Utilizing Vibration Analysis to Detect Gearbox Faults

hello and welcome to this presentation on utilizing vibration analysis to detect gearbox faults my name is Jason Tranter the founder and managing director of Mobius Institute I hope you enjoy this presentation in this presentation we're going to take a look at a quite a number of things actually and it is a little bit long because there are so many things to talk about but I hope you find it useful first we're going to look at just the basic question of why gearboxes are more difficult to monitor with vibration analysis than other components like pumps or fans and so on so there are some issues related to the designs of the gears and the construction of the gearbox itself and just making sure that you get vibration readings that tell you what you want to know but also techniques that you can use to make the vibration in his less complicated for a more complicated gearbox hopefully that'll become clearer as we go through the next question we want to look at is just to understand some of the failure modes and how vibration patterns will change now this isn't a detailed discussion of all the failure modes of gears and so on it's really just looking at some of the more common fault condition and how the vibration will change we'll look at a little bit of time waveform analysis and also how the spectra will change and finally how to use spectrum analysis and time waveform analysis including techniques like circle plots and time synchronous averaging as well as enveloping techniques to detect the faults and because of the time constraints we don't talk much about the enveloping techniques but anyway let's let's get on with it so let's first look at the challenge that you opposed with when looking at gearboxes there really are a number of challenges and I've not listed them in any particular order but often gearboxes are very complex but critical components in your plant and the challenge therefore is to make sure you get it right because the ramifications of failure can be quiet vier so you know you got to look at the criticality of your all your equipment and make sure you're able to perform the diagnosis correctly perform your measurements and diagnosis correctly but certainly gearboxes can be extremely critical and therefore it makes it even more important that you get it right the vibration from the gearbox can be quite complicated because the gearboxes themselves can be quite complicated now of course you've got to think about the gearboxes in your plant and you know how complicated they're but you can see in this case that we've got you know these rolling element bearings and of course you may have sleeve or journal bearings but you've got all these mating gears here that all generate their own sources of vibration and you want to be able to detect if there's any problem you know on any one of these gears you've got to consider the forces that will be generated you know radial forces and axial forces you've got to think about how the vibration is going to be transmitted from a mating pair of gears to the point where you are taking your vibration readings we in addition to you know taking the measurements correctly we need to be in a position to detect various kinds of faults you know the more severe faults such as you know tooth damage or there could be multiple tooth teeth that are damaged there could be anything from cracks to complete loss of metal as in the case here but we are also looking for increasing where we're looking for misalignment in the gearbox backlash tooth load and one that I haven't listed up there's a centricity where this run-out or actual eccentricity or for whatever reason maybe we're in the bearing which is moved shaft or but in any case we're looking for those different sources of vibration that indicate these different types of faults and we've got time a form analysis and spectrum analysis at our disposal and you want to make sure that you use both techniques that is probably the most important message out of this you need to do time waveform analysis and I know I mentioned it later but you should also be performing oil analysis to make sure that lubricant itself is good and where particle analysis to see if there are any signs of where you know let's move along you've also got the complication that a gearbox can have you know multiple shafts that are turning at different speeds so on the one hand you may have a shaft that's turning at a relatively high speed in a normal speed that you may be used to dealing with but depending on the gearbox the output speed might be a very high frequency in which case you're dealing with limitations of transducer response and some because it's a very high frequency on the other hand you may be dealing with a very low frequency a matter of a few rpm and that creates its own challenges when performing vibration analysis now in this presentation I mean that's why we run training courses you know we can't cover absolutely everything there are challenges associated with very high speeds low speeds there's details that we could get into regarding modulation and time synchronous averaging in these various techniques that you can use how to detect bearing faults which of course you'll find in the gearboxes as well as your faults themselves so anyway we'll focus mostly on gears and and mostly on just normal speed years anyway so the other thing you have to consider is just what is the design in the in the gearbox therefore what teeth are meshing with each other what speeds will you see what frequencies do you expect to detect and therefore how do you set up your spectral measurements and your time waveform measurements in order to capture the data you need in this case we've got you know a bevel gear which creates sort of more actual vibration and a spur gear which creates more radial vibration and you know we've got certain measurement points that we can use so we just need to be sensitive to what the forces are like and therefore you may in this instance taking an axial reading as well as a radial reading but of course you've got to look at you know what mounting points you have available how thick is the metal are you going to be dealing with resonance some people actually use the potential of resonance to help them because they figure well if the case resonates it's going to amplify vibration that's an interesting approach as long as you've got good repeatability and a good transmission path from these forces through to the point where you take the measurement but as I say good repeatability then you should be in good shape you know as I say gears come in different designs and you need to make sure that you're aware of of exactly what you're dealing with and not just treated as a generic gearbox without knowing what's going on and of course we can have quite complicated gearboxes in this mobius Institute designed wind turbine I'm just kidding we just make animations not wind turbines but here you've got an epic cyclic or planetary gearbox with you know two additional stages to the gearbox we can see it in a bit more detail here but you can imagine that with a gearbox like this let's say we've got a an Isetta on what amounted just at this location where I'm moving my mouse you know you've got these gears turning it at very different speeds you've got gears that are moving in position relative to the location of micellar Amida if you think about this gear here it's now moving away from my accelerometer and it's going to move back towards my cell rahmanir you've got to think about the different failure modes and that's certainly something that so we'll help you with vibration analysis with your particular gearboxes is to consider what the failure modes have been in the past if you've got a long history with the machine so at least you can be absolutely making sure that you can detect those those fault conditions as everything with vibration analysis sometimes you have to make compromises you'd prefer not to but in this particular instance with a planetary gearbox in a wind turbine you may not be able to mount as many as Celler ah meters as you might like you know in a case like this you are most likely going to be permanently monitoring permanently mounting the accelerometers and probably permanently monitoring the system as well but so if you understand the design you understand all the issues related to detecting gear faults then you can make the appropriate choices so a few other just quick points I'll make number one you do have to understand the failure mode as I just mentioned but also you need an idea of the PF interval that is basically the time between when you can detect the fault with any technique and the time when it has functionally