I have seen some analysts refer to collecting a "special time waveform" overall in Gs and using it as an overall leading indication of bearing condition. If you collect peakvue data is this extra waveform necessary?

Regards,
Alan

Depends on what they mean by "special". A time waveform in g's is basic raw data which will show repetitive events if the time span is sufficient. I have used a "gSE" (similar to PeakVue)filter on time waveform data to highlight the impacts. This is a high frequency demodulating filter which removes synchronous elements from the raw data and leaves the non-synchronous (sometimes synchronous if amplitude is high enough)events which could be bearing defect frequencies. Hope this helps.

I can see one difference between "gSE" and PeakVue. The recommended setup for PeakVue is one (1) average. This means that random events as well as periodic events will have the same weight.
One of CSI's techs uses four (4) averages, but this will remove the random impacts.

I think this is an extra waveform that is collected (specified in an AP set?)soley for the purpose of comparison to an extra alarm band that is set up to look at total peak to peak width of that data. I guess if there is a lot of impacting it gives you another way to tip you off about approaching problems. Please correct me if I am wrong on that. My new analyzer will have the peakvue analyzing capability, my older 2115 did not, therefore do I need this extra bit of data?

Special time waveform is a TWF collected in addition to that which is used for spectrum calculation. It can be set with collection parameters different from those defined by spectrum AP. It can be analyzed and trended in the same way as a "normal" one. Its usage does eliminate/substitute usefulness of PeakVue TWF.

The special twf designated in the ap set is used to give you a waveform in acceleration when collecting a spectrum in velocity.

Since PeakVue is collected in acceleration already the use of a special twf is not necessary.

Good Luck,

Al,
for example you might want a longer waveform on the slower stuff, there should be some demo peakvue ap setups in the example database that comes with the software,
respect

As David mentioned, the special waveform has no relation to the stored spectrum, so anyone analyzing your data other than yourself, may not be aware that the waveform they are looking at is not directly related to the spectrum, close maybe, but not directly.
Not to say a special waveform is not good, it is great, especially when you might want to look at an area higher than what the "normal" spectrum-related waveform shows.
Example: The spectrum covers 1000 Hz and you want to see what is happening in a waveform if the spectrum were out to 3000 Hz.

Peakvue is limited to your HP Filter or BP filter, so to say, "having both the special waveform and Peakvue data is not necessary" is not correct, in my opinion.

As Ralph and captnb stated there are instances where you would want to obtain the special twf and I do so frequently, but rarely with PeakVue. I also should have added that when using analog integration, such as SST is when you need it for the reason I stated earlier and that PeakVue uses digital. I think that SST use was the original reason for creating this feature.

Like Ralph said, you won't be looking at the twf from which the spectrum was calculated if you use it.

Good Luck

The special time waveform is a frequency limited waveform in sensor units. If the Fmax is set at 3,00 Hz (as an example)for the special time waveform, the antialiasing filter is set at 3 Khz. If a fault (such as fatiguing or friction) which is represented with high frequency (short duration)events are present (10 + Khz), they will not be readily evident in your special; time waveform. PeakVue does not employ low pass filters (antialiasing filters) and will be equally sensitive to all short duration events (impacting, fatiguing, friction, etc). I do not feel a special time waveform is useful in addition to PeakVue. The PeakVue time waveform saved should always be that from which the PeakVue spectra was computed. The PeakVue time waveform is used for severity assesment and separation of impacting (which are periodic) from friction (generally random).

Jim Robinson, Emerson Process Management

quote:

PeakVue does not employ low pass filters (antialiasing filters) and will be equally sensitive to all short duration events (impacting, fatiguing, friction, etc). I do not feel a special time waveform is useful in addition to PeakVue.

