OK, I've read most but not all of this thread. First of all, most belt-induced vibration is not due to belt resonance. Most belt-driven machines, unless the sheaves/belts/alignment/tension is spot on will exhibit vibration at 2X belt rpm which is insanely close to the fan turning speed on many fans, especially NYB fans (can't figure out how they do it). Often there will be several other harmonics as well... 4X is not unusual. I usually take a 200Hz, 3200 line velocity spectrum to be able to seperate everything out.

Belts can slip continuously without making a sound (at least that can be heard above the other noise). Soemtimes it will be only one belt slipping. A single belt can slow down enough so that it's 4th harmonic matches 2x fan rpm. Just food for thought.

A shaft bent external of the drive end bearing will generate vibration 2x rpm. The belts reach maximum tension as the "high side" of the sheave rolls around and then tension is at a minimum 180 degs. later. This produces 2 pulses per rev. Check the shaft runnout near the end of the shaft.

The number one "belt problem" is not belts at all, probably not even alignment, but rather worn sheaves. Sheaves are seldom ever changed, partly because of the work involved, but mostly due to the perception that it's too expensive. But on a fairly large drive with multiple belts, a set of sheaves (minus the bushings which can usually be reused) only costs about as much as 2 sets of belts. It's cheaper to change the sheaves in the long run.

Finally, do these fans use "power band" belts with multiple V's on a single outer band? If so, get rid of them. They are too heavy, too stiff, and serve no useful purpose. Belts don't have to be exactly the same length since they will all be the same length on the taunt side if the sheaves are in good condition. The only thing good about the power band belts is they require the motor actually be moved to install the belts (instead of rolling the belts on and off).

Finally, a "cocked bearing" can cause 2x. So can a cocked coupling half or a cocked sheave.

In retrospect, the "quiz" format was probably not the best choice. I wrote this post very shortly after we got the results of a first check I had requested where we deliberately set the belts at the high-end of belt tension band. The 2x vibration shot through the roof as I had predicted. I was elated (that's the first time I have ever cheered for high vibration!). I thought I had it all figured out, and so I posted the quiz so I could share my new-found knowledge. But after talking through the many possibilities that many people had suggested that I hadn't even thought of, I think I need to keep a more open mind and tap into the tremendous knowledge on the forum. To do that most effectively, I should have summarized all my facts up front.

The question of belt matching did come up while we were analysing this machine. All belts were Gates Super HC 5V1400 (5V cross section, 140"). All three belts had the 5V1400 stamped on them. They also all had another number stamped on them (which we later found out was a date code). That second number was the same among two of the three belts on the machine, but different on the third, which led someone to believe we might have a belt matching problem. We called Gates at their Texas customer service center, got through to a real live person who was tremendously helpful. They confirmed for us that all their 5V series belts are Super HC and all their Super HC are in their V-80 program, which means they are manufactured so that all belts within a part number are made to close enough tolerance (the Rubber Manufacturer's Association recommended tolerance) so that you dont' have to worry about matching as long as the belts are all from Gates and all replaced at the same time.

While we were sorting through that confusion about those extra numbers, someone at our site had recommended going to the powerband. They said: "what could it hurt?". I had an old post from some guy named Rusty, and it was enough to help convince them that we don't need to go changing things if there is not a good reason (can always introduce new problems).

I am reasonably confident that what we're seeing is not a harmonic of belt speed. We see this 2*fan speed at the same spot in the spectrum forever (we have data back to 1989). It doesn't wander around in frequency. Over that time we have replaced and adjusted the belts many times. We have occasionally done thermal images of the belts and don't see any huge difference among belts.

Sheave wear is something I have been thinking about. I don't really think it is tied to 2x fan (do you?). But nevertheless it's something we don't check. We don't have a sheave wear indicator. I don't see anything wrong with the sheave from visual inspection, but I'm not sure how accurate that is. Are there any tips for inspecting the sheaves without a wear indicator? I would expect if it were worn I might be able to see some kind of ridge at the deepest point that the belt reaches? I certainly don't see any ridge like that. If I get a chance, will post some photo's of the sheave grooves and the way the belts fit into them.