failed which may be catastrophic failure but it may be a point before there where you are basically forced to stop running that gearbox because it can no longer perform the the function that it's intended to perform so you've got a certain window of time when you can detect the fault and of course the earlier that you detect it put you in a position where you can decide what maintenance action you want to take which could be anything from low running the machine under lower load it could be an option it could be to you know do an inspection of the gears it may require ordering parts replacement gears replacement gearbox so everyone's situation is slightly different so the earlier you detect it is always a good thing but you've got to think about the PF interval and if it happens to be short then you need to be monitoring it much more frequently if it's a longer failure mode then you've got more time but of course you've got to consider all the failure modes and whatever is the shortest that's going to determine how frequently you take your measurements and as part of all of that because you have to take all of these readings that has a cost associated with it we may also do where particle analysis and oil analysis we really should be performing those condition monitoring techniques and that's where the asset criticality ranking is important to understand the criticality of the gearbox which therefore justifies what effort what investment you make to monitor that gearbox of course there's a lot more we could talk about in terms of making sure the lubricants are clean and everything and as part of that we will perform oil analysis to not only make sure that the lubricant condition is correct but also that it's not contaminated which can be a very big issue in relation to gearboxes and then we can look at where particle analysis to see if there are any signs of where you may see signs of wear before you detect it in the with the vibration readings and as part of all of this is making sure the correct tests are being performed to make sure that the samples are being taken correctly but also you know that the reports are being reviewed by a competent person who can understand what those reports are telling them so that action can be taken so let's have a look at the gearbox vibration measurements I mentioned this briefly before but in the case of bearings for example we've got the the the bearing is generally at a location where you can place in the cellar aa meter and there are all sorts of issues you know regarding monitoring of the bearings trying to get as close as you can to the load zone considering the design you know what's the thrust versus you know radial component and taking the measurements correctly but with gears of course the forces that we're trying to detect inside the gearbox so how is that vibration that will change as the gears as the gear teeth deteriorate or if there's any a centricity and misalignment or anything how is that vibration going to transmit to the available measurement points so you ideally need to know about your gearbox the actual internal design but whatever the case is it's not a bad idea to start by taking measurements at multiple points and reviewing those measurements to see what sort of data you've got and just see what you can see in that data and make sure you can pick up the gear mesh Peaks and so on that we'll talk about a little later so let's just talk about just some of the real basics if you're new to vibration and this is this is important with any gears whatever the design you've obviously got two shafts turning typically at different speeds so you've got to consider you know what those speeds are and what where they're going to show up in a time a four-mile how that how many cycles of rotation you'll see in the time waveform and then also in the spectrum where those Peaks are going to appear what sort of resolution you have and so on and of course in a more complicated gearbox will have multiple shafts but then I'll just start that animation again you've got the mesh as well and we could talk a lot about the tooth design and and so on but ideally in this mesh here we don't have any backlash which is the case where the back of the teeth are sort of making contact that can happen for a variety of reasons from you know poor design and assembly of the gearbox even to VFDs if there's sort of shuttling of the speed you know sort of fast slow fast slow slow slow then can cause backlash to occur and we can detect that in the vibration analysis but ideally what we're going to have is some sliding contact and rolling contact in other presentations we talk about lubrication issues but there's less than a micron between these two teeth so the lubrication is incredibly important with with gears but from a vibration point of view if I haven't actually counted these teeth recently but let's say we had 31 teeth or something like that then 31 times the speed of this rotation we expect some variation in vibration now even though that's nice smooth a little bit of sliding a little bit of rolling contact we do expect some variation in that force and therefore a variation in the vibration at this frequency so let's say that was 31 teeth 31 times the speed of this shaft let's say this was 19 teeth as I say I haven't counted them I should have done that but it'll be 19 times the speed of this shaft that will come out to the same frequency it has to that's just just physics um but either way this there will be a source of vibration at the gear mesh frequency which is the number of teeth times this speed or the number of teeth times that speed so we can you know look at this in a little more detail to see what other sources of vibration and I will talk about other sources of vibration that we might see in addition to just these two shaft speeds in the and the gear mesh frequency now what I'm trying to show here is here's here our teeth our gears turning and the cycles you see in here a simplistic view of the mesh force so we've just got this you know up down up down up down as the teeth mesh so tooth mesh tooth mesh tooth mesh each one of these little cycles in here is a tooth mesh that's the variation in force as the teeth come together I'm not representing here the vibration related to the turning speed of this shaft and the turning speed of that shaft the vibration you measure will be a combination of all these sources of vibration what I'm focusing on here is that tooth mesh forces themselves so that's this cycle so as I mentioned if there were 31 teeth here then I'm going to get 31 of these cycles per rotation of this shaft and if there were 19 teeth here then it's actually 19 cycles per rotation of this shaft so the result is of course one source of vibration now if you look closely you might notice that the vibration seems to rise and fall just slightly there's some modulation now when I originally recorded this webinar there were a lot of questions about modulation so I am going to create a separate presentation focusing on that modulation question and you'll be sent an email on that or if you've just come across this in our mo mobius area then just look for that presentation shortly but it's not uncommon to see a variation in this EMS force because of any run-out any eccentricity any slight imperfection in the forces as these two gears turn if they perfectly rotated inner in a circular motion absolutely perfectly then there's no reason for this rise and fall and vibration so it may be slight as we will see that creates sidebands in our spectrum so we expect when we start looking at vibration analysis and we consider the spectrum we expect there to be some vibration at the turning speed of the well let's just say one shaft and we've identified another peak here which is the turning speed of the other shaft now I've called that one XP pinion by definition the pinion is the smaller gear and therefore that turns faster so for the sake of this little graph here that we're going to build up over the next few slides I've just made it a slightly higher speed than the input speed now it could be the output speed the point is there are two sources of vibration for the sake of this little demonstration here I'm just putting those close to each other because that leaves room for all the other things I'm going to show you so first we have those two peaks but that's just normal vibration of course there could be another one of these Peaks if there's an intermediate shaft and if there are two intermediate shafts then we'll have another two peaks okay so then we get a peak at the gear mesh frequency so as I've described that's the input speed times the input gear tooth count and that's the same as the output speed times the output