A recent vibration inspection showed low frequency noise (< 15 Hz) in the textbook ski-slope pattern on an accelerometer reading integrated to velocity.

An additonal data collector and accelerometer were used with similar results.

Previously, a velocity transducer was used for monitoring at this location, with notably less low frequency noise (~ 10 times less).

Mass of the accel is ~ 49 grams
Mass of the vel trans. is ~ 145 grams

The monitoring location is on a steam turbine driven pump at the turbine bearing. The accelerometer temperature is ~ 150F after monitoring the three locations on the bearing.

Readings taken with the accelerometer on the pump bearings were consistent with historical trends.

The data sheet for the accel shows an approximate 8% deviation in accelerometer sensitivity at 150 F.

Amplitudes at frequencies above 15 Hz are relatively unchanged on the accelerometer readings.

Do any of you have experience with hot accels on your walk-arounds? Is the deviation at low frequencies and consistency at higher frequencies typical? Is this a stabilization issue?

I've got a call in to the accel vendor, but no response, as of yet...

Skislopes generally result from 2 conditions - overloading and CHANGE in temperature. Levels greater than 70 g peak will result in overloading. This is probably not the case here but on turbine-driven machinery, turbine blade rates may excite the accelerometer's resonant frequency resulting in overload.

I think in this case, the skislope is the result of a temperature change. If the accelerometer is used with a portable data collector, it might be at 70 degrees F. Being applied to a 150 F location results in big low-frequency levels until the temperature has stabilized. If the accelerometer is indeed temporarilly mounted here, allowing several minutes for thermal stabilization should reduce the ski slope.

IF the accelerometer is permenently mounted, it's still possible to get this thermal ski slope if cooling air is swirling around.

To reduce the effects, first be sure a shear-mode accelerometer is used. The older type compression accelerometers are much more responsive to thermal changes. If a portable accelerometer is used, consider installing a permanent accelerometer so it will reach thermal equilibrium with the measurement point. Finally, if it's a problem with a permanent accelerometer, arrange the cooling airflow to provide steady conditions with either a direct flow across the accelerometer or no flow across the accelerometer. Industrail accelerometers usually tolerate 257 F, so no harm will come to the accelerometer from limited cooling.

Jon
Spintelligent Labs

Hi Jamie,

Let me get this straight, you took readings on the turbine ... the LF slope occurred ... and accel temp was +150F ... then you took readings on the pump and they match historical data ... ???

It sounds more like an overload situation, where the accel wasn't allowed to settle ... are you using magnetic mounts? If so, try "rolling" them onto the machine instead of the usual *SMACK*.
(I know some of you guys that use the "bolo" method might disagree)

Also, I have seen this with old cables, where connections were loose.

Any chance you can post some data pics?

I also agree with Jon, on the thermal issue.

How many still use the rubber boot for thermal stabilization?

Sam,

I don't believe the rubber boots are still in use. They were a PCB "innovation" for dealing with temperature fluctuations on compression-mode accelerometers.

Industrial accelerometers available today from CTC, IMI/PCB and Wilcoxon are all shear mode as far as I can tell, but many compression-mode accelerometers still in use.

I had some interesting results when testing an outdoor cargo crane a few years back. The job took every acceleometer I could lay my hands on, including a couple old Dytrans. These made great event markers, because I got a big spike on every transisition between sun and shade as the crane traveled down the track!

Jon
Spintelligent Labs

May be a case for thermal boots afterall. I still have some - even new.

Why is it that accelerometers respond this way to a change in temperature? Is there a brief 25-words-or-less explanation?

Stress; basically same as cable strain to be short and dirty.

These little baby's are sensitive. I like microdot cables and absolutely do not move while collecting data and hold the cable so that it doesn't induce strain into the transducer - overkill? I'm calling shots to within a week when something will fail or two weeks at the most generally to provide predictive maintenance (scheduling range within 3 weeks). So with my neck streched, I want every advantage I can get to having reliable data.

Temperature change or bending may impose "bas strain" in the sensor material inside the accel. Sensor will output signal from whatever force that hit it also from temp change that make material move or bending from the mounting, also depending on sensor design. Olov

Let me add a bit to what Sam and Olov have said....

In a compression-mode accelerometer, the piezo crystal is sandwitched between the accelerometer base and the inertial mass. Any movement between the base and inertial mass compresses the crystal and results in an output. Because the crystal is directly against the base, any distortion of the base (i.e., base strain) results in compression of the crystal and an output. Temperature changes of the accelerometer result in uneven expansion, having the same effect.

In a shear mode accelerometer, the crystal is arranged differently to reduce or eliminate base strain and thermal effects. Rather than depending on the compression of the crystal to generate an output, the crystalography is changed to produce an output when the crystal experiences movement in the shear direction. In a shear mode accelerometer, 2 or more crystals are attached to a post extending up from the base of the accelerometer. The inertial mass is a ring which compresses the crystals to the post. Now, instead of compression being measured, shearing of the crystals produces the output. Distortion of the base and temperature changes don't result in any shear force across the crystal.

A simple experiment can be done to show the difference in sensitivity to temperature. If you connect a compression mode accelerometer and power supply to a real-time analyzer or o'scope and look at the time domain, simply blowing across the accelerometer will result in large voltage swings! Switch to a shear-mode accelerometer and this effect is almost nil.

Keep in mind the displacements an accelerometer can measure are tiny. At 10 kHz, 1 g rms is only 0.0003 mils pk-pk of vibration! And the noise floor of an accelerometer is typically in the milli-g range, meaning an accelerometer can respond to displacements in the mirco-inch range.

Jon
Spintelligent Labs

A couple sketches will help clarify.....

Great info. Thanks Sam, Oli, and Jon.