Also, I still have to find some photo's of the other end of the fan as Danny had requested.

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

Pete, sounds like you've thought a lot about the belts. What if the sheave wears very evenly, where there is no "ridge" or anything to give a visual clue that wear is excessive? (hey, it could happen). Even if the sides of the grooves remain flat, if the V is too wide, the belt is not going to lock up on the sheave properly, no matter how much tension there is on it. What actually causes the belt to "lock" onto the sheave is the change in cross section of the belt as it's bent. Basically the sides of the belt "pooch out" (don't know how else to describe it) and push against the sides of the groove. If the groove is too wide, there's not enough force to keep the belt and groove in contact. Also, it's not unusual for the wrong belts to be used on a given sheave. Usually, the "correct" belt is not in the tool crib, so someone gets a set that are close to right. The next guy that changes the belts justs uses what he finds installed. Happens all the time.

Your belt supplier should be able to get you a set of "groove gauges" for the various cross-sections. Tell him you want the "free" ones!

Have you tried less tension on the belts?

As mentioned by others, the bearing supports look awfully flimsy. Hard to imagine there's not a resoance somewhere. Have you tried a TSA using the fan. If you take enough averages, at a high enough resolution, the peak will eventually drop out if it's not phase-locked with the fan.

EPete,

The easiest way to check for sheave (or belt) wear is to see how far below the outer diameter the belts are riding in the sheave. With a new sheave and new belts (the right size by the way), the outer circumference of the belts will be just a little bit (thickness of the top cover for example) bigger than the OD of the sheave. Now this of course is assuming that you are using th right belt for the sheave (not an "A" belt in a "B" sheave).
Some of the 5V sheaves, with "B" size belts, are hard to tell, but that's an exception.

You can also take your fingers and "pinch" two different sides of two adjacent grooves and MOST of the time feel the ridge, or the worn areas. The sheave has a flat taper when new, and you can readily tell the taper is worn.

Dave

Thanks Rusty and Dave.

I think frame/support resonance may still be an open question, even though we couldn't find any when bumping. It seems unusual that the whole structure moves so much. Also thinking about beairngs and various internal looseness as remote possibilities.

Back on the subject of sheave wear, I did find some info about worn grooves in one of Heinz Bloch's books . It suggests the typical pattern would be "dishing out" as shown in attached powerpoint. So you are probably right that absense of a ridge doesn't mean much. And it shows that the belts will sink deeper if the belt or sheave is worn as Dave said.

We have replaced the belts many times over the life of these machines, but never replaced sheaves or even measured the sheave wear (other than visual inspection) to my knowledge.

I've been thinking some more about whether a worn sheave groove can be related to the magnitude of the 2x vibration. I think maybe it can, if it affects the ratio of tight-side to slack side belt tension. For example the mu shown in equation 1 here: http://www.mech.uwa.edu.au/DANotes/V-belts/kinetics/kinetics.html#top

I'm not positive I reallly understand how that all fits together, but it definitely seems credible that change in sheave profile could possibly affect the ratio of tight-side to slack side tension. If it does that, it will increase the radial loading required to transmit a certain power. We already know our vibration is sensitive to sheave radial loading, so it could be the reason why this machine is recently vibrating more than the sisters, even when we adjusted it down to min belt tension.

Since it appears to be an easy check, I will definitely request sheave depth to be checked with a groove depth gage.

Dave – thanks for that info. I didn't understand this part however: "You can also take your fingers and "pinch" two different sides of two adjacent grooves and MOST of the time feel the ridge, or the worn areas. The sheave has a flat taper when new, and you can readily tell the taper is worn."

Are you saying that the belts will be loose in the groove... so you can pinch the belt to one side and feel a gap on the other side?

SheaveGrooveWear.ppt (180 Kb, 23 downloads)

I have hard time understanding why right hand case (in the previous attachment ) in general signifies poor sheave condition. IMO, as long as belt is capable of being wedged into the groove without slippage during operation, thus, transmitting proper torque, there is nothing wrong in comparison with the left hand case. Proper wedging requires only flat walls and consistent angle in the groove, IMO. Maybe somebody can elaborate on that.

