, before balancing I have never had to work with displacement.

Hi Andy, mil is 1/1000 in. We Use µm or 1 millionth of a meter. 25.4 µm to 1 mil. In general you would use displacement as peak to peak. (The analyser set up should say which) You would generally use displacement measurements for low frequency i.e shaft speed related issues and it would also depend on the transducer you are using. If you are using a proximity probe which measures relative displacement then mil or µm peak to peak is used. If using an accelerometer which measures absolute vibration then you would do a single integration to get velocity for shaft related vibration (balance, alignment etc) veleocity is generally peak or RMS in/s or mm/s. For high frequency i.e rolling element bearings gear mesh etc, then acceleration in gs or m/s/s peak or RMS is used. You can use dispalcement, velocity or accel for balancing depending on the transducer but in general dispalcemnt for prox probes and veleocity for accelerometers.
So to answer your question displacement does coincide with the actual motion of the shaft relative to the transducer. The parameter you use depends on what you are looking for and the transducer. You would be wasting your time looking for high frequency vibration in terms of dispalcement as the signal would be lost in the noise as dispalcement is related to acceleration as the inverse sq i.e if you have 1 g at 1Khz from a accel then you would have 0.06 in/s (1.5 mm/s Pk) or 0.02 mil p to p (0.5µm p to p) The converse is true low frequency signals show up best in terms of displacement. If you are using an accelerometer for low frequencies then best to use one with a high output (1000 mv/g) so you can see the signal in the transducer/cable noise.

Regards

MDE

There are others here who are probably more of an expert on this than me, but I have some ideas about this. When I graduated in 1981 I went to work at a large coal-fired power plant. The only equipment with continuous vibration monitors were the main turbine/generator sets, the boiler feed turbines/pumps, and the boiler fans. All readings were monitored/recorded in displacement (mils, peak-peak). The turbine-generator used shaft riders which measured actual shaft movement which was peak-peak (total) displacement. The boiler feedpumps had prox probes which measure total (peak-peak) shaft displacement. And the fans all used velocity pickups, integrated to read peak-peak displacement. One set of fans used Dodge sleeve-oil bearings, but the other two, larger sets had rolling-eelement bearings. The fan motors all used sleeve bearings, but had no vibration monitors.

The largest and most critical fans were the Induced Draft (I.D.) fans which were 7500 h.p., vane axial, overhung fans. These fans still used "displacement" when I left in 1995.

I think vibration monitors probably originated on large turbines and compressors and the first were probably prox probes. G.E. used shaft riders on their turbines... not sure why except maybe they thought they were "better" since they measured actual shaft displacement instead of measuring "gap voltage" as prox probes do. They may have done this to be different from Westinghouse who (I believe) always used prox probes.

I think the velocity sensor was developed for use on machines that did not use sleeve bearings, and thus had very little actual shaft displacement. Using displacement for 'monitoring' machines that have rolling element bearings was a terrible idea, but I think was done as a holdover from the original displacement monitors. Measuring displacement on turbines and such still makes sense because you are generally more concerned with whether or not you are exceeding the available "clearance" at the rotor tips, steam glands, and oil seals. Throwing "velocity" into the mix on the R.E.B. machines would have been confusing at the time, and often is today as well.

I got a frantic call one day back then to check one of the large I.D. fans because it was running around 3 mils (toward the upper limit, but not that unusual). I asked why they were so excited, and they said the oil filters on the fan lube skid were plugging up about every hour or so. "Plugging with what?" I asked. "I dunno... some sort of grey, powdery stuff"... which of course was the metal from the almost total disintegration of one row of the double-row bearings (the thrust bearing had not been shimmed properly, so the motor-side row of the radial bearings worked as a thrust bearing -- 7500 h.p. vane axial fan). By the time I got a sensor on it and got my meter turned on, I was reading around 1 in/sec (peak). A 1x component of 3 mils (pk-pk) would be equivalent to 0.14 in/sec (peak) at 900 rpm. So the overall value of 1 in/sec tells you there was either an awful lot of higher frequency energy or a lot of low or mid frequency broadband vibration.... in this case it was both.

I don't think "displacement" is a "preferred" unit of vibration, except if you are using prox probes or for balancing.

I use displacement for balancing because it makes the "geometry" easier for me to deal with (lag angle, weight placement, etc), and because I just like using larger numbers. I use velocity for general vibration analysis & trending. But I rely a lot on acceleration for R.E.B. trending and analysis. Really we should be aware of the proper use of all three units and the limitations of each.

As Rusty said, displacement is often preferred for sleeve bearings relative displacement monitored by prox probes. One thing it does is gives you an easy comparison of the shaft movement compared to the bearing clearance.

Another situation when some people prefer displacement is on slow speed equipment. The reason stems from typical severity guideleines. For example NEMA MG-1 gives limits for new motors tested on rigid bearings of 2 mils pk/pk displacement, 0.12 ips, and 0.8 g's. Assuming the vibration were single frequency, anywhere below about 1000 cpm, you would hit the 2 mils limit before you hit the 0.12 ips limit. Between 1000 and 24000 cpm, you hit the 0.12 ips limit before the 2 mil or the 0.8 g's. Above 24,000 cpm, you hit the 0.8 g's limit before the other two. So it makes some sense to choose the units based on the frequency of interest (displacement below 1000cpm, velocity above and up to 24,000 cpm, acceleration higher).

