OK, I’ll take up the challenge and start this new thread.

The “2130 glitch” thread, Ralph and Rusty comments why they don’t like order based Fmax’s and IanS - I know where your wife shops ;-) - intrigue as to why this approach is not liked, prompted this. – I am equally intrigued….deserves a new thread IMHO.

What are we trying to see? I suggest it is a minimum of cage frequency. Typically, between 0.38- 0.47 orders.

To resolve cage frequency you need at least 70 orders, 800 lines, which equates to just below 12 lines per order. OK, this lines per order resolution does not quite resolve cage frequency clearly but if a rolling element defect exists, it will show in normal acceleration spectrum as 2x BSF plus cage sidebands which appears as little “haystacks” or “humps” with centres at 2x BSF. Clearly identifiable – visually – and I am a supporter of visual analysis.

1600 lines will see more clearly but we trade this off against added data collection time, in the light of routine surveys.

This holds true for machines running at all speeds. Regardless of 60 Hz. US, 50 Hz. the rest of world – most of, and particularly variable speed or for shafts at other speeds, intermediate gearbox shafts, output shafts, fan shafts / pulley reductions, again particularly variable speed, not to metion 6, 8 and greater poles for each LF.

In my experience 70x 800 lines is a great starting point, for 80% of machines, gearboxes higher orders, say 100 or even slightly more and a minimum of 1600 lines.

Why, I hear you ask. Because a bearing shows its true colours, fault frequencies, order based. They do not care about Fmax in Hz.

It is much easier, for an 8 rolling element 63** series bearing, to remember very close to the number 3.08x…(and inner race 4.92x).than remember 76.69 Hz. 50 Hz line freq. 4 pole, 89.64 hz.,60 hz lf, 4 pole 60 Hz, 153.69 hz. 2 pole 50 Hz., 181.7 hz. 2 pole 60 hz LF. Or other values if slip is different. Or a myriad of other numbers you care to dream of when looking at complex gearboxes, particularly variable speed.

Please contribute answers- Cat among pidgeons…..!?

This is all similar for the various bearing types with common diameter and width series commonly encountered.

To me, 5.69x means a 22312 14 rolling element bearing. Roughly same for 22315 or 22318 etc..(add or subtract cage frequency for an additional or 1 or more less rolling element will get you those fault frequencies) 18.26x immediately points to a Timken tapered roller with 39 rolling elements or thereabouts, and almost all other bearings in common use have families of fault frequencies that fall in associated regions on an order based spectrum.

What if you use an Fmax in hertz?

You are keeping Hewlett Packard in business through buying calculators and do not build an in-brain fault frequency database and associated “gut feel” for whether the issue is bearing or not. You are not using your software / hardware to best advantage.

I hear the Hz and CPM people cringe…….with so many years of experience in the way they have always done it……..(red rag to a bull??)

When you look at a paper mill, if one day it runs 500 metres per minute, 900 the next, your Fmax should alter accordingly (nearly double) to maintain, the fundamental need to resolve the minimum frequency of interest – cage frequency. Otherwise you are facing different lines per hertz every time.

I know the above example is extreme but it does happen.

The fundamental need exists. Separate synchronous from non-synchronous. Order based analysis gives this to you immediately; it takes care of the calculator. 3x in a pump to me means vanepass, 13x in a gearbox input shaft means input gearmesh, 20x in a vacuum pump means vanepass, 93x on output shaft is gearmesh. All this cannot be bearing in vast majority of cases. And we are using orders here.

The Hz’ers, CPM’rs do this instinctively with calculators, order based people use their software………….I hear those cringes…..

Of course, Hz. Has its place, resonances, line frequency, DC drive 300 or 360 hz, etc. etc. but overall orders rule.

Think of when you are looking at a complex gearbox. Will the client be happy if the bearing and gear info are not available? – no. you will still need to resolve their problem. Do I have a bearing issue or not! Think of a machine that has had wrong bearing info entered, or bearing with a bearing with different number of rolling elements…the huge different numbers of frequencies involved is overwhelming. Not in orders.

Understanding fault frequency families for different bearing types through order based analysis will allow you to do this, more simply and instinctively.

And you will never forget 3.08x, 11.24x, 18.26x, 9.69x, 12.69x and the list goes on.

Sad but true….

Look forward to hopefully a flurry of responses which no doubt will spawn a whole lot more.

You need to make sure your speed is correct, and know how to check it is correct from spectra, another thread…

Regards, John.

John,
I'm not sure if you are talking about an order based fmax while collecting data, or while viewing data that is already collected, or both. I'll address the part about collecting data. We used order based AP sets when we had single channel CSI 2120s. We upgraded to the dual channel 2120 and found out quickly that the order based fmax was bad for us. A speed change will bump up your fmax and knock it off of the simultaneous data collection and revert back to consecutive data collection. The simultaneous collection sets were saving us 30-40% on our route collection times and we couldn't afford getting bumped out of that. I now have a 2130 which does not have that restriction.