gear tooth count and with a more complicated gearbox we may have multiple gear mesh frequencies the next question you may have noticed these little Peaks sitting around the gear mesh frequency that is due to amplitude modulation it can also be due to frequency modulation now again I'm not going to dwell on these topics too much right now but modulation simply means of a variation a cyclic 'el variation in the vibration amplitude modulation means that that variation is in the amplitude of the vibration and we saw that just a moment ago and soon we'll see it in in more detail frequency modulation is a variation in the free see so imagine that instead of the two gears turning at exactly the same speed from one moment to the next imagine that there's a very slight speed up and slow down speed up slow down as as those meshes occur so each tooth comes into contact it might just slow down just that little bit and then it speeds back up again and the next tooth comes into Misha just slows down just a little bit we're not talking about a lot just talking about a little bit of variation and that's enough if you visualize a sine wave for the cycles to just get a little bit closer to each other further apart closer further and that also creates these side bands that we're seeing here either way we expect to see these sorts of side bands but the amplitude of the side bands and the number of side bands and the separation the difference in frequency of these peaks you know between here and here and here tells us about the nature of the fault condition I am in addition to the gear mesh peak and the potential for those side bands we may see a peak at twice the gear mesh frequency and three times the gear mesh frequency and in fact we may see one at four times and five times the gear mesh frequency now a lot of the sort of classical vibration analysis wall charts and things like that focus on what happens with the gear mesh twice gear mesh and three times Girish frequency the relative Heights and how they change over time it gives us an indication of the different fault conditions and you know you can use those rules and and let them provide an indication of what's going on I would always recommend that you also look at the time waveform I'll talk a little bit more about that in a moment but at the very least your F max the maximum frequency displayed which I have shown here is orders multiples of the running speed but whether it's in whether you listed in Hertz or CPM or in orders either way it must spend at least three times the gear mesh frequency and you need it to be a bit higher so that there's room for these side bands and so on now when we talk about time waveform analysis I'll try and convince you to actually make it more than four times the gear mesh frequency because that'll make sure that what you see in the time waveform is is more useful and the detail you see as they teeth mesh together okay now just briefly there is another source of vibration that we can see depending on the design of your gears okay so here's my gear box little simple simulator now what I've simulated here is a situation where I do have 31 teeth here 19 teeth here and what I'm going to do is damage one of the teeth you can see it moving there I'll just make it a bit easier to see but what I'm simulating in this case is that this tooth when it meshes it's actually going to slightly damage that tooth and then when this tooth comes into contact it's going to slightly damage that tooth and then when this tooth comes back into mesh it'll slightly damage that tooth and then that damage is that and this will damage or they're going to mate with each other and this will touch there and if we keep watching we'll see an interesting thing happen that from now on notice that there are a few points on this gear that are red and a few points oh no sorry no I'm getting ahead of myself if we keep watching we'll see that eventually every single tooth will go red so I'm getting ahead of myself for the next point I wanted to make okay so if we speed up the animation just so that we're not waiting too long you can see now that all the teeth are red now that means that if any tooth is slightly damaged for any reason it may be that the lubricant is contaminated it shouldn't be but let's say it is and some hard particle goes through and slightly damages a tooth and that makes that tooth imperfect and then that mates with the tooth you know on the mating gear and that has the potential for slightly damaging that tooth and as you can imagine these gears are turning turning turning turning turning turning and so this is just repeating repeating repeating same could be true if when the gearbox was transported or installed or whatever you know to mating teeth may just damage each other and as the gear gearbox rotates that damaged that imperfection has the potential to spread around all the gears now in this design which is how gearboxes are supposed to be designed the number of teeth are prime numbers so we got thirty-one teeth now thirty-one is divisible by 31 it's divisible by one but no other number you know you can go through one tooth or two three four five six seven any time you try and divide 31 by two or three or four or five or six etc all the way up to 31 itself if you divide it through you won't end up with a whole number and the same with 19 19 is divisible by 19 and one but by no other number so that makes them prime numbers in gearboxes ideally will be designed with a prime number of teeth and this is the reason the wear spreads more slowly the wear is spread evenly across all of the teeth and therefore as this wear spreads very slowly as a normal process the fact that all of the teeth are sharing in that process it means the gearbox has the a longer life now let's look at a different situation now we're going to do something else we've got actually 30 teeth on here and 18 teeth what I'm going to do is fix the teeth only that easy and will damage one tooth and will just speed up the animation a little bit so if damaged that one little just a little bit they've damaged that one that damaged that one that damaged that one etcetera etcetera now if we keep watching in this instance we will see the strange thing that I still have to mention before and that is no matter how long we wait no additional teeth will come into contact with those red teeth so let's say the gearbox was being transported two teeth were damaged those teeth now potentially wear against other teeth but now only five teeth on this gear and three teeth on this gear are taking that extra wear and that means that those teeth will wear far more quickly than would be the situation if there was a prime number of teeth now we can get into all these calculations you see we have these non prime number C 30 is divisible by one it's divisible by two 30 divided by two is 15 15 is divisible by three 15 divided by 3 is 5 that leaves five is the sort of remainder and we can look at 18 to visible one by two by three and then that final result is divisible by three so we have these common factors of one which is always a common factor two and three and we multiply them together and we get six so I'm rushing through this part you know allies this webinar would go for hours but the key thing is here that if you look at this one and you go through this calculation of the prime factors notice that the largest common factor of type prime factor is five one two three four five there are five teeth on this gear that are going to take the wear in this instance in this one the highest prime factor is three we've got one two three so you can imagine that three times the speed of this gear is potentially going to create a source of vibration so if we actually simulate that then we expect to see these Peaks show up the gear assembly phase frequency is what it's called gear assembly phase frequency so if you like it's three times one frequency so it's three times this one or five times this one it's the same frequency five times this three times that remember this is turning faster so three times that is equal to five times that and so we may see a peak you might see a peak in your spectrum and so what is that it happens to be a sixth of the gear mesh frequency so there's my product of my common factors they are the prime factors that are common to both gears and so our gear mesh frequency divided by six is this and we got harmonics of it of course there's six subjects one two three four five six anyway all the math work out the point is you can see these Peaks now as I said ideally a gearbox will be designed with the you know the prime number of teeth but you know if there's a specific design that was required there may not be a combination possible with prime