Dear Colleagues:

Could anybody qualifiedly recommend the best software for professional deriving of trends in vibration spectra? I would list the required key features as following:
1. Produce spectra from subsequently recorded waveforms and display those spectra in a waterfall screen, with the recoding date for each spectrum.
2. Operator (or software) would mark a particular peak (which may be not of exactly the same frequency in each spectrum) in each spectrum.
3. Tips of selected peaks are connected with a trend line -- manually or automatically.
4. Trend line is displayed in a Magnitude-vs-Time plot.

Of course it would be very nice to also have all other bells and whistles in the software, like enveloping, bearing- and gearbox frequencies calculator, order tracking and extraction, tachometer processing, harmonic- and sideband cursors, orbit, Bode plot, polar plot -- etc. but the main thing is a good trending tool.

Please do advise, because I am a bit lost among glossy brochures and shouts about "revolutionary" stuff.

Thank you very much.

Hi Bobcrad,

I'll take a bite at this. Please afford me a "reading between the lines", for as much as is possible from reading your post, rather than endless questions.

1. So you would like to replace or bypass an expensive spectrum analyzer with an MP3 device or the like in order to record time waveforms which you then import into software which performs the FFT necessary to produce spectra. In turn displaying these spectra on a waterfall plot with associated timestamp.
2. Behavior you describe might be summarized as needing an order based approach - is your machinery variable speed like so many paper machines in kiwi land? "(which may not be of exactly the same frequency in each spectrum)" i.e could it be same order value from spectrum to spectrum??? but different Hz? Your initial waveform would need knowledge of the rotational speed of shaft at the time of recording to do this succesfully. Otherwise, how would a spectral peak (or software) know that from one survey to another the different Hertz values are due to the same source?? and hence connect them.
3. See 2 above, if my suspicions are correct, an order based approach may be needed. There are some 3 letter brand software that can do this (draw a line from peak to peak in a spectrum on a waterfall plot - but all need their respective analyzers), provided the waveform is FFT'd with knowledge of original running speed - but I do not know of simple tools to do this, only the usual suspects, and not all of these exactly as you request.
4. Are you talking of possibly a different trend line for each of the selected peak trends above? Again some can do but not all of the well known and they may not in a simple trend plot, instead overlaying the connected line on the waterfall plot. But if you do have variable speed, how are the amplitudes effected by naturally occuring resonant frequency ranges?? which could lead to false reasoning of "it has increased" or "decreased".

For some reason I get the feeling you are after a simple tool (read low $$). To develop such a tool, why would soft/hardware manufacturers confine it to "simple" when so much groundwork to get that far would mean the bells and whistles are not much extra effort.

And think of the extra you could achieve with those "bells and whistles" the vast extra info about the nature of faults compared to fault judgements made on increasing lines.

More information is required about your application to see if the usual suspects are what you need or simpler methods can add something, if that is what budget allows. Consider the cost of having someone "running routes" to collect basic info. For only a little more effort they could collect so much more. Are you thinking of using existing operators to do more - in preference to vibration specialists/technicians running the same routes as well??

Be careful of the glossies, they will promise the world and you will only find out short comings after you have received the product! Speak to someone in your industry who has substantial experience with this sort of work, whether that be a colleague or someone else you can trust.

With info you provided, I suspect you will be after the usual brands, but there may be someone who has something I am unaware of. Good luck!!
p.s. the revolution hasn't happened yet!!

Hi BobCrad, the CSI AMS software is able to do this. However the line through the waterfall is through the same frequency and not manually selectable on each spectra. If your data has any variability in the machine speed then you will obviously as previously mentioned have an error. You can however elect to display in terms of order and as long as you have carefully selected the 1x the corresponding peak will line up. This software has the advantage that it automatically trends up to 12 user definable frequency ranges. I have attached an example, a bit crude but you can see the result. The main disadvantage is this package does not come cheap. I am based in NZ so if you want to discuss this further or would like to visit email me at simon.hurricks@genesisenergy.co.nz

Regards

MDE

Waterfalltrend.pdf (138 Kb, 7 downloads)

Dear Vibe-Rater,

Thank you for your detailed reply.

quote:

Originally posted by Vibe-Rater:
Please afford me a "reading between the lines", for as much as is possible from reading your post, rather than endless questions.

