No, its not a myth. But the design difficulties associated with it are very dependent on what type of planetary  it happens to be. The most demanding types are those involving compound planets, where each planet is a cluster. In such cases, the clusters must all be made such that the angular tooth timing between their constituent gears is the same for each planet. There are many design tricks used to reduce the criticality of this - for example, making use of axially floatable helical gears, or locking the gears together after assembly with friction devices or castable resins. Even in a planetary with simple planets, the ring gear and/or sun gear are often made so that they can float radially, thus improving the load sharing. So it all depends on a combination of the quality of the gears, the type of planetary, and the degree to which the designer has dealt with potential load sharing issues.

English Muffin, thanks for your comments.  If I read you correctly, a pair of rigidly connected helical gears of opposite hands will act like a herringbone gear, but they would be easier to produce.  The gears may be of different sizes if differential action is required.  This compound gear will shift axially on its spindle until it begins to take load.  Then it will stop shifting and carry the load.


I suspect that friction devices for setting up the gear timing might be somewhat complicated and expensive, and  castable resins might not be able to carry the full load, at least for very long with high reliablity, in the larger gearboxes.

However, I am not seeing how any of these methods would improve load sharing between the planets.  It seems that the first planet to take load would remain more heavily loaded than the others, just as it would if spur gears were used.  Some method for improving the load sharing would still be needed, such as floating sun and planets as you mentioned.
  
I have noticed that the floating technique is often used in smaller planetary gearboxes, say under 50 hp.  Would you use it in units rated 2000 hp and above, with the high speed shaft turning at 1800 rpm or higher?  If not, and the system must be fixed, how would you achieve good load sharing?  It can be improved in a fixed system by using higher quality gears, say AGMA 13 when AGMA 10 would be good enough otherwise, but that can be expensive and may not completely solve the problem.

Regarding cost, I understand that the expense of a ring gear increases much more rapidly than the increase in its size, because of the difficulty of heat treating larger ring gears.

Suppose that, just by chance, load sharing is good in a new planetary unit of the fixed design.  Would it deteriorate as the gears and bearings wear, if no method was used to adjust for the wear?

The Niigata Co. uses a bending planet spindle to achieve good load sharing in large planetaries (www.niigata-converter.co.jp/NICO e.l pg.htm).  This appears to be a new idea, and I am wondering if anyone has heard anything about it.  Forty years ago an Austrian inventor, Felix Fritsch, patented several mechanisms for balancing the planet loads.  These were licensed by Cincinnati Gear, but they apparently proved to be too complicated for commercial use.  By the way, Cincinnati Gear has gone out of business, apparently a victim of the Enron collapse.

On a related subject, most of the gearboxes in the largest of the current wind turbines (2000-5000 hp)are planetary  for the low speed stages, with a single-mesh parallel-shaft arrangement for the high speed stage.  I am only speculating on the reason for this, but unequal load sharing in a high speed planetary stage might cause excessive dynamic loads.

Thanks to all who comment on these questions.

Let me clarify a few things. First of all, a good place to see what I am referring to when I say simple and compound planetary is machinery's handbook - there are three pages of different arrangements in most editions, and the figures below refer to these. Now suppose you have a typical simple planetary, such as that shown in figures 9 thru 12. Suppose further that the sun and ring gear are made perfectly and are perfectly concentric. Now, if for example you place three perfectly identical planets in the space between the sun and ring gear, the load sharing will be very dependent on the positional accuracy with which the spider is bored for the three (equispaced) planets, because each planet has two meshes. If you are using helical gearing, whether herringbone or not, it will not of itself improve the load sharing situation at all, although frequently herringbone gears are used in planetary boxes because they eliminate axial loads on the bearings and can remove the need to have any axial thrust bearings in some cases. Allowing the sun or planet shaft to float radially will help the situation, and is not new -the idea of having a flexible planet carrier shaft has been around for ever as far as I know, and this feature  often occurs automatically to some degree because bearings are not infinitely stiff, and most designs have an overhung bearing arrangement of some kind. Of course, you have to be concerned with angular misalignment as well, so the amount you can get from this is limited - depending on the geometry. Now suppose you have a compound planetary such as that shown in figures 16 thru 18. In this case, each gear only has one mesh so the situation is different. Here, you have the problem of angularly aligning the two gears making up each planetary cluster, such that their relative timing is the same in each case. As I said in my previous reply, there are a number of ways of achieving this, and I don't claim to know them all. One way is to just make the gears very accurately in matched sets. Another is to make the two components of the cluster separately and jig them together, locking them with friction devices or sometimes castable resins with crude splines. For this type of planetary, you can make use of some properties of helical gears. What you can do is arrange for the leads of the two members of each cluster to be very slightly different, such that there will be an axial force produced - this usually means that the helix angles will be the same hand -not opposite (you have to think about this a bit). You can then have lightly spring loaded thrust bearings on each cluster so that the clusters can axially achieve an equilibrium position. Of course, you have to figure everything out so that the axial travel does not exceed some limit at maximum load.
There are also all kinds of assembly issues in this type of box - it is very easy to mis-assemble such a planetary and sometimes load sharing issues can be traced to that. I hope this clarifies what I was saying. And bear in mind - all planetaries do not have to have ring gears, or even more than one planet!
Regarding the reliability of planetaries, I would say that they can be one of the most reliable designs made - one of the most unreliable things in gearboxes is often the bearings, and the balanced design of most planetaries can help things in regard to bearing loads. And of course, as you probably know, most American cars have planetary transmissions.
Regarding the wear issue, I can't comment from practical experience, but my guess would be that modest wear would help if the load sharing was bad to begin with. However, in my opinion, in a properly designed and lubricated gearbox, gear wear due to sliding should be minimal. What you are more likely to get is pitting due to subsurface shear, and of course bearing failure, where the catalog life ratings are also based on that same thing occuring after a given time. I am sure a lot of people will take issue with this, but that's my view anyway.

Regarding some of your other points -

Yes, ring gears can be expensive - particularly if you want to grind them.
There is a French company called Andantex which makes cluster gears by connecting them with castable resins. Some of their deigns are not true planetaries - they are just multiple load sharing designs. And yes - they do have problems with this on occasion, although they have been doing it for decades. I did not mean to imply that I endorsed this method, only that it is done.
I have used the friction clamp method myself in planetary designs with success, although in that case it was combined with a spline and helical compensation - it is not necessarily expensive (see Ringfeder shrink disks).
For quantity production designs the lowest cost method would be one piece, although you can run into manufacturing problems with close coupled clusters, especially if you want to grind them.

Thanks for your detailed reply.  The Ringfeder Shrink Disc is interesting.  I have seen taper lock devices before but none that dispensed with the key.  Apparently the shink disc can handle as much torque and shock as any keyed assembly.  That and the timing ability make it a very useful device.

I looked at the planetary arrangements in Machinery's HB and have a better understanding of your comments.  Cluster gears must be of the same hand to obtain opposing axial thrust because one is driving and the other is driven.  This is not like a herringbone gear.

I have several questions about designing a 5000 hp planetary speed increaser for a wind turbine.  It would be of the type shown in Fig. 12 of Machinery's HB, 26th Ed.  It would be three stage, three planets per stage unless a greater number is feasible, 1:90 ratio, 1800 rpm output, 20 year life.  Can the sun and planets be allowed to float in such a large unit?  I am thinking that the parts might be getting too heavy for that.  Dynamic effects might start to become serious, especially for a unit that must last for 20