numbers so they have to use these non prime numbers of teeth and therefore this is a source of vibration that you may see what I'll also mention just while this is going around is that when you have any source of vibration there is potential to excite resonances and if you imagine taking this gear out of the gear box suspending it on some sort of cord and hitting it with a hammer it's going to ring it is a metal object now when it's in the gear box even though it's mating with the other gears it's on a shaft which is supported by bearings and so on it still has a natural frequency it may be different because the stiffness is just changed and so on because of the configuration but it still has a natural frequency and therefore we have the potential to see additional peaks in the spectrum related to that natural frequency and if we have any fault conditions which can excite that natural frequency you know any impacting sort of faults with damaged teeth then that natural frequency will be excited with a higher level of force if you like and therefore that peak can rise we can also see modulation in that vibration you know instead of it being a solid impact if that force of that impact rises and falls then our gear mesh natural so if our gear natural frequency may also have sidebands anyway this just explains why sometimes you might look at gearbox vibration and see Peaks that you can't quite explain they could be related to natural frequencies they could be related to this gear assembly phase frequency okay so moving right along here is an example actually a photo of a damaged gear but you notice that it's damaged one two three four perfectly good teeth damage one two three four let's go in the other direction one two three four damage one two three four damage one two three four damaged chances are if we could you know count the number of teeth here and and look at the mating gear then we would see these same common factors and this is a rather severe example but hey you can see what I'm talking about there now there's one more source of vibration that sometimes comes up is called the hunting tooth frequency now it's related to this very same issue however what I'm going to do here is is damage one tooth and what we're going to look at is how frequently does this tooth come into contact with that tooth so we're not talking about how the wear spreads we're just interested in how often at a individual tooth on one gear comes into contact with an individual tooth on the other gear if we keep watching are they finally came into contact with each other and if we wait for a while they'll finally come back into contact with each other so anything that doesn't happen frequently is going to be a low frequency so from a spectrum point of view if I put in my hunting tooth frequency it could be quite a high peak at quite a low frequency that's just because you know that the those two individual teeth don't come in to mesh very frequently so again if we go back to that possibility that when the gearbox was transported two teeth were were mating at the time and they were damaged then you can get a particular source of vibration every time those two teeth come back into mesh you know that's going to be the sort of worst vibration just when those two teeth come in to mesh when those two teeth mesh with other teeth there's going to be a heightened amount of vibration because the teeth aren't perfect but when those teeth come into contact with each other then the vibration is even worse and so you may hear this low-frequency kind of growling sound coming from the gearbox and this could be the source of that vibration there possibly other reasons why those two teeth were damaged but either way this is the nature of the vibration and there's a complicated calculation you can go through which involves all these prime numbers and all the storage as to determine but basically we can have a peak in the spectrum and it could be you know quite a high amplitude peak and we'll see it in the time waveform as well if we've set it up properly okay let's have a look at some simplified spectra now here we're talking about just tooth wear so the two teeth I have become damaged just so I should say all the teeth potentially worn to some degree and we will see an increase in the vibration so you got to imagine that these teeth are just you know instead of being a nice smooth contact in the gear mesh they are now rough if I can just put it that way and therefore that vibration will change now I've pointed out a couple of things in this sort of classic type of spectrum here's our gear natural frequency peak it could sharp anywhere you know it's whatever the natural frequency is and because of the nature of the vibration we might see some side bands around that that's one thing but we're also just looking at how this gear mesh peak changes in amplitude you know if the teeth are are wearing there's no real reason for a change in the sort of cyclical force of the gear mesh you know the force between each gear mesh doesn't necessarily change from one point on the gear to another points as the as the gears turn there for modulation isn't the real big issue in this case member modulation occurs when the vibration the force in that gear mesh changes per rotation but if the teeth are all wearing basically evenly then we have a heightened amount of vibration at the gear mesh frequency now here's where I'm showing you no gear mesh twice gear mesh three times gear mesh you may see three times gear mesh rising amplitude these are our guides they've sort of been published the vibrations being experienced but particularly with gearboxes that are potentially quite critical components I would be reluctant to make a call purely on what I saw in the spectrum number one I'd want to look at the time waveform number two particularly in a case like this I'd want to look at the the lubricant itself because if there is wear that means there will be metal particles of the metal that you know used to make the gears I expect to see that in the inner lubricant and a experienced analyst or lubrication analyst where particle analyst can look at the shape and size and color of those particles and tell you yes where is occurring so yes vibration may be an indicator ideally you would have already been doing oil analysis and you should be getting warnings from that technique as well but every additional piece of supporting information you can get is always helpful when you're making a call like this so obviously with tooth wear there should be clear signs and lubricant you know the root cause may be also present in the inner lubricant as well may be the viscosity is is incorrect maybe the wrong lubricant has been put in there perhaps the lubricant is contaminated with water or with particles that's are causing that wear to occur tooth load is a different situation where the forces in the mesh itself are higher so you can look at the possible reasons for that to occur but you know we're just driving that gearbox harder the force as each tooth comes into mesh is higher and therefore which bet the gear mesh peak to go up it's just a normal source of vibration it goes up again you know we can look at that and say well yes if there was any air centricity then that's you know the effects of that air centricity would be heightened but because we're focusing on tooth load the real indicator is just that the amounts of vibration in this gear mesh pick will go up now we might always expect to see twice gear mission three times gear mesh because they're harmonics anytime you've got nonlinear vibrations just not smooth vibration we expect to see those Peaks go up but the real telltale sign with to float is the MSP rising up now in the instance of gear s intricity the situation I'm trying to depict here is where if you look very closely at this gear it's turning correctly around the axis of rotation of this shaft so the shaft itself is not moving the gear itself as it turns is is moving in a perfectly circular motion but when we look at this one in actual fact if you look closely the way I've animated this particular situation the shaft itself is moving in a circular orbit so if I leave my mouth in one so reference position you can see that the shaft itself is moving in a circular motion that means that the force in the mesh here will increase in force then decrease in force increase in force decrease in force now this could be because of a bent shaft for example we may have a situation where the shaft is turning correctly on its axis but there is a centricity in the gear itself maybe because of manufacturing maybe because of the way it's been installed you have to sort of consider the possibilities for why this is occurring but anytime we