Mr Robinson,

I am not clearly understanding what you are saying here, especially where you say "PeakVue does not employ low pass filters (antialiasing filters)and will be equally sensitive to all short duration events (impacting, fatiguing, friction, etc).".
If a HP filter is at 2K Hz, and the normal non-Peakvue route Fmax is 1K Hz, will Peakvue also see and include all impacts below this HP setting? I was under the understanding that Peakvue pays no atttention to anything below the HP filter. It is true, maybe, that the HP filter should start at the end of the normal non Peakvue Fmax.

I guess maybe, rather than "looking" below the HP Filter, you are saying that Peakvue does not stop at a predefined frequency as does the special time waveform example of 3K Hz, but continues out to and beyond 10 + KHz.

Also, is it an absolute fact that, for example, a 2K Hz HP filtered Peakvue shows us, within a 2K Hz Fmax spectrum or any frequency Fmax as far as that goes, and the associated waveform, all impacts, fatigues, frictions, etc. that it (Peakvue) finds between the 2K Hz HP and

quote:

"10 + KHz"

or to whatever the limit is for the meter is -------- or does it, in reality, only show us what it sees from the HP Filter of 2K Hz to the predefined Fmax, in this example, of 2K Hz, in other words, does it only show us the first 2K Hz of data located between 2K and 4K Hz? If not, how does it determine what to display within the 2K Hz Fmax?

Sorry for the long winded questions.

I can't comment on PeakVue directly, but there are some important considerations when using time wavefroms. If you've ever looked at time waveforms "live" using a spectrum analyzer, you've probably seen this limitation.

Consider a clank from a damaged gear tooth. IF you set Fmax fairly high and look at the time waveform, once in a while you'll see a burst of energy when the clank happens to occur in the sample you're viewing. With most analyzers, you can press hold to have a careful look at the burst of energy generated. That single burst doesn't help you determine the repetition rate to determine which gear has the damaged tooth.

To understand the cause, we need to see 2 bursts on the screen to calculate the timimg. Simple, just lower the Fmax to lengthen the time period. Except when you try this, you can't see the bursts! When you lower Fmax to lengthen the time, you cut off the hgih frequency noise that the damaged tooth generates. Time length and frequency response are not independent.

Demodulation gets around this problem by converting the energy of the clank to a varying DC voltage. Change Fmax to whatever range is needed to see the length of time record. You can set Fmax really low to see a whole string of pulses to check for uniformity.

I believe that PeakVue is somewhat similar. If I understand Robinson's comment correctly, PeakVue switches out the antialiasing filter, with is the reason that the high frequency noise is filtered out when looking at a conventional time waveform.

Jon
Spintelligent Labs

Jon is correct in his representation of PeakVue. If one captures one of the 'clanks' he refers to using a high sampling rate and expands, there will be a fundamental frequency which rings down (usually requires only fractional to a few milliseconds for the 'clank' to disappear). The ring down frequency is somewhat depedent on the type event which initated the event. The ring down frequency will range from around 1 Khz to several Khz. If this raw signal is passed through an antialiasing filter set below the ring down frequency, the 'clank' will not make it through the filter.

In PeakVue, the raw signal is always sampled at a SR corresponding to 40 Khz (or 40,000*2.56=102,400 s/s). The 'Fmax' selected for a PeakVue measurement is used to define time increments equal to the inverse of 2.56*Fmax. As an example, assume Fmax is chosen to be 40 Hz, then the absolute peak value from the digital raw signal corresponding to the 40,000 Hz band over the first 1000 points will be inserted into the first bin of the PeakVue TWF. This process is completed until the PV TWF is filled. The PeakVue 'spectra' is then computed from the PV TWF. A better labeling would perhaps be 'rate' instead of 'frequency' (as pointed out by Jon).

Jim Robinson Emerson Process Management

quote:

assume Fmax is chosen to be 40 Hz, then the absolute peak value from the digital raw signal corresponding to the 40,000 Hz band over the first 1000 points will be inserted into the first bin of the PeakVue TWF.

Mr Robinson,

Are we dividing the 40,000 Hz raw signal band by the 40 Hz Fmax to get the 1000 "points" you mentioned??