For other than low speed equipment and sleeve bearing prox probes, I think velocity is the most common indicator for overal machine health. Rolling bearing defect monitoring of course is another story.

Shaft riders pre-date proximity probes. The shaft riders that I've seen used seismic velocity probes; probes that have a coil in a mass/spring suspension system with an additional magnetic field where the low frequency resonance of the mass/spring determines the low frequency response of the system. There have also been versions of these which used a diect mechanical connection to the moving mass/coil that aren't limited by the low frequency resonance of the mass/spring.

The main reasson that these velocity probes were used is that they had a lot of output to drive the early tube type instrumentation. These devices were like dynamic microphones modified to have a large mass attached to the moving diaphram.

Displacement is useful as a parameter if you wish to emphasize the low freuency parts of a vibration signal. In general, it's probably the easiest parameter for a user to visualize but probably the least useful parameter to use for assessing machine condition.

Some of the early vibration vibration analyzers/balancers such as those made by IRD in the 1950s only had displacement for output.

You can balance in any parameter you wish. I use g's to balance in conjunction with an accelermeter and photo tach due to its ease of use. Some balance in IPS while others use mils.

Duncan's answer is to the point and very good; also in keeping with frequency characteristics. When using g's on low frequency and your only interest is low frequency the use of filters, say a 100 Hz LP, allows you to view low frequency signals more easily.

quote:

I think vibration monitors probably originated on large turbines and compressors and the first were probably prox probes. G.E. used shaft riders on their turbines... not sure why except maybe they thought they were "better" since they measured actual shaft displacement instead of measuring "gap voltage" as prox probes do. They may have done this to be different from Westinghouse who (I believe) always used prox probes.

Circle bar W adopted the newer technology sooner. One could tell the Cirlce Bar W machines from the General's, because the General's turbines had the shaft riders 30 degrees off vertical (60 off horizontal), and Circle Bar W's were top dead center. Besides the spun they each spun in opposite directions.

Of course one can balance with any units, but some caution is needed when using optimization type balance software, e.g. least squares error methods. The programs don't attempt to bring all vibration values to zero, just minimize some objective function, e.g. the sum of the squares of the residual vibration. As such, displacement does (or may) not compare equally with velocity or acceleration; so, the measurements should be weighted to properly reflect the mix that you desire to 'balance.'

Weighting can be used for other purposes as well. Just converting everything to displacement may work, but often shaft relative (or ablsolute -Yyes, shaft riders are still being sold.) and casing measurements don't compare either, or perhaps what is important to balance is one or the other.

For example if one had a turbine to balance and one was comparing bearing and coupling readings, one might not want to weight the coupling readings as high as the bearing readings for the balance program. If one balances at multiple speeds, one may use less weighting near a critical speed than at running speed. Of course all these selections are the responsibility of the balancer, and this helps to make balancing fun.

When you guys mention shaft riders, what are you referring to?

you guys are a wealth of information!

What about the transducers that are basically a load cell? Same as displacement, or is there a fundamental difference?

I've seen a balancer similar to what Duncan mentioned, A mass attached directly to a piezo buzzer to create a crude transducer, seemed to work pretty well.

Man you vib analysis guys have so many variables, I thought I had it hard just trying to extract 1x imbalance! My balancer only locks on the peak amplitude once per rotation, and I'm low passing the dookie out of the signal. The rest of the low pass is done with DSP, and it's still so easy for other frequencies to dominate the signal.

Now this may be a stupid question, could displacement be used in a controlled test to do a "rough" calibration on an accelerometer? Possibly using a test rig with a decent amplitude low freq vibration and a 10ths indicator. I ask because my accel's
are not calibrated to any accurate figure. The G's I use are based on calc's of the rated output of the MEMs chip/amp gain/filter attenuation etc. I don't think I need perfect accel
calibration, as accuracy in my application comes from the linear response of the transducer and a known trial weight. It would just be nice to have some way have a rough number to back up the theoretical.

quote:

I thought I had it hard just trying to extract 1x imbalance! My balancer only locks on the peak amplitude once per rotation, and I'm low passing the dookie out of the signal. The rest of the low pass is done with DSP, and it's still so easy for other frequencies to dominate the signal.

One has to be careful how one computes the 1X if you care about either amplitude or phase.

Reinforcing what Bill said, you need to make the 1x amplitude and phase measurements in a very consistent manner so that the vector difference calculations will be meaningful.

Andy; yes you can use displacement to do low frequency calibration data. Most accelerometer calibration data is acquired with hardware that does not lend itself to very low frequency operation because of power requirements. Some, if not many or all, use signal generators to drive the accelerometer amplifier circuitry and assume that this gives a correct result; if you think this gives questionable data, you are correct.

