This is what happens when you design a fan so that the natural frequency of the blade is very close to the blade passing frequency. This is a heavy fan, 3/8" plate for the blades, 3/4" thick gussets, heavy hub, lots of weld which looked to be well done. More Photos

Here is the data. Bottom plot is for the fan running with the new wheel about a year ago. Impact test is for a 'good' blade on the failed wheel; resonant frequency may be a little low since there may have been a little cracking at the blade root.

And here is the waveform from the impact test.

Rusty,

The link to the photos works fine but I can't get to the data. It gives me the ghostbuster sign. Nevermind, It's all there now.

I'll bet that was pretty exciting. I hope you weren't there to witness it. I assume that you balanced out the two giant square head set screws set side by side Now that I look at it again, they aren't even over the key.

Danny,

This type of fan hub usually has a set of set screws over the key and another set at about 60 to 120 degrees from the key. The picture doe not allow us to see if there is a set screw over the key, but there likely is.

Regards,
John J

If clear: look at the welds - or, lack thereof. The flats have no penetration or it appears that the blades pealed off and left flats. Home-made; no doubt and probably not stress-relieved. System excitation although close to a component will fail the weakest link and not necessarily the component. What made the link weak?

quote:

I can't get to the data. It gives me the ghostbuster sign

You have to be "logged in" to view the graphics. Danny, when I bring up this web site, it says that I'm logged in, even though I'm really not. To get to a login screen, I just click the "Reply" button, login, and then close the reply window.

There may be set screws over the keyway, I don't know. Doesn't really matter... the wheel was not loose on the shaft and I balanced it after it was installed last year. This is the wheel shown in the How NOT To Attach Fan Weights photos. This wheel was built by a shop that did not have a balance stand; they sent it to a motor shop for balancing and it came to the customer with the weights shown in the photos. We cut all those off and started over. And, Yes, I do notice that the two failed blades have balance weights on them, but I don't think that had anything to do with the failure.

I'm not a welding expert or metallugist, so what are your opinions of the welds on this wheel? They look to be well done, but the penetration looks inadequate to me. These blades just peeled off.

Still, I've seen full penetration welds fail when subjected to high-cycle fatigue as was the case here. Sam's comment about stress relieving (heat treating, right?) is interesting. Is that usually done on a fan wheel like this?

No, I wasn't around when this happened, but a couple of maintenance guys had just walked away. I assume they were out around it because it must have been shaking the world just before the blades flew off. It usually takes about 30 grams to balance one of these, so I can imagine what the imbalance must have been, and how bad it must have been shaking.

I wondered about the penetration of the welds.
I've seen two fans that had blades break from poor penetration of welds when a mig was used. If they were made with a mig and by machine (or human for that matter) they may look real good, but not have any penetration at all.
It's easy to run a nice bead with a mig, but difficult to get good penetration on a 90 deg. angle like those welds are. I pine for the good ole days of stick and chipping slag.

D

Dave, good point about mig vs. stick. I'm not much of a welder, but I've wondered about that myself. Seems it's a lot easier to get high heat with the much larger electrode of a rod vs the relatively small wire of a mig.

quote:

natural frequency of the blade is very close to the blade passing frequency

In the stationary world we see blade pass. In the world of the rotating fan it is something else. I am puzzled.

Those welds look like they were made with smaller rocket rod and cold. Notice the hump and roll. Also a weave to cover extra surface not necessarily deeper.

But, MIG or TIG torch in the hands of the right weldor may see deep penetration.

I have my opinion on the welding but the issue is this rotor.

Stress relieving is not necessarily 'heat treating'. Although 'heat treating' is stress relieving.

Yes you can stress relieve by vibration, actually works and some machines do that naturally.. and it may not give the same result. It can be used for large size things where a furnace is not around. OLI

The welds do appear to be made by a competent welder however it does seem that the proper welding procedure was not in place. The metals possibly have different chemical content such as carbon levels (4140 4160 ect.) and in these cases preheating and other special procedures should have been in place.

Yeah: I saw lack of bevel and overlay with hump. When they saw that, they should have had a hard look.

A preheat is necessary on many metals.

I find it interesting in a Vibrations Forum that there is more discussion (actually gross speculation from photos) about the welding and heat treating than there is about the vibration data presented!!
The "Bump Test" revealed a natural frequency at 13,105 cpm, and that is near blade-rate pulsation at 1683 rpm x 8-blades = 13,464 cpm. The Bump test was on the broken fan rotor after removal, so some difference in natural frequency could be expected from testing a undamaged rotor in-place on the shaft and on the fan bearings. Apparently the new rotor was not Bump tested at all, so we don't know if the same problem exists or was corrected. Vibrations were low with the new rotor operating, but that doesn't indicate anything about the actual blade vibrations. Blade natural frequencies can change with speed (gyroscopic stiffening) and temperature (metal strength), so the actual natural frequenies are different from the cold shut-down Bump test. Finally, as mentioned by Bill, the rotor is a rotating coordinate system, so a blade natural frequency would have to be equal to Blade-rate +/- shaft speed to be excited by blade-rate pulsation frequency. All things considered, the failure could have been initiated by fatigue from resonant vibrations. We could speculate that the replacement rotor could fail again, unless the design was modified or manufacturing (welding and heat treating) procedures were changed.