Ralph, that is correct.

The sampling rate for the 40,000 Hz bandwidth is (40,000*2.56=102,400 S/S). The time interval over which the absolute peak value is selected for PeakVue with an Fmax of 40 Hz is the inverse of (40*2.56=102.4). Hence divide 40,000 bu 40 to get number of samples in original to extract the absolute peak value from.

Jim Robinson Emerson Process Management

Thanks, Mr. Robinson.

quote:

Hence divide 40,000 bu 40 to get number of samples in original to extract the absolute peak value from.

Does this division mean that the data displayed from the first "1000 points" of the original raw data is coming from the begining of the HP Filter and stopping at 40 Hz above the HP Filter?
In other words, if the HP Filter is 1000 Hz, does the actual data displayed, after the 40K is divided by 40, have to come from between 1000 Hz and 1040 Hz, when a 40 Hz Fmax is used, or does it come from anywhere between 0 and the orignal 40,000 Hz span or where does it actually come from?

Alan, for 'normal' equipment (not low- or high-speed) I am usually set up to take a "special" timewaveform. I use 3000 Hz, Acceleration, 2048 lines. This gets about 90% of all bearing problems. For some machines, I have to kick it up to 4-5 kHz. I trend 'peak' acceleration as a paramater, not 'peak-to-peak'.... my rational for this is that true "impacts" occur in a single direction, not in a peak-peak fashion as mechanical faults do (imbalance, misalignment, etc).

If I detect a low-frequency, mechanical problem when I'm taking data, then I acquire a velocity waveform for further analysis. This is usually a low resolution spectra (200 Hz, 100 lines for example) so I get good waveform resolution at the low end without all the "noise" that may be present at the higher frequencies.

I have tried to use Peakvue, but not had much success. It must work because so many people here swear by it. But I am very seldom looking for "early detection" of bearing faults... my customers usually want to wring every last bit of life out of a machine. Peak Acceleration allows me to do that with a lot of confidence.

quote:

Does this division mean that the data displayed from the first "1000 points" of the original raw data is coming from the begining of the HP Filter and stopping at 40 Hz above the HP Filter?
In other words, if the HP Filter is 1000 Hz, does the actual data displayed, after the 40K is divided by 40, have to come from between 1000 Hz and 1040 Hz, when a 40 Hz Fmax is used, or does it come from anywhere between 0 and the orignal 40,000 Hz span or where does it actually come from?

Ralph,

You should not ask these sensitive questions. No one is going to say that what you are asking is true or false or are they or maybe they can't understand what you are asking or it's too hard to answer or--------.

Looks like you and Rusty agree on the 3K waveform thing. What exactly can it see that Peakvue can't?

Buddy

Rusty, you said that you have not had much success with Peakvue. It seems that I remember far back that you do most of your data collection using a stinger. Do you use that for Peakvue? If so, that may be the reason. Peakvue is much more sensitive to mounting than "normal" data.

Buddy M., Peak Vue is patented technology so it´s open for anybody to read the patent that may be as clear as pea soup or crude oil, getting some more understanding on the idea behind may bring some light on the various behaviour you find using different technology. My experience was that I used PV always with SST and in some cases catched the snake with one or the other method, rarely on both, mainly on slow speed machinery. So this clarification may give some light on these things people find. So thank you Mr Robinson for one more clarifying statement. So if there is no anti aliasing filter at 40KHz you will have aliased signals folding down from any frequency as high as the transducer can provide, and that may not be so much more anyway.
Ralf,
I think one of the key elements in the PV working is the extraction of the
"peak value from the digital raw signal corresponding to the 40,000 Hz band over the first 1000 points will be inserted into the first bin of the PeakVue TWF. This process is completed until the PV TWF is filled."
So it´s not as simple as dividing and taking every 1000nd value but the highest peak within those 1000 values in the 40Hz band. That is only my interpretation and may not be the way it is. Olov