have this variation in the gear mesh force we see more of that modulation in this animation that we see here we've got the two gears meshing together eccentric lis so you know it says I've already said it could be because of a bent shaft because of the gear manufacturer or whatever but if you look closely we see the forces are much higher now and therefore we get more a higher amplitude of vibration when the gears are further apart we get a lower amplitude of vibration so you can see there's more force in the mesh less force in the mesh more force in the mesh less force in the mesh now with my animations and everything I'm exaggerating things and I'm simplifying things so on the exaggeration part you know I'm showing this this mesh it's a tooth meshing a tooth mesh tooth mesh so that's what this cycle is here as I described earlier and then in my animation I'm exaggerating the amount of movement I mean if I didn't exaggerate it you wouldn't see it but what I'm trying to illustrate here is that if for any reason one of the gears is is moving in this sort of circular motion then we will see the same gear mesh frequency there is still the same number of teeth on the gear so we see that same gear mesh frequency but the amplitude the force in the gear mesh rises and falls per rotation of the eccentric gear so in this case we're seeing this rotate around well that's the you know from from the top here to the top there is one rotation of this eccentric gear so rises up when the force is greater right falls down when it's weaker rises up when it's greater and when we see this sort of vibration we can see that in the time waveform it won't be as clean and pretty as what we're seeing there but you will see it in the time waveform but you will see it in the spectrum as well so if we have a look at this what we're seeing here is here's our good old gear mesh peak and some vibration of twice and three times gear mesh but the real telltale is that our sideband amplitude has risen up we get all these side bands in part because of amplitude modulation that I've explained in part because of frequency modulation that as those gears come in as each tooth comes in to mesh we slight speed up slow down speed up slow down the result is you are looking for these side bands we may also see the gear natural frequency being excited it may also have side bands around it but the real focus we're looking for here is that the gear mesh peak itself may not rise up a lot it will rise because there is overall you know greater force in the mesh so we may see that go up but we're really looking for the side bands increasing in amplitude and because it's nonlinear vibration we always see multiples of a source of vibration so multiples of our gear mesh frequency because it's not nice simple smooth vibration in the case of misalignment we've got one gear which is not aligned correctly to the other gear it could be an assembly error it could be that there's excessive wear in one of the bearings as various possible reasons but if you look at this exaggerated animation we're not really expecting there to be a lot of modulation in that vibration you know the two gears are sort of at a fixed relative position assuming that that is the case and therefore our gear mesh force will have changed because of the misalignment but it's sort of constant per rotation so in this case we may still see side bands as normal but we see a change in the vibration and just as we often do in relation to normal misalignment between motors and pumps and so on we may see that twice gear mesh frequency will increase in amplitude more than just the normal gear mesh peak but the bottom line is you know I guess with all vibration analysis one of the things that I think is the most important ability of a good vibration analyst is to be able to transform yourself down and think about what's happening inside the machine you can do it in one of two ways you can think about the various faults and think how that's going to change the vibration what forces will change what modulation will occur is their impacting you know how is the vibration changing and that way you can imagine why the spectrum changes the way it does but on the other hand you have to be able to reverse that process you need to look at the vibration and say hmm why is it so why am I seen that these changes why is the modulation changing why is the gear mesh changing and and go backwards that's the truth in the case of gibb analyzing gearboxes but it's the truth no matter what you're doing with vibration bearings pumps you know whatever the vibration is it's really helpful to be able to relate what you're seeing in the vibration to the machine because then you don't have to remember wall charts and so on you don't have to sort of just take guesses because well it sort of looks like that I mean this is a big call you're making you know do I stop this gearbox do I have to order parts I mean and they could take a long time to be delivered it could be very expensive you know it's really important that you you get it right but obviously it's also important to take your measurements so that you can see those you know Peaks up to three times Girish frequency and as I mentioned soon I would go to four times but also so that you have enough resolution to see the sidebands so let's talk about time waveform analysis I've mentioned it a lot of times it is a very very important tool in relation to gear analysis time the time waveform is the truth in vibration as long as it's measured correctly it is really telling you what is happening inside your machine from one moment to the next so as the teeth meshed together as the rolling elements roll around you know if there's a damaged tooth if the teeth are worn if the forces of rising and falling whatever the situation is like the time waveform is telling you exactly what happened from one moment to the next and then we do an FFT and that just looks at all the time that all of the vibration and says hey I'll give you a summary of what happened so in this instance we're seeing something very very important in this time waveform and I'll elaborate on it shortly but you know in here this low amplitude vibration is the vibration we expect to see as the teeth that as good healthy teeth meshed together but the spikes relate to a damaged tooth coming through the mesh so the damage tooth comes through healthy teeth damaged tooth healthy teeth damaged tooth healthy teeth etc etc etc our time waveform tells us exactly what's going on we can see that there are spikes if we know if either we set our x-axis up in rotations of the shaft we would see that for the gear with the damaged teeth this is one rotation two rotations three rotations four rotations and so on in this particular instance the time waveform has been beautifully set up so that I can see enough of these events and I'll talk more about this in a moment but there are two keys with time waveform analysis we want to see the events occurring we don't want to miss the events but we don't want to have so many of the events in our time record that they kind of get smooshed together and we can't see them so here you know for the speed of this particular gearbox we needed just over three seconds of time to capture one two three four I won't count them all out but you can see there's a limited number of teeth coming in to mesh so that I can easily see what's going on that's perfect and I'm also sampling fast enough so that I can see these events if I wasn't sampling fast enough I could miss them or not get a detailed picture and I'd certainly get the amplitude wrong but I'll elaborate on that in just a second so time waveform analysis is important one of the reasons it's important one of the reasons why it is important is because the fft process has limitations the fft is looking at this whole time record here and and trying to come up with one set of peaks in the spectrum that represents all of that in some cases does a great job you know with normal unbalanced and simple sources of vibration does a good job and we can see harmonics and side bands in in the spectra and try to interpret what that actually means in terms of the machine but and therefore what is really happening in the time wave form now at the time wave form is telling you but when we get even more complicated time waveforms or therefore vibration from the machine the fft becomes even more limited what you got to imagine inside the machine is we have this vibration just coming from the machine whatever it looks like and the analyzer grabs a chunk the time of this chunk this time record is based on the lines of resolution setting so grabs that it windows it calculates an fft and grabs another chunk and there may be overlap