BTW, does anybody know how to measure belt slippage?

A few things that can go wrong from sheave wear:

If the groove wear is uneven, then a matched set of belts may act unmatched... share the load differently.

If wear proceeds to the point that the bottom of the belt touches the sheave, the belts will overheat.

As far as how it will affect the effective friction coefficient between belt and sheave and the radial tensions while under load, I'm not sure. I do think the geometry was originally designed to maximize the friction coefficient between belt and sheave and anything that alters the geometry will likely decrease it from that optimized design point. If I am understanding the link above right, higher friction coefficient gives higher ratio tight/slack side tension and lower radial force for given transmitted load.

Although I'm still thinking about that... it seems a little contradictory that I can predict the ratio of tensions knowing only the friction coefficient and the geometry. If that is the case, then if I add another equation P = (Tension1-Tension2)*BeltVelocity, I should be able to solve those two equations in two unknowns (tenion1 and tension 2) to determine the tensions (assuming I know horsepower demand and belt velocity). But during that whole process I never considered the static tension adjustment... so I'm sure I must be missing something.

To measure belt slippage, you could measure the rpm of both sheaves with a strobe. If there were no slippage, then the linear velocity of each wheel at the pitchline is the same:
Pi*F*D1 = Pi F2*D2
where F is speed expressed in revolutions per second.

If there is slip, the driven velocity will be lower and the total of the sum slip velocity at both sheaves is
Vsliptotal = Pi*F1*D1 - Pi F2*D2
(assuming 1 is driver and 2 is driven)
To express it as a fraction, divide it by driving linear velocity:
slip fraction = Vsliptotal / Pi*RPM1*D1

I believe infrared will be pretty good at detecting slip as hot belts exiting the smaller sheave, especially when there is something to compare it to (different slip among belts on a machine, multiple similar machines).

There are in fact at least two kinds of slip. One unavoidable kind is called creep. The belt entering the driving pulley is at a higher tension than the belt exiting the driving pulley. And therefore the tension changes from max to min as the a section of belt rotates with the pulley from tight side to loose side. But the belt is elastic like a rubber band, so changing tension equates to changing lenght. Since the belt is growing shorter as it travels from tight side to loose side of the driving pulley, it must creep in order to accomodate that expansion. Similar situation on driven pulley except the belt is growing longer as it travels from loose side to tight side.

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

quote:

Originally posted by electricpete:
Although I'm still thinking about that... it seems a little contradictory that I can predict the ratio of tensions knowing only the friction coefficient and the geometry. If that is the case, then if I add another equation P = (Tension1-Tension2)*BeltVelocity, I should be able to solve those two equations in two unknowns (tenion1 and tension 2) to determine the tensions (assuming I know horsepower demand and belt velocity). But during that whole process I never considered the static tension adjustment... so I'm sure I must be missing something.

I think maybe what I'm missing is that equation i of the link above is the maximum ratio of tensions before "gross slip" occurs. I'm guessing machines normally operate somewhere below that maximum.

quote:

Proper wedging requires only flat walls and consistent angle in the groove, IMO. Maybe somebody can elaborate on that.

The belt really doesn't lock up via a wedging action, if you are thinking of a flat sided belt making straight line contact with a flat sided groove, pulled into contact by belt tension. When the belt bends it is not flat on the sides, but is actually rounded as it expands in width. It's really more about groove width than it is the sides being perfectly flat. If the belts are riding 'low' in the groove (good point), it means the grooves are too wide (or the belt is wrong) and it's time to change the sheaves.

When I hear a fan start up, I like to hear a little belt slip just before the fan comes up to full speed (and load). This tells me the belts are not too tight.

Increasing belt tension on badly worn sheaves is not going to stop the belts from slipping, at least not for long. It's just going to create vibration and fail the bearings from overload.

And changing to the "notched" belts is not going to help for long. The belts are still going to slip, get hot, and start cracking in the bottom of the notches (which are just "stress risers"), and eventually the "nubs" are going to start stripping off the bottom side of the belt.

