Just attended a Into to Predictive Maintenance Technologies Course and what the instructor said on setups has me scratching my head a bit. Like Fmin and Fmax. I was under the impression from a prev vibration course that a good rule of thumb for Fmax, is 40-50 times running speed of the equipment that your wanting to monitor. Lines of res of atleast 800, avg of 4, and overlap of 50%. From what the instructor said, he req a Fmax of 10 times running speed, 800 lines, 1-2 avgs and 0% overlap. I'm in the "early" stages of setting up specific equipment that will be monitored and entered into the data base. I have been using a few specific equipment/machine trains to test out various setups. But again, I having a hard time deciding what is good and what it not. (Esp after this last course) Is too much good or bad or is too little good or bad. Here or some sample machine trains that I have been experimenting with. Example (A). Circulation Fan. 2Hp motor, 1760 rpm. Belt drive to the Fan shaft. Example (B). Blower. Motor 1780 rpm, Belt drive to Blower 1693 rpm. Blower is a positive displacement Gardner Denver 2 lobe blower. Example (C). Mill roll. 100 Hp 1180 Rpm motor. Driving a Gearbox 5.38:1 ratio. Output of gearbox is driving the Mill roll. I grabbed rpm readings of each component. Motor rpm is 735 rpm, Gearbox input is 550 rpm. Using the gb ratio, I calculated the output rpm as 102 rpm. I know that roll speed can change depending on output speed desired, but from what I have gathered, the fasted would be 105 rpm and the slowest around 90 rpm. So for Example (C), my thinking is that the motor is a 1180 rpm (np) but operated at 735 rpm. Is it best to use the 735 rpm ( +- say 10%) and set a Fmax at 50X? Or is this too high? If a Fmax is too high or too low, what would I be missing? Same ? for the gearbox and Roll. Since each component is operating at a different rpm, each Fmax would be different? So for a quick recap. Should one base Fmax on motor name plate (or actual running speed) times 50 or a lower value (X)? The lines of res, I have a understanding on.. ie more lines being clearer or more to see. If Avg are lowered, what is lost? If Overlap is too high/low, what is lost?

Thanks in advanced,
X

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

XRACER
If you know all the technical details for example (you’ll be lucky)

The bearing manufactures and reference numbers
The number of pump impeller vanes
The number of fan blades
The number of rotor bars
The number of rotor bars
The number of gearbox gears and gear teeth
The number of teeth/length of drive belts

You then need to assign a collection spec that will capture data that will include all the fundamental fault frequencies along with any harmonics that may be present if a fault situation occurs

Gearboxes and high speed compressors need particular attention and Technical associates of Charlotte offer some very good advice
3.25 times the gear mesh frequency if the tooth count is known or 200x the shaft speed if unknown

You may have a lot of machines to monitor most of them will hopefully show no signs of problems so using a Fmax of 60x rotational 400/800 lines 4 averages 50% overlap might be adequate

On the equipment that is showing sings of problems use high resolution if it is necessary i.e. if you spot something untoward on your collector go off route and take the appropriate action, high resolution time waveform etc however collecting data using a higher resolution will take more time

A colleague I knew used the following collection spec on most equipment excluding gearbox and compressors
Velocity Fmax 30k Cpm 1600 lines of resolution, 4 averages 50% overlap
Acceleration Fmax 120k Cpm 800 lines of resolution, 4 averages 50% overlap

His reasoning
Velocity Fmax 30k used to determine the presence of non integer of rotational speed activity; “fundamental bearing fault frequencies and harmonics”

Acceleration Fmax 120k used to determine the presence of various higher fault frequencies and harmonics “lubrication rotor bar, cavitation”
He felt he could monitor the equipment adequately using these specs

In my opinion it’s all down to training and understanding how the equipment that is going to monitored works, whatever parameter set up you use has to do the job; as for variable speed base your Fmax on the worst case scenario but also add a suitable lower Fmax and collect if necessary.

If the Fmax is too low you may miss vital high frequency data “lubrication issues, rotor bar, cavitation gear mesh frequencies and harmonics”

If the Fmax is too high and you don’t increase you lines of resolution you may miss vital low frequency data “fundamental bearing fault frequencies and harmonics, line frequency etc”

It is my belief that if the number of averages, are lowered you may miss vital data
If the overlap is too high the time to collect your data will take longer if is to low then vital data will be missed

I have shown the above information on collection specs and the reasons purely to illustrate that my colleague was at ease using them; they are general collection specs and are NOT to be taken as gospel

Hopefully someone else on the forum might offer some further information on signal processing

I hope this makes sense and helps, if I have made a mistake it is not intentional and I’m sure someone will put it straight

Mike.

Thank you for the reply. I'm still in the early stages but the wealth of information that you provided will help me onto the right path. I have been testing out different setups on different equipment, but I believe my problem is that I'm trying to do it all at once. Trying to find that magic setting the will work for all, but know realizing that I should be focusing on one type at a time, and get it complete before moving onto the next type of equipment. My biggest fear, is missing something that could have been caught if setup properly. Anyway, thanks again for the info.. I know that I will be post more with many questions.

Good advices so far, may I add mine, this is what I use to do:
Start with acceleration, hanning average = 1*, max. freq. selectable (20 khz), max. number of lines (3200 or 6400), minimum freq. = 0 hz, trigger = free run
Reduce max. freq. to capture all frequency components present. So do not stick to any x .. y times rpm rule, listen to the machine to determine the frequency span of interest. New max. freq. / new number of lines preferably < 1.
Look for amplitude modulations. If present then this determines your filter when enveloping (demod = accel, hanning). Often 2 khz to 10 khz satisfies to capture bearing defect noise. When electrical machines involved this > 2 khz filter setting will often exclude the 2*fline magnetic modulations that have nothing to do with a mechanical or electrical defect.
Store acceleration with averager = 4* lin.
Store demod spectrum with averager = 4* lin, max. freq. 20 * rpm will capture ball bearing and roller bearing defect frequencies and one or two multiples thereoff.
Store time wave form, accel, over a time span that at least captures a couple of revolutions with max number of samples.
Store 0 .. 1000 hz velocity spectrum.

These set ups will provide good optimum for type of data (accel, vel, time, demod), memory load (not too much), measurement time (not too long).

With the sensor mounted on a machine frame by means of a magnet you will often find a ‘haystack’ of frequencies around 2 khz or 3 khz. This is a result of resonance of sensor/magnet and its non-stiff coupling to frame. This frequency area contains enhanced bearing defect energy however this makes the measurement less reproducible.

Practical experience is best learning curve so exercise as much as you can and experiment with alternative settings to learn what you miss or win!

Good luck!
Arie Mol