and it windows that and calculates another FFT and does it again and does it again and we end up with these four time waveforms and we are sorry for spectra and we averaged it together so with normal vibration this is just fine there are limitations and i've talked about those limitations in other webinars and certainly now training and so on but the fact is that for normal processes that's an acceptable process but when you got more complicated vibration from a gearbox all this windowing and averaging and all of that can really sort of lose the information that we ultimately want to see when you determine a form analysis we are keeping one of these time wave forms and analyzing it or we may take a separate time wave form measurement which is ideally suited you know we might take one measurement which is ideally suited for the spectrum and another one which is ideally suited for the time wave form I'll elaborate on that a second but the fact is that you know when we take those four spectra you know we average them together we say you know depending on the technique we use we end up with just this average and that's fine for normal vibration analysis but for time may form a narrative so for gearbox analysis I think it's too limited you need the time waveform coming back to this time waveform we were looking at before this was the situation so every time this damaged does these two damaged teeth came through the mesh we got the spike of vibration all these other healthy well-lubricated for now or healthy for now just depends on what happened to the chunks of metal and so on but the bad teeth come through we get the spike the good teeth the bad teeth the good teeth the bad teeth this tells us exactly what's going on I know which gear is damaged I get a sense for how badly it's damaged you know you can look at these G levels and in get an indication of the severity but you got to remember this transmission path issues you know how's the gear vibration attenuated and and so on the tongue waveform is telling you exactly what's going on just a quick primer on the time waveform you know I'm taking way too long on this this presentation but I hope you're finding it useful imagine that what we're seeing going past here he'll slow it down what we're seeing here is the real vibration imagine if I could sample it a billion times a second so that we nearly exactly what was happening from one split second to the next let's just assume that that is the real vibration on the machine where we've placed the vibration sensor now what I'm going to do is I'm going to come along with this analyzer and what it has the ability of doing is at certain time intervals it's going to say what is the voltage now from my cellar ometer the accelerometer is sensing that vibration on the gearbox and there are issues we could talk about with the mounts resonances and the response time with your cell rahmatan all sorts of things but so the analyzer says I'm going to take a voltage reading now now now now etc you can see them but if those time waveform if the sample time the time between each one of these samples is too slow you can see that the waveform this green trace here is what the analyzer will remember and then put through the FFT and display to you in a time waveform window well this green trace has nothing to do with the real vibration does it it's obviously poor we are obviously sampling too slowly so we speed up the sample rate the number of samples per second from our FFT parameter you will either have the option to change the number of samples per second or you'll have the option to change the F max this has nothing to do with the lines of resolution setting at all it's all to do with your F max setting or you may have direct control so what I'm going to do is I'm going to increase the sample rate faster and faster and ideally we'll get to a point where we're sampling it fast enough that our green trace looks well ideally identical to the original real blue vibration but this is what you have to select with your measurement settings you have to say how fast is fast enough so that I can see exactly what's happening is these teeth come into mesh because if I've got one damaged tooth then as that tooth comes into mesh and the forces all change and therefore the accelerometer is experiencing that I want to get lots of these digital values these voltage readings from the sensor as that tooth comes in to mesh I might like ten little voltage readings as that tooth comes in to mesh so that a I can see it clearly in my tongue waveform and be more importantly you know as that vibration changes we want to capture it if we've got any really sharp spikes we could take a voltage reading on either side of the spike and never see the true vibration I can't illustrate that easily but if you look here you'll see just here that there's a blue spike that goes up but we didn't capture that the Green captured waveform is lower we ideally wanted to get a digital sample up top here you can see it there as well we didn't capture these spikes we sampled even now we're sampling too slowly we might just flip it you can see here we flipped it and captured the peak of the time wafer of the true vibration but here we didn't I hope you can see the difference between the green and the blue here we didn't see that spike well you could be having 20g spikes but never know it because you're sampling to a to lower rate here's a perfect example look at this great big spike here but our captured waveform looks nothing like it so we need to sample very very quickly and as a rule of thumb I wouldn't want to have anything less than 10 samples per event that I was interested in in this case tooth mesh I'll explain how you can set it up correctly in just a second anyway that's the time waveform process so I want to sample fast enough to capture all the lovely detail that I'm interested in but when I look at my time waveform I don't want to have so many of those events that they all get smooshed together like I mentioned before I would like to have six or ten or maybe fifteen on my time waveform plot so that I can see that these things are actually happening and then I can zoom in and look at them a bit more closely if I want to but at least I can see them if if the time waveform is not long enough I may not see the event at all there are situations I mentioned the hunting tooth frequency before you might listen as in with your ear to a gearbox in here click click click as two particular teeth come into mesh - each with each other your time waveform could start and end between those little clicks in the in the gearbox or you might in the time of your tone waveform have lots and lots and lots and lots of teeth meshing together and therefore that variation in vibration is just not seen because there's too many events per cycle now let's consider another way of collecting our data in a normal analyzer setting when you walk out to the machine even if it's even if your analyzer setup correctly and you press the start measurement button it will just start collecting the data just sort of like I described before so here's the vibration whatever it is and these little marks indicate revolutions of the shaft and down here I've got my tax ignore just for now as a reference so let's just zoom in on that so there's just two revolutions of the shaft and the vibration I'm seeing here in my little simulation is I've got 1x vibration 5x vibration 7.2 X vibration now it might be you know bearing frequency for example and I've got noise on top of that so that's what that's what's in this waveform up here now when the analyzer normally works what it does is it grabs a chunk it might be 1024 2048 4096 it just depends on your number of lines of resolution it's going to grab a chunk take that out and do the windowing and calculate the spectrum it's then going to grab another chunk now I'm what I'm illustrating here is no overlap but the point is the analyzer as soon as it's got it just keeps on sampling it just keeps on taking chunks of the tone waveform and if I was to look at each of the time waveforms in this little time record in this time record in this time record let's just call it average one average to average three average for well if I look at them sort of stacked above each other you can see that there's no sort of commonality between them in other words this one goes high at the same time as this one sort of at the midpoint in its cycle this one's going high here where this one's at the sort of the midpoint and so on anyway point is that if I was to use my analyzer to average the time waveforms not per spectra but the time waveforms you can see that would be a pointless exercise because these things are just going to average each