In summary, and not to put too fine a point on this discussion, change the sheaves.

Belt_Wedging.doc (235 Kb, 15 downloads)

For those still hanging with this topic, following is some data on the 3-head sander I mentioned. Basically it shows pretty conclusively that there's a roll resonance being excited by the 5th harmonic of the drive belt turning speed.

Results of the Impact Test on the roll

Baseline, TSA on motor, TSA on roll, TSA on drive belt:

Thanks Rusty. With the benefit of your discussion, we can recognize your first spectrum the belt pass harmonic pattern. 2,3, 4, 5, 6, 7, harmonics of belt pass all show up in your first spectrum (labeled attached), peaking at the 5x belt that you proved was resonant by bump test.

I don't see any similar pattern in our spectra that would indicate our suspected 2x fan is a harmonic of anything other than 1x fan.

Rusty.ppt (76 Kb, 18 downloads)

Attached are photo's of the fan / motor / support / belts etc from various angles. Let me know if this gives any ideas.

There are two fans mirror image of each other. One on the right of the first picture (closer) and one on the left side (farther). The right one rotates CW and the left one CCW and they discharge into the common header going up above. They both take suction from a common header in back.

Danny - I don't know whether the sheave is a QD bushing or a taperlock bushing. I'm not really sure what the difference is between those two types. Can you tell from the photo's (slides 5 and 12)?

PhotosSmall.ppt (564 Kb, 30 downloads)

QD

Thanks Denny.

Attached is the data showing change in 2x fan vibration as we adjusted the belt tension.

The first slide reviews the various frequencies of interest on this machine:
Belt 640 cpm
Motor 1780cpm
Fan 1210 cpm

The remaining slides show waterfalls where we adjusted the spectrums.

The earliest data is on bottom, latest data on top of the waterfall.

The bottom is first at min belt tension (labeled 8#)
The middle is the second at max belt tension (labeled 12#)
The top is the last at min belt tension (labeled 8#)
All this was done within about two days.

1H, 2H, 3H, 3V all show the pattern pretty well – the 2x fan increased when we tightened to max tension and decreased when we went back to min tension. Although the 2nd time we went to min tension (the top spectrum), the 1H and 2H did not come back as low as original, and on the 1A and 2V positions the vibration went up at that frequency (go figure ). Still overall I'd say it is a pretty strong pattern, particularly on the points that show the highest 2X fan.

The sheave/belt data:
motor sheave - 8.0" diameter
fan sheave -11.8" diameter
center-to-center distance 54.4"
Motor horsepower – 40hp.
Three belts per machine.
Belts are 5V1400 (Gates Super HC).

We tension with the force displacement method. We adjust so that the perpendicular force at belt midspane required to get 7/8" deflection (1/64" per inch of belt span) is 8 to 12 pounds.

VibVsTensionSmall.ppt (364 Kb, 21 downloads)

Pete, I may have missed it, but did you check the radial runnout of the fan sheave/shaft? The plot below shows a roll with a large 2x vibration in the presence of nothing else really (after we loosened the belts and balanced the sheave). The sheave had 0.008" TIR radial runnout. I believe that what happens is the belts go tight and then slack (180 deg later) when the "high spot" on the sheave passes, giving 2 "pulses" per rev.

It is worth revisiting the runout discussion.

The tolerances from Machinery's Handbook's Table 2b = Narrow V-belts (includes 5b) are listed as follows:

• Radial Runout (TIR) - Up through 10.0 in. outside diameter 0.010 in.; For each additional inch of outside diameter add 0.0005 in.
  
• Axial Runout (TIR)- Up through 5.0 in. outside diameter 0.005 in.; For each additional inch of outside diameter add 0.001 in.
  
• So for our 12" diameter fan sheave, this limit translates to 0.011 inch max radial runout, 0.012" max axial runout.
  
• Danny's limits were in the same general ballpark (OD Eccentricity tolerance would be about .010" and side wobble and runout .001"/1" of sheave diameter)

Our results were as follows

• We have 0.011" radial/rim TIR on the sheave (at the 0.011 limit).
  
