Hello all,

I'm making a research on methods which are in use for condition monitoring of bearings.

I already spend a lot of time in figuring out what is the main principle on which PeakVue method is working. I think I figured out the procedure for envelope method but for PeakVue I didn't find a "recipe" jet.

Acceleration Envelope:
1. Signal from accelerometer
2. Band-pass filtering
3. Hilbert transform
4. Fourier transform (FFT)

PeakVue:
1. Signal from accelerometer
2. High-pass filtering
3. ????? - what is going on in this step??
4. Fourier transform (FFT)

Thank you in advance for any kind of help.

Mateo.

For acceleration envelope, most, but not all, commercial systems don't use hilbert transforms. Instead, the signal is demodulated by rectification, then usually low pass filtered, and then that waveform is sampled at a substantially lower rate than the input waveform. For peakvue, data is grouped in successive blocks of some length that corresponds to the ratio on the input sampling rate to desired sampling rate of the output waveform and the peak value of each group s found and sent to the output. Other systems use simple decimation or else select an average value of each group. The advantage of using an averaged value is that, if done properly, the spectrum of the output can have a much better signal to noise ratio than peak detection or simple decimation; for machines with high background noise signals, this can be the difference between having successful diagnostic measurements and not having them.

Hilbert transforms systems have much greater memory and computational requirements compared to the demodulation by rectification methods and most commercial data collectors can't support hilbert transform methods in a practical manner. Practically, there is no advantage for hilbert transforms over properly done rectification demodulation measurements.

Thank you very much Mr. Duncan for taking time to help me with those methods!

With your help I think I figured out how those two methods works. In attached one file in which I tried to explain those two methods (PeakVue and Acceleration enveloping).

I have to more question:
1. Regarding to acceleration envelope method:
When we get the envelope of the original signal (signal from the accelerometer) the envelope is the same length as original signal. Are there any guidelines on how much lower rate we can sample this envelope?

2. Regarding to PeakVue method:
I have one question about the step in which we brake apart the signal into the time blocks. How long are those time blocks? How many samples we have to take to make one block? Are there any guidelines?

Mateo

Envelope_PeakVue.pdf (138 Kb, 161 downloads) PeakVue and Acceleration enveloping method

Mateo,
The time blocks will be the inverse of the chosen f-max times 2.56. In other words a 100 Hz f-max would be a time block of 100*2.56 or 256 samples per second (.0039 seconds). It would hold the maximum peak in this time block. Peakvue always samples around 100,000 samples per second.

Thank You Jim! This is all I needed to hear.

Thanks for your help!

Best Regards,

Mateo

dear all peakvue practician,

i have have several question about peakvue.
i intended to analyze bearing by using MSA (multi channel signal analyzer) and then take the data into *.txt format to process it in MATLAB (making peakvue program in MATLAB).

actually i have read a lot of information about peakvue, and one of them that make me more clear about this method through *.pdf by Mr Mateo, but i have still confused about that.

my question:
1. it is said in the 3rd step we must split into time block after rectify, but why there is still original signal when we are taking mean or max value (not based on the rectify signal)??

2. if i'm not mistaken (if i use MATLAB especially using 'max' command for taking the max value),it means that the data will be reduced, just the max in the certain split box only and how can i plot the graph in the same area like the original one?

3. actually, i have read too some paper called "682A05 bearing fault detector" and the paper said that it must taken into log amplifier first after full wave rectify. is it true/ the same like Mr Mateo *.pdf???

4. actually, what kind of information that we want to get from peakvue???

a. is it one value that show the vibration level [Gs], suppose when we want to identify BPFI freq of bearing and compare it to the recommended limit like in Peakvue conference paper?

b. or we just see both the time and frequency domain like other vibration data to look at that peak at certain frequency??

which one is true???

can someone help to get out from this confuseness please???

The description of PeakVue by Mateo is not correct.
PeakVue is patented.
What is your objective, reason, or purpose for making a demodulation or PeakVue analysis?