Walt

Repeated blade forces can add at the rotor to a single force as indicated by Walt. However, just looking at a single blade or point on the rotor will show what a blade sees. If there are m diffusers then the blade sees a repetition rate of m per rotation. If only 1 then the fundamental is at 1x with harmonics for the individual blade.

For a stationary point one gets the number of blades times rotation.

The destruction process was probably quick and likely was missed in a survey type of data collection.

Therefore, I wonder if blade resonance would manifested itself in the vibration data if it was read continuously?

Picture an individual blade which goes into resonance while passing the diffusor. In other words, in a certain space location a passing blade mass starts oscillating in tangential direction due to air pulse excitation.

Question: Would it produce a radial dynamic force sensed on the bearing?

Dave

Bill,

I did not say or imply this:

quote:

Repeated blade forces can add at the rotor to a single force as indicated by Walt. However, just looking at a single blade or point on the rotor will show what a blade sees. If there are m diffusers then the blade sees a repetition rate of m per rotation. If only 1 then the fundamental is at 1x with harmonics for the individual blade.

Blade vibration is a global property of the fan rotor. A single blade gets a pulsation force at 1xSS as it passes the cut-off, but it also sees the cumulative pulsation forces of all blades passing the cut-off. The issue is not whether the blade has a NF at 1xSS (usually not), but whether it has an NF at BPF +/- 1xSS. I'll dig out a technical paper that proves my point.

David G,
Would it produce a radial dynamic force sensed on the bearing?
Not much vibration is radial direction on bearings from tangential blade vibrations, until a significant crack develops that alters unbalance at 1xSS. More likely to see tangential blade vibrations from torsional vibrations; that hardly anybody measures!

Walt

(Walt, this is not in response to what you posted... we were apparently writing at the same time).

Dave, these fans are not routinely monitored. Initially, I checked them several times a year, but as problems subsided, they had me come less often, and now it's only when they have a problem (well, until this happened). This is a small, marginal plant; they were within a week of being closed down as recently as November.

A friend who is a welder confirms what Sam suspected. The hub was probably a hardened material and they should have used preheat. No beveling produced lack of penetration. You can actually see this at the hub where the welds just peeled off. Failure probably started at the hub welds, and when those let go, the centrifugal forces just peeled the blades off of the backplane.

As to resonance/vibration: these fans, like many paddle fans, operate with a very high blade pass vibration. This can often be seen at the bearings, but the housings will often have hellacious vibration, usually requiring additional stiffening to prevent the housings from breaking up. To me, action/reaction dictates that the high forces on the housing equals equally high forces on the wheel. Essentially, as a blade passed the cutoff, there is a "pulse" which really just acts as an "impact" and excites the blade resonance, regardless of what that frequency is, or what the fan speed is. Notice on the impact TWF the long ringdown; this was with a very light tap from a rubber hammer. You could actually "feel" the blade ringing. Remember, the only thing that failed initially was the fan wheel.

Going out on a limb here, I will assert that these blades actually vibrate at their resonant frequency, continuously. This is a function of poor fan design, possibly endemic to paddle fans in general, and perhaps accentuated in this particular fan because of lack of engineering. As Sam said, "home made" even if done by a commercial company.

The fact that blade resonance frequency is proximate to blade passing frequency probably doesn't help, but is probably not a primary concern here. So I don't think that just changing the operating speed is really going to help (as I originally thought). If the original wheel (which is in place now, and which was removed due to a suspected crack, and which the owner says is NOT there now) was made by the same shop, using the same procedure, then it will likely fail as well.

This is another example of what I call "convergence"... a poor fan design, coupled with a poor wheel design (many of the paddle wheels I see have a stiffening 'ring' on the inlet side), coupled with improper welding procedure, "converge" to produce a failure. If any of the three conditions had not occurred, you have no failure.

That is my take on this, and of course, I could be completely wrong. But given the information I have, that is what my conclusion is.

Reading this thread, I realize I have just repeated much of what you guys have said.... sorry if I gave the impression that this was original with me.

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

Walt,

quote:

Blade vibration is a global property of the fan rotor. A single blade gets a pulsation force at 1xSS as it passes the cut-off, but it also sees the cumulative pulsation forces of all blades passing the cut-off.

A blade may see the other blades. The extent results from the transfer function. I have seen strain gauge tests with blades seeing what I have said, i.e. the number of disturbances created by the stationary components. Blades can act as if they are isolated, which in cases they are by the boundary conditions.

I have seen failures that resulted from regularly spaced stator created forces on the blades. The strain gauges said, "number of stationary disturbances."

In fact if the forces added the same from each blade pulsation to an individual blade only phased, the forces would sum to zero at the blade due to the phasing.

The forces and vibration that sums on the rotor is another story. In this case it is fair to assume that each blade contributes similarly with only a phase difference and a geometric angle. For the individual blade and one stationary pulsation source, the blade sees a 1X individual pulse to a different location to the other pulses coming through the shaft into the base.