I have an antique ( Knowles Instruments mechanical exciter circa 1965 ) that generates good sine waves in displacement at amplitudes up to 25 mils at frequencies below 40 Hz down to as slow as you can turn the shaft. If you are trying to use this kind or data to calibrate accelerometers, you need to be extremely careful about the speed measurements. AFAIK, these have long disappeared from the marketplace.

This message has been edited. Last edited by: Duncan Carter,

quote:

One has to be careful how one computes the 1X if you care about either amplitude or phase.

William,

I assume you are referring to the stability of the amp and phase readings?
Of course before a reading is taken, averaging
is done and other variables are considered before one can assume they are looking at the peak 1x. I would assume if one were attempting to balance and after low pass and tracking could not give you 1x, maybe you should be looking at the other dominant frequencies prior to attempting to balance as a solution. Soon I am going to try a windowed sync filter, which will completely isolate any frequency of interest.

Duncan,
I can run a signal generator through my amp circuit to verify gain, it's just that the acual ic datasheet states "approx X mV per G" at X volts, along with a ratiometric output related to input voltage, and also an output resistance that can vary as much as 15%. Couple this with high amp gain, and, well, you get the picture. I would like to come up with a simple, semi-accurate calibration for my accels. Now the extremely sensitive MEMs have a 2G measurement cap and are extremely sensitive, and DC coupled you can actually just tilt the accelerometer vertical and measure the voltage change. I have done some testing with these but the problem here is if mounted vertical, half of the dynamic range is lost, and signal clipping could be a problem on anything approaching 1g.

(Low Pass) Filtering will alter the phase and would need correction to get what most know as phase for balancing.

It is not clear what you mean by tracking. The techniques for this vary.

How do you get the amplitude and phase?

Andy:

Flipping the device is one way of calibrating some old style mechanical vibration switches. The method is crude but so is the hardware.

You are the first person on ths board who uses MEMS that I've enountered. I was really thinking about common ICP accelerometers which are internally capacitively coupled devices so that they have at least first order rolloff below some frequency. Usually, these ICP devices are gain trimmed by the OEM but they can get damaged with use.

I have a phase offset least squares fit curve, a "generic" lag amount as well as the phase offset due to low pass filtering is added or subtracted to the signal according to the frequency and direction of rotation. Amplitude is the 10 bit 0-peak value of the 1x phase location. 0 phase is identified with photoreflective tape and a laser tach.

Do you use a trigger or a two channel relative phase measurement and do you calculate phase from ffts?

I capture the entire waveform (100 samples per rotation), and phase is located by sorting through the sampled data. No FFT yet, a little DSP though.

It doesn't sound extremely precise, but considering that I've balanced in a balancing stand with dead measurement electronics and no drive motor and only a dial indicator for an indicator, who am I to complain?

If you do use a fft or some other averaging method, be aware that you need at least 4 periods of the signal at the shaft frequency to begin to get reasonable accuracy.

There are also pitfalls in getting triggered data, especially with older A/D hardware.

What kind of machinery are you testing?

I don't get the direction of rotation and adding or subtracting a phase correction

What happens when a low frequency component is present? Has this been checked against a known technology or instrument? One needs the filtered 1X component for balancing.

Bill, of course it's checked against known data, thats what builds the best fit line. The known deviations are then combined with the raw phase angle to arrive at the actual phase angle. I shouldnt have mentioned direction of rotation, that is more relevant in the actual functions and subs that the application uses to calculate, and I don't want to get into that

You know, I'm going to make a video tonight of the balancer in action, I will provide a link to the vid and you guys can pick it apart from there...seriously, I do need the input.

I know the subject has deviated, but somehow we will get back to displacement!

Getting back to displacement. I am at a coal and gas fired power station with 4x 250MW single reheat Parsons turbo-alternators (late 1970s vintage). The OEM vibration was pivoted coil veleocity transducers the signal from which was integrated to veleocity and the peak value plotted on chart recorders. The OEM terminology was that amplitude was always displacement 0 to peak. These machines had a single prox probe on each of the three turbine shafts to measure eccentricity. The problem is that these machines are mounted on all steel foundations which created a resonance on one of the generator bearing at 100Hz (2x machine speed) When monitored in terms of dispalcement the level was higher than disirable but within the OEM alarm limits but if looked at in terms of velocity was not acceptable. The integration process was attenuating the significance of the 2x component. When one looked at the OEM alarm settings they aligned with international overall vibration standards when you converted the velocity levels to dispalcement at 50 hz (running speed) During the run up because of the lower frequency the alarm limits were quite conservative. We have since removed all the old monitoring system and installed accelerometers single integrated to velocity. The signals are monitored with an online monitoring/ protection and diagnostic system which profiles the alarm levels for both 1x and 2x. The generator goes through 1st and 2nd critical plus a bearing resonance the last two being close together and just below running speed which makes balancing "interesting". See attachment.
What I am alluding to is that when using a single parameter to guage the condition of the plant be aware of using displacement as the severity id frequency dependent.

Run_up.pdf (29 Kb, 19 downloads)