Walt

The following is from an Emerson site.

quote:

Emerson Process Management, PlantWeb, AMS, DeltaV, Ovation, PeakVue, CSI and Machinery Health are
marks of one of the Emerson Process management family of companies. The Emerson logo is a trademark
and service mark of Emerson Electric Co. All other marks are the property of their respective owners.

Other places say that PeakVue may use patented methods. It looks like PeakVue is a marketing symbol.
------

Reading the above, I could not decipher the intent if there was or is any.

The word PeakVue may not have technical meaning. From what I read or the way I interpreted it, the signal processing that Emerson labels PeakVue could change, becasue PeakVue is just a name (for marketing). To my understanding one does not know for certain that a PeakVue implementation on one product today will be the same as another implementation in the future.

If one says Fourier Transform (with settings) one should (and generally do) know what you get.

Emerson should have no reason to divulge all the specifics. They would consider that to their comercial advantage.

This message has been edited. Last edited by: William_C._Foiles,

The PCB/IMI vibration transmitter that uses PeakVue method claims to be patented:

THE MODEL 682A05 BEARING FAULT DETECTOR
(U.S. Patent No. 6,889,553)
A New Approach for Predicting Catastrophic Machine Failure
by
James Robinson – Emerson Process Management
Mitchell Illig – IMI Sensors
Thomas Brown – NSK Corporation
Ray Limburg – PCB Piezotronics

The above title is from White Paper describing the new product.

Has anyone installed and used this product??

Walt

Here are some links to detailed technical information on the PeakVue method:

actual PeakVue U.S. Patent #5,895,857
All of the technical details of PeakVue are divulged in the above patent (if you can follow the legal wording of the claims).

Description of PeakVue - CSI paper
Good comparison between demodulation and PeakVue.

Walt,

Here is the patent for the PCB/IMI PeakVue transmitter you mentioned:

Method and apparatus for vibration sensing and analysis - U.S. Patent #6,889,553

Regards,
Rob Laschinger
Planet Instruments, Inc.

actually, i just want to prove peakvue "pure" for education purpose only.

the one that i "most" do not understand is why we must use High Pass Filter??

i have a problem to detect bearing which rotates at 900 rpm (15 Hz) and known BPFI = 90 Hz.
according to the recommended measurement, i must use high pass filter at 1000 Hz for using peakvue and also Fmax recommended is 4*BPFI = 360 Hz because i want to detect the vibration level of bearing with BPFI using peakvue.

of course it will be sample at 2.56*Fmax = 921.6 Hz. and i think for all vibration common analysis, it is strange.

i said it is "strange" because the data that will be show from 0 - 360 Hz (or say until 400 Hz) and what will we get if we apply HPF at 1000 Hz? (of course all of vibration data before 1000 Hz will be swept out)

i dont know if my interpretation about peakvue is wrong.
is there anyone can help me how to solve this or give some explanation/recommendation about the measurement???

thank you

Read the article, especially the first few paragraphs, at:

http://vibrotek.com/article.php?article=articles/dlcvi93/dlcvi93.htm

There are other bearing related articles linked to:

http://vibrotek.com/articles_bearings.php

These articles don't talk about PeakVue but the information they contain is relevant.

hide,
You said,
"i said it is "strange" because the data that will be show from 0 - 360 Hz (or say until 400 Hz) and what will we get if we apply HPF at 1000 Hz? (of course all of vibration data before 1000 Hz will be swept out)"
I know it sounds like you are filtering out everything in your spectrum and will end up with a blank graph, but you won't. The attached document is a 900 rpm bearing like you mentioned. I am using the 1,000 Hz high pass filter you mentioned, but I am using a 1,000 Hz fmax instead of your 360 Hz fmax. The top plot is a "regular" reading and the bottom plot is the PeakVue data.