• We have 0.015" axial/face TIR on the shave (above the 0.012") limit.
  
• We also have 0.010 radial TIR on the shaft. (I don't know a limit but this sounds high to me... and the shaft runout may be responsible for the high sheave runouts).

Also some clarification of my earlier facts - there is one of the four machines (the one with the runout and all the results discussed so far) that exhibits higher 2x than the rest and perhaps a slight increasing trend in 2x over time although I'm not sure.

So it brings some questions:

• 1 - What would you use as a limit for shaft runout?
  
• 2 - Can excessive sheave runout or shaft runout cause high 2x fan? In my posts 21 February 2008 11:05 AM and 21 February 2008 11:50 AM I have argued from theory (rightly or wrongly) that it shouldn't. I think others are suggesting that it can. I would like to understand that better. How does it happen? Have you seen from experience that it does happen? Rusty gave one example directly above where it seems to be the case for a roll. I'm not familiar with rolls, but it seems like there is a mechanism for 2x due to runout on rolls that is not applicable for sheaves: Assume the belt for a roll is very long axially along the shaft. If the "node" of the bowed shaft is in the middle of the belt, then during one half revolution the shaft on one side of the node puts the belt in tension; while during the other half revolution the shaft on the other side of the node puts the belt in tension. That would give 2x, but I don't think it would apply for a sheave.
  
• 3 - How would the shaft get bent? Danny mentioned hammering, but I'm not sure what activities would involve hammering on the shaft (has something to do with the type of sheave QD or taper-lock?). Could it be a result of overtensioning at some point in the past (note slide 10 of the photo's above show bent structural member supporting motor slide rail tensioning screw... suggests possible excessive force.... also I calculated that if we apply the nominal 10 pounds tension for 7/8" deflection, the static belt tension is approx 100-160 lbf resulting in 600-960 pounds on the sheaves).
  
• 4 - Do we need to straighten the shaft? (that would be pretty big effort I think... don't want to recommend that unless needed).

Some of this is a re-hash of previous discussion. I have a short attention span ;-). Please feel free to repeat/reinforce your earlier comments if you think I am overlooking something.

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

Pete,

Yes, that is a QD bushing. I can't tell from the photo, but if the bolts are not pulled down evenly and torqued properly, you can get the bushing twisted which will show at the split in the bushing opposite the keyway. The two sides of the split should form a smooth surface.

There are gauges available to check for sheave wear. Your friendly Motion Industries salesman will probably give them to you.

Hammering on the shaft occurs during assembly most often, usually as a result of poorly machined parts. I tried to balance a fairly large, center hung fan mounted on a 4 7/16" shaft. It had a huge axial component (don't recall for sure, but I think it was 1 x). After much head scratching, I asked about what work had been performed recently and they said they replaced the shaft and they had a really hard time. When the 12 pound sledge hammer failed to drive the shaft into the fan hub, they decided they needed some professional help. That came in the form of a 20 pound sledge hammer.

The hammer marks could be on the shaft, bearings, fan hub or even the bushing (although it probably would have broken).

So if anyone tells you that they can't bend that shaft with a hammer, disagree.

I would agree that you have evidence of past over-tensioning. I have actually witnessed overtensioning kill a motor bearing and have seen several instances of bent motor bases, bent shafts, bearing pedestals, etc. All but one were caused by worn sheaves or belts.

Another thing to try is to make a white mark across all the belts. Watch it with a strobe as it rotates (or not) and you will probably see them slip at different rates. Shut it down again and check to see if the line is still there. It probably won't be because the bend in the motor base makes accurate alignment and even tension unlikely.

Other comments:

Your drive is rated for about 25 hp per belt using 5vx belts which are the norm.

Ultra-V belts and sheaves require higher tension than conventional v-belt drives.

Ultra-V sheaves and belts (may) wear more quickly due to the higher tension and the exposed cording in the sides of the belts. Conventional belts have a cover that makes them less abrasive.

Spring operated tension gauges are also available for about $20 or less. The "guitar tuner" unit that I described earlier was about $500 a few years ago.

Laser belt alignment tools are cheap and very fast.