Without any irony, this your effort is very kind, but excessive: I would be delighted to straightforwardly answer your any questions.

quote:

Originally posted by Vibe-Rater:
...to replace or bypass an expensive spectrum analyzer with an MP3 device or the like in order to record time waveforms which you then import into software which performs the FFT necessary to produce spectra.

Yes and no. We are contracted to monitor few large industrial Sites where each machine already has one or several two-axis solidly mounted accelerometers, each already connected to its individual logger which periodically records the signals into PCM WAV (non-compressed) stereo files and can store them for a few months until the on-board memory is full. By default, 20-second records are made each 8 hours. The Sites are totalled to 300 machines. From time to time, Operator collects files to his laptop for machinery condition monitoring.

We are restricted to this existing data collection system.

We were also restricted to WAV waveform file format, but recently they allowed our access to the firmware. We may now implement any codec into their recorders and use any format. It is still undesirable, because the huge logged arrays have to be converted and replaced in the existing database. Frankly, we would prefer to tolerate drawbacks of WAV, if lucky to find a suitable WAV-compatible software. But by now, we do not know a WAV-compatible software, that can expeditiously build trends for selected peak(s). Several WAV-compatible programs we tried, cannot derive a trend, cannot envelope the signal, build an orbit, derive orders and correlation, etc.

So we decided to re-start our search from the most important feature: trending. Then we will try to match the formats. Then, if we still have more then one choice, we would sort-out other features. Or, if not, we would write our own code using LabView "bricks". But I hate to think about the latter, because we do not have a 30-year experience in this field and I am not sure if we know WHAT to write. Hence I am here.

quote:

Originally posted by Vibe-Rater:
...variable speed like so many paper machines in kiwi land? "(which may not be of exactly the same frequency in each spectrum)" i.e could it be same order value from spectrum to spectrum but different Hz?

The sites are abroad NZ. Variables gain their popularity, but asynchronous inductive (squirrel-cage) motors still dominate. Their RPM slightly vary with load. That is why, a spectrum array should not be sliced at a strict frequency, but rather linked to peaks within some narrow band, or order based spectra is to be used. Sometimes we do not want order based spectra because we want to count for amended load.

quote:

Originally posted by Vibe-Rater:
And think of the extra you could achieve with those "bells and whistles" the vast extra info about the nature of faults compared to fault judgements made on increasing lines.

We are a hot supporter of all "bells and whistles" and know their benefits. We will also supplement Client's system with ultrasound (acoustic emission) warning tools. But if advanced features are not in the Vibration Analysis Software we chose to our major requirements, we will survive.

quote:

Originally posted by Vibe-Rater:
Consider the cost of having someone "running routes" to collect basic info. For only a little more effort they could collect so much more.

Let me note that logging the waveforms means databasing of quite comprehensive info. Nothing precludes you then of using all advanced post-processing technologies.

And our Client was not too stupid, solidly installing wideband accelerometers. At least their reproducibility of measurements is superior.

Also yes, we do not believe in glossy papers and "revolutions". That is why, we are here rather than in noisy traders' offices.

quote:

Originally posted by Vibe-Rater
With info you provided, I suspect you will be after the usual brands...

Yes, if no other choice, but we are very open-minded and would be glad to accept any novice's suitable software.

If somebody has a reasonable outlook in existing software, please advise.

Thank you very much.

Bob Crad

Bob, I have a software that is not really there but in a way not so far. I enter raw data from my box, not yet .wav but it may be possible, trend but only ISO, unfilt, Brg, Temp, RPM etc. currently. It converts raw data and display FFT with marker and zoom. Working on swapping graphics to be able to plot waterfall. It has a standard database to put imported data to. Email a snail mail address and I can send a CD to look at. Olov

Hi BobCrad,

Years ago we used a soft package called goldwave to play a wav file to a spectrum analyser - but that approach does not eliminate an analyser. See http://www.goldwave.com/ Try that it's only US$45.
Rather than taking waveform from a system such as yours, we powered an accel and fed signal into goldwave on a laptop to record. Then emailed elsewhere and played back using same soft to fft analyser. In the end we gave up since we could not control line in levels (amplitudes) on laptop but from a good soundcard it was possible - but that was impractical to carry around. Regards.

This message has been edited. Last edited by: Vibe-Rater,