other away that's the normal function of an analyzer but we can use something called time synchronous averaging and the difference is that the analyzer will when you press the Go button it'll say okay I'm going to start digitizing internally but I'm not going to save the data until I see a tack pulse come along from the shaft of interest so the tack comes and says oh good let's go and it grabs our 2048 time samples and when it finishes and during that time there'll be a certain number of revolutions of the shaft it'll finish but rather than capturing this data as well as up wait wait wait wait wait up bang there's another pulse that's come from my tax ignore start saving these samples again and it'll go again so what we've done is we've synchronized our sampling to the rotation of the shaft we've just waited this little bit and started sampling at the same point in the cycle or the rotation of the shaft so we do it once we do it again we do it again we do it again so this is this one expanded etc for that one and this one for that one now if we look at these and we overlay them on each other you can see that the dominant sort of once per revolution vibration has now been synchronized you know those waveforms overlay on each other but what happens is if you average those or in this case we've just got three but let's assume we did it a hundred times it takes quite a few of these averages to weed out the data what happens is anything that's not synchronous with our ones per revolution attack signal gets averaged away so our bearing frequencies get averaged away the noise gets averaged away and the gears that are turning at a different rate also get averaged away so that the gears that are turning at the rates where this tack signal came from will remain any vibrations synchronous with that will remain everything else will be removed and so if we were to look at what is inside this well we've got the one X plus five X because one X + 5 X are synchronous and integer multiples of 1 X so you can see that the time signal is averaging average data is exactly the same as just the 1 X + 5 X data but the the this shows the 1 X + 5 X plus 7 point 2 X the 7 point 2 X has been averaged away anyway we don't really have time to keep yakking about this but hope that sort of clarifies why you'd want to use time synchronous averaging now one of the big questions that often comes up as well how do I do it ideally you would stop the machine you'd put a reflective tape on the input shaft you would set up your analyzer and say okay one time singleness averaging but my trigger signals coming in from that input shaft and you would take the measurement you need lots and lots of averages and you will end up with a nice clean time waveform you can analyze that time waveform we might use a circle plot which I'll describe in just a second but we can look at it just on a normal plot as well but if it's just a simple single stage gearbox then that's fine however if you've got you know multistage gearbox well you just average the way a lot of gear vibration that you are interested in so now ideally you would also get a tax ignore off the intermediate shaft and or in every shaft that's in the gearbox now you may not be able to do that but thanks to the wonders of software and electronics and so on there is an easier way you still need a once per Rev sort of reference signal that comes from the gearbox you still need something that tells you at least for one of the shafts you know that you need this one tach signal but you can either use a tracking ratio synthesizer or your analyzer may have the internal smarts to do it and what it basically does is it says okay this is the ones per Rev of the input shaft and I know that the intermediate shaft speed is that four point six seven times that speed or might be you know at a lower speed whatever it is you tell the tracking ratio synthesizer or the analyzer what that ratio is and if you like it synthesizes the tach signal that would have come from that intermediate or upper chaffed okay so it's it's creating the time the tach signal that you are unable to measure so again you know you will have multiple measurements for each shaft however you only need the one tach signal the analyzer will do one measurement then you set it up and say okay well the ratio between these two shafts and the gearbox is whatever it is speed increase or speed decrease and it will give you a good time synchronous averaged time waveform which again you can look at as a spectrum but you sure want to look at it as a time waveform as well because any damage to any teeth will show up very very clearly so as I mentioned before we would ideally capture a minimum of ten samples per tooth mesh now you can do all kinds of calculations in terms of what's the speed of the machine you know what what times it take for a shaft to rotate how many teeth are there therefore you know what proportion of time are the teeth in mesh and so on there's another simple way to do it and takes a bit of an understanding of how the analyzer works the fact is that when you choose an F max for your analyzer the analyzer itself is going to retain for you a time waveform that is sampled at two point five six times the F max so let's say we had a 1,000 Hertz F Max and therefore the analyzer is going to sample at two thousand five hundred and sixty samples per second now that means that if there was a peak in the spectrum right at the right-hand end of the spectrum so this is a time way from my apologize imagine this is a spectrum and we've got a peak right up here let's say that peak is the gear mesh frequency so that relates to the you know the the gears meshing together well if the peak was right here at the very end of my spectrum that means that as the teeth mesh together I'm getting two point five six samples per tooth mesh you can sort of do all the maths if you like but that is that has to be the way it works out so then if I increase my F max I am increasing the samples per second which means I am increasing number of samples per tooth mesh as these tooth come in to mesh I get more samples well if I change if I doubled my f max now that gear mesh peak will be in the middle of the spectrum and therefore I'm going to get you know two times five two point five six samples Mich in other words we've 5.12 well what if I put my gear mesh piec a quarter of the way along my spectrum so therefore the number of samples per gear mesh will be four times if it's a quarter then it's four times two point five six which is approximately ten samples per tooth mesh so the simple thing is if you set up a spectrum that is tip a minimum of four times the gear mesh frequency you will have at least ten samples per tooth mesh that should mean that the spectrum you get is good because it's you know you'll see four times gear mesh in your spectrum that's good and the time wave form should be good now the other part of it though is that the resolution dictates how many samples you see as you increase the resolution the analyzer keeps on sampling at that same rate it just samples for a longer time and uses that for your FFT now for the spectrum you do one nice high resolution because you want to see all those side bands but for the time waveform you only want to see you know six 10 15 of those tooth measures of those rotations of the shaft in the time waveform anyway so that hopefully gives you an idea of how to set it up so you get a good time waveform okay now we can analyze the tone waveform in a normal sense with the x-axis as time in seconds or milliseconds a lot of software will let you change that to be revolution so we can sort of see what's happening per revolution of the shaft and that's sort of what I'm seeing up here so this is this is just some vibration I'm getting one two three four cycles per revolution or four events per revolution so I can look at that as a time waveform so yep that tells me what's going on but remember you know when we look at a time waveform what we're interested in here is what's happening as the shaft turns the shaft is round so wouldn't it be cool to display that time waveform in a circular format so that we can really relate what happens as the shaft turns well that's what a circle plot does so if I just expand and say what happened per revolution one two three four right just as I said let's wrap that around a circle so we wrap it around what we're doing is we're saying you know this is one revolution of the shaft so as the shaft turns this is one revolution the shaft turns at a certain speed in rpm or Hertz and we can calculate that by saying 1 divided by the value in Hertz is the number of seconds so that's you know the number of seconds that took to rotate if we wrap our time