David Eason

900_rpm.doc (96 Kb, 64 downloads)

Mr David Eason,
may i conclude from your experiment in *.doc file that it is OK if i use Fmax in 1000Hz for showing in frequency domain instead of peakvue recommended in 360 Hz (or more,suppose in 1500 Hz maybe for Fmax)?

i really concern to HPF because i want to make some program in MATLAB and i stuck with HPF setting when i must decide nyquist frequency

f nyq = (f'cut off' HPF/(f sampling/2)) and the result must between 0 and 1

so, in order to use HPF 1000 Hz, the minimum value for the denominator has to or equal with 1000 Hz,right?

and for the result of that,can i regard it as a common vibration data,except the usage of this more refer to time waveform data rather than frequency domain data and use some logarithmic amplifier to amplify the signal??

i think peakvue is powerful because when i read your data,if we compare the upper plot and lower plot spesific in "Route Spectrum", the result is totally different. i really curious why it happen? (esp. when i see the range from 0 - 1000 Hz, but you make it in 2000 Hz for the upper plot,right)

i really want to know the reason because i want to modify the data like the upper one in order to make it like the bottom.

thank you

This message has been edited. Last edited by: hide,

hide,
Yes, my upper plot, which is my "regular" data has an fmax of 2,000 Hz. I have it set that high because I use that position to look for motor rotor bar problems. For PeakVue, I would use an fmax of 360 Hz only if my running speed was 200-400 rpm. Let me stop and say that this applies to most "common" bearings. You would want a higher fmax if you were checking a bearing with a large number of rollers in it. One of the rules of PeakVue is that the filter has to be equal to or higher than the fmax. This means that if you change your fmax from 1,000 Hz to 1,500 Hz, the filter will move higher also.

David Eason

Now I'm little bit worried and for this are two reasons.

First reason:
I did not want to steal the patent, if thats how things looks.
I was just making the research on methods which are in practice most commenly used for fault detection and condition monitoring of bearings. (for educational purpose)

Second reason:
What is wrong in my description of envelope method and PeakVue method. Because I don't wont to spread a "lie" around on this methods.

Please help! Thank you in advance for any kind of help.

Best Regards,
Mateo

In my experience and feel free to correct me, When I see bearing defect harmonics in the gSE spectrum, gSE being the Entek version of HFD, I can quickly take an additional acceleration spectrum with an Fmax of 5kHz or 10kHz, 5kHz being the high pass cut off for the gSE filter, and most often can see the bearing harmonics residing way out there. So, based on this, I believe that the HFD or Peakvue or gSE or whatever you want to call it is simply a simplification method to bring the defect harmonics which are "ringing" the bearing down to their corresponding frequency in the spectrum. So, a cut and paste of harmonics so to speak.
I am not saying that it doesn't work because it works very well but if I took acceleration readings on all my bearings with a high enough fmax, I would see the harmonics anyway.

Where I have found gSE to be extremely valuable is in detecting insufficient lubrication. Whereas there are no specific defect frequencies but the the overall level is trending up or high. I have had levels 10g's (gSE) on a bearing with a raised noise floor, added grease and see the level drop to .5 instantly. To me that is a very dramatic and useful bit of information.

quote:

Originally posted by David Eason:
One of the rules of PeakVue is that the filter has to be equal to or higher than the fmax.

That is true for PeakVue. Is it done in order to avoid aliasing?

Rob,

Thanks for posting the CSI patents. The PeakVue patent references another CSI patent #5,249,139 for an analog preprocessor for demodulation signal processing. I have one of these called the CSI Model 750 that fit on the back of the old CSI 2110 Analyzer. When PeakVue came out, I compared several measurements with the 750 and did not find a great difference. In fairness to PeakVue, I probably didn't set the PeakVue parameters in an optimum way, and I certainly did not try a wide range of measurement conditions for comparison. Now I exclusively use PeakVue and have on few projects made demodulation measurements with Dasylab software.

In my opinion there are few people that are using PeakVue at all and of those fewer use it (setup and interpretation) correctly. An even fewer number of folks actually understand the PeakVue signal process.

Good luck trying to simulate the PeakVue signal process! It is possible to demodualate vibration data with Dasylab, Labview or possibly Matlab software. I suggest trying one of these programs to learn demodulation before trying to program the more difficult PeakVue process. Quick quiz: do you know what signal rectification is and how to implement it with a digital signal process?

Walt