David Eason

Hertz is the standard as is Seconds with time. Why is there a need to re-invent the wheel? Why do some want their own language and convention of doing things? What's wrong with normal? Why are there so many negative comments on this board against Nspectra? And still there are a lot of CSI users including me. Two times running speed equals the first harmonic and widely used in acoustics. I hate cpm as you have to divide everything by 60 and deall with very large numbers so it isn't obvious until you get it to standards units. Or, maybe it's a personal thing.

Hz, rpm, or orders? Well, you use all 3 of course, depending on what you are looking at.

Quick.. how fast is 13.67 Hz? Heck, I don't know... but I have a good feel for 820 rpm. And hey, what is that promiment peak at 5477 rpm (91.28 Hz)? Stumped again, except I know that it's 3.07 orders (sure, I took hi-rez data to resolve it) so it's BPFO of the 6313 bearing common on many motors. Gosh, what is that peak out there at 21,600 rpm... it's 38.3 orders of my 564 rpm motor speed (9.4 Hz) but what is it? Oh, it's the 360 Hz "buzz" common to many DC motors (regardless of the turning speed).

So, properly, you use all three units.

As for orders, the bearing "noise" that often shows up due to bearing damage or lack of lubrication is actually the bearing components "ringing" at their natural frequencies as the bearing defects are "impacted." These natural frequencies (not to be confused with the mechanical defect frequencies) are indepent of the shaft turning speed. If you set your data collection for 70 orders, then on a 685 rpm fan your Fmax will be 800 Hz (isn't that easier to wrap your mind around than 48000 cpm ?) and you will not "see" the bearing "noise" that is often generated.

I see distinct bearing defect frequencies on maybe 20% of the bad bearings I find. I use the appearance of bearing noise (typically from 1000 - 3000 Hz) along with the "maximum peak acceleration" value of the acceleration time waveform to detect and trend significant bearing problems.

This is just one example of how order-based Fmax will let you down. BTW, I would never use less than 1600 lines... the small savings in time is just not worth it.

As for the Fmin, 0.35 orders is probably good enough except on belt-driven machines. You want to be able to "see" the belt 'turning speed'... most belt problems show up as 2x belt rpm, but you want to be able to see the 1x belt rpm so you can better verify the 2x belt rpm peak. BTW, CSI's Speedvue sensor makes determining belt rpm "child's play".

quote:

I hate cpm as you have to divide everything by 60 and deall with very large numbers

Sam, don't the very small numbers you get when you analyze mechanical defects using acceleration also bother you?

Another reason that CSI users tend away from setting fmax in orders for variable speed equipment is the effect of having discrete fmax settings especially for PeakVue.

An example would be a 4 pole motor on a vfd. The first time it is operating at say 53 hz which translates to about 1580 rpm. Using 70.5 orders would make the fmax 111390 cpm which would default to the nearest discrete fmax which is 120000 cpm. PeakVue would apply a 2000 hz hp filter. The second time it is operating at 60 hz or about 1790 rpm. 1790 * 70.5=126915 cpm. The fmax for a velocity reading would go to 150000 cpm but for a PeakVue reading it would go to 300000 cpm and change to a 5000 hz hp filter and you would very be taking readings on a critical piece of equipment with a failing bearing and miss it and get fired

Well it might not turn out that bad, but I think you get the idea. It is not just a preference for one way of doing things that drives these decisions.

Danny,
Why don't you just set your peakvue for it's own reading with a fixed fmax and filter? I know in Odyssey, even when all of the readings are taken with a single button push , the spike energy readings have their own filter and their own collection spec.

Rusty, small numbers never bother me; it's the big one's that do! Somehow, I'm not making sense out of what you're saying? An accel producing a voltage signal translates to an EU. You loose so much by integrating, your PdM program will suffer greatly IMO. You can't analyze what you can't see - a picture is worth a 1000 words. Bearing analyses is so much easier with acceleration why bother with integration then process and reprocess into various formats because velocity is inadequate --- consider it! Your velocity is not good enough so you process in peakvue and, and.

Ron,

I do. Is saving a push of the button really that big a deal? It doesn't seem to save any time, in fact it seems to take lots more. Maybe the anticipation of what your spectra will look like is the appeal.

Personally I use units of Hz, but, having said that I come from a generation that used a sliderule for many years before the calculator came along. Because of that expertise, and the fact that I've been doing this for close to 40 years, I don't find it difficult to rough out in my head the various conversions. "Head" work was very common in the days of the slide rule; I can remember many a meeting where a gearbox design was roughed out in conversation and later verifed by a maniframe computer.

It would be interesting to take the poll and plot the results against age/experience. I'm betting that the young folk out there prefer the units that take the least amount of "head" work.

John from PA

Any takers for pole-pass sidebands in Hz?

Now that is simply masochistic!

I, like Rusty, use all of them depending what I am doing. Most of the software makes it a simple click of the mouse anyway.

That being said, my standard is setup in Hz. I am almost 50 so I don't know where that would put me in your poll John. But my head is filled with enough stuff so the least amount of "head" work the better.