waveform around that well what do we see 1 2 3 4 events per rotation and because it's sort of nuts nice smooth pattern we know that it is exactly 4 if it wasn't exactly 4 events per rotation and these would all overlap if I had a lot more time to tell you about circle plots then I could demonstrate that in this particular case we've got I forget now but quite a few events per rotation let's just called 31 like our gear before nice prime number so if we expand that out so this is now a situation like our gear mesh one gear mesh two gear mesh three gear mesh and if we wrap that around our circle hey it even looks a little bit like a gear doesn't it because there's a gear mesh gear mesh gear mesh gear mesh gear mesh etc etc and because of course with one rotation with a gear we've got an integer number of events per rotation hey there's an integer number of teeth therefore our circle plot looks good you know we've got our time waveform it's wrapped around the circle plot and in this beautiful simple case that I'm showing here you know we can see very clearly that the vibration as the gear turned is uniform each mesh is the same but I can see it very clearly now rather than waiting for that to finish now I'm going to show you a different sample so here's the same gear mesh frequency but once per revolution I've got a little spike in my in the vibration let's imagine that there's a damaged tooth or something like that so here's the same thing we looked at before but once in the revolution we've got this spike now yes it's very simple the way I've depicted it but the point is that when you wrap that around the circle plot we can see quite clearly what's going on you know if we wait long enough they will just keep overlapping on each other and the big spike so here it comes there's the broken tooth coming down it will sit right on top of that same spot so if you look at a real circle plot from a machine with a damaged gear a damaged tooth you will see you know some normal sort of amplitude vibration around the plot but in one spot you'll see this big spike if there were two teeth next to each other you might see you know a broader spike year or even two spikes but let's say there was a damaged tooth and then there were a few good teeth and another damage to tooth in you see a spike here and a spike there and with time synchronous a purging how you can even figure out where they are because you know that this is the start of a tack signal you can even say well okay you know if you know one two three four five six you know seven teeth around from where the tack reference was is where the damaged damaged tooth is now that'll really freak out any of the guys working on the gearbox when you can say yeah I mean not only do I know that there's a damaged tooth I can tell you which one it is man they're going to be impressed but anyway I'll just briefly mention that you know we use techniques like enveloping demodulation peak view shock pulse etc typically we use that for bearing analysis the reason is that in the early stages of a bearing fault we get this impact but it's a very low amplitude event you know if we can visualize the bearings rolling around there's a little bit of damage on the outer race the ball rolls across that so maybe you know six point seven times per revolution we get this little spike in the normal spectrum it's kind of being swamped out because there's all other higher sources of vibration it's a very short event it may not be picked up very well in the FFT there's a lot of issues there but because we see that as a spike which causes excitation in the higher frequencies we're exciting - see natural so higher frequency natural frequencies in the structure or in the sensor itself we can use a demodulation technique whether it's a peak view or shock pulse or whatever we're taking advantage of that factor we're filtering out all that lower frequency higher amplitude vibration we go through an enveloping process which allows us to see that not only is a time waveform but as a spectrum but bearings aren't the only thing inside a machine that can cause that impact which can excite those natural frequencies if we have damaged teeth then the same thing occurs if we have looseness and we're getting a once per revolution impact the same thing occurs if we have broken rotor bars the same thing can occur the impact is exciting the - sheet natural frequencies the natural frequencies that exist in the machine up above you know to say two or three thousand Hertz we can talk about this in terms of stress waves and shock pulses and so on the bottom line is we can detect that in high frequency in the high frequency part of our spectrum if we use these same techniques that we use for bearing analysis and you may have seen it already you might be looking at a peak view spectrum or wave form or your shock pulse or whatever and c1x vibration thing well why am i seeing an increase in one X whilst because there's probably looseness or it could be one tooth damaged in the gearbox once per revolution we're getting this the stress wave so we can use those same techniques to detect gear faults as well as bearing faults and once we have the time waveform you know from that enveloping process we can look at the we can look at the time waveform we can look at it as a spectrum we can even wrap it around a circle as we did just before so they're very useful techniques anyway this has been a very long presentation I hope you've hang in there with me I normally try to keep them a bit shorter this is a big topic so I hope you found it useful a few quick factors you do have to think about gearboxes very carefully to make sure that where you place your sensors you're capturing the frequencies of interest there's a good mechanical transmission path to that point and that your measurements are being set up correctly spectra are useful enveloping and the various techniques are useful time waveform analysis is very important you know have a look at time synchronous averaging have a look at circle plots if your software has it think about the criticality of the gearbox and how important it is that you get all this right think about the failure modes and the PF interval to make sure you're taking your measurements frequently enough and also think very carefully about oil analysis to make sure it's being lubricated correctly without contamination and we're particle analysis excuse me so that you can detect those particular contaminants but also detect any where that's going on so thanks for viewing this presentation or mini training course however you might want to look at it as you probably know we've got lots of other presentations on our website on a range of topics I'm going to do one on modulation to help explain that because that's something that sort of comes up quite frequently as a question of course you can click off at this point but I'll just tell you that we've got lots and lots of training courses and products that explain all of these principles in much greater detail it is really important that you understand how to take measurements properly ranging from how you mount the sensor to how you set settings like resolution and F max how you analyze the data how you determine when phase analysis is important time waveform analysis spectrum analysis but you know there's a bigger picture story here that relates to reliability it's fantastic to use vibration to detect faults it's very important you look at your maintenance practices and the way the machines are being operated and the selection process for new components and lubricants and so on so that we find these faults less frequently we need to improve the reliability so that we don't have faults to detect the most important thing is that we increase reliability so we reduced downtime we increase the confidence we have in the plant we increase the safety in the plant and that's just great for job security you know you keep the plant and the business financially secure and your job is financially secure and it's a safer work environment anyway we have lots of training that cover all those sorts of topics in I learn reliability and our category one two three and four training courses anyway thank you very much for listening to this presentation bye

Channel: Mobius Institute

Views: 126,686

Rating: 4.9140186 out of 5

Keywords: Mobius Institute, certification, ISO 18536, mobiusinstitute.com, vibration, iLearnReliability, iLearn Vibration, iLearn Reliability, PdM, Diagnosing Gear Faults, iLearnVibration, vibration training, mobius institute, condition monitoring, vibration analysis, Gearbox Vibration Analysis, accredited, training, Jason Tranter, CBM

Id: 0TH5SLghYPY

Length: 83min 48sec (5028 seconds)

Published: Mon Jan 20 2014