Hey guys ... thanks for all of the responses. I was in training all of last week, so only replying now...clarifications:

The pump readings were taken before the turbine readings, so no temperature effect on the the pump readings - only on the turbine readings.

The accelerometer used was an Entek (I believe that its a PCB made accel) 9000A, which I believe is a shear mode accel.

The velocity transducer was a Wilcoxon 793V, shear as well.

A plot of the data is attached. Keep in mind that the 'normal' readings for these points are extremely low, so there's not much happening in the spectrum ...

We will be conducting some tests hopefully this week and will post the results here.

Thanks again.

793V_and_9000A_Plots.doc (58 Kb, 28 downloads) 793V and 9000A Plots

Jamie,

One point I must admit that I missed in your orignal message is that this measurement is on a turbine position.....

Overloading the accelerometer results in clipping the signal and ski slope patterns too. The levels shown in the plots you posted are low, but the frequency range isn't high enough to see turbine bucket-rate vibration. Tubine bucket-rate or harmonics usually won't be high enough to result in overload on their own (about 50 g peak on a 100 mV/g ICP-type accelerometer), but if any of these peaks is close to the mounted resonance of the accelerometer, it's trouble!

The resonance of the Entek 9000A is 28 kHz. Depending on how the accelerometer is mounted, the resonance will be lower than this. I'm guessing the sensor was mounted with a magnet, so the natural frequency will be much lower. Excitation of the resonance frequency is a very real possiblity in this case.

To rule this out, first check out the spectrum up to 20 kHz or higher to check for turbine burket rate and harmonics. If the trubine has more than one stage, each stage will have a diferent forcing frequency.

To change the natural frequency of the accelerometer, try some different mounting methods. The best case is to drill, tap and spot finish to stud mount. Almost as good and much less effort is to glue down a threaded mounting base - just make sure the mounting surface to cleaned down to bear metal and use the thinist layer of rigid adhesive possible. For a short-term test, "Superglue" is an excellent choice. Either of these mounting methods will raise the resonance frequency compared to a magnet.

On the other had, and this is the only time you will ever hear me suggest this, is to try a measurement using a stinger probe. The stiffness of a measurement like this is even worse than a magnetically-mounted accelerometer, which will lower the resonance frequency....with any luck, the blade rate vibration will be higher in frequency than the resonance frequency, eliminating the ski slope.

If stud mounting solves the problem, stud-mounting a permanent acceleroemter at this location will solve the problem (just be certain to buy an accelerometer with a high natural frequency!). If stud mounting doesn't solve the problem but using a stinger does, the answer is still stud-mounting a permanet accelerometer, but a "high frequency" type must be used with a higher natural frequency.

Despite Rusty's comments (and others), I would never recommend a hand-held stinger for vibration readings. The lower frequency peaks, such as 1x and 2x rotation will be accurately measured, but higher frequency peaks, such as turbine bucket rate can be all over the place depending on how the probe is held, the angle and pressure applied and the exact location of the probe tip.

Jon
Spintelligent Labs

Update - I work with Jamie and we're still struggling with this problem. We appreciate your great comments and feedback.

The Entek 9000A is a shear mode accelerometer. We also tried a Wilcoxon 793 compression mode accelerometer and got similar results.

We are using quick-connect (3/4 turn) pads epoxied (Devcon titanium putty) to the bearing housing.

We took measureemnts using a magnet adjacent to the pads and the spectral data looks consistent with our historical data.

When collecting data, the overall (analog) measurement is having difficulting settling. We are using 8 averages (0% Overlap) for the spectral data. When we acquire spectral data, we see the large low-frequency ski-slope and having frequent overange-settling iterations. Eventually, the data collector produces a spectrum, but it doesn't look real. I'm seeing this as a temperature/integration problem with the accelerometer and data collector.

We took some high frequency spectrums that show a peak at around 4160 Hz, which is approximately 70X. We suspect that this could be a turbine blade passing frequency or a mounting issue with the pads.

The amplitudes from the timewaveform are approximately 7 g's with the magnet and 35 g's with the quick-connect pad, which indicates a high-frequency effect.

We would appreciate any additional comments.

There were many fine comments on thermal, high frequency overload, cable motion, measurement time, frequency range, mounting looseness, and accelerometer type. My guess is that the problem lies with the Quick-connect pad. There is either a resonant amplification effect or a looseness effect. I suggest removing the pad and adhesive. Attach same accelerometer onto a thinner threaded metal disk (stock glue disk) that is monted to same location with epoxy or Super Glue adhesive and make comparable measurements.

I have worked with a lot (1000's) of plain flat steel disks 3/4" diameter and 3/8" thick attached with epoxy adhesive. On a rare occasion a goofy measurement can occur from a disk-type mounting pad. Remounting pad and slight change in location usually solves the problem.

You may also want to use analog integration (before spectrum) when trying to compare velocity from an acceleromter to a velometer. The integrator pole frequency should be set to give comparable frequency response to the velometer.

Based on Steve's comments, I believe the accelerometer is overloading from high vibration levels. For some reason in my previous post, I didn't think to mention trying a lower-sensitivity accelerometer. Brain fade I guess

A 25 mV/g accelerometer will buy you 4x the headroom, at the cost of a higher noise floor. 10 mV/g accelerometers are also available providing even a greater range.

Jon
Spinteligent Labs

Steve--

The high frequency peak you're seeing may be what I've called a "flow effect" related to velocity of steam through certain parts of the turbine (to be really unscientific). I have it on various turbines at my plant and it seems to come and go (meaning I haven't related it to a certain speed). Historical testing with order-based data has discounted it from being a vane pass phenomena.

However, I've also had plenty of experience with the temp effect--especially when steam is leaking through the seal at the turbine bearing (as evidence by water condensing on the transducer--good time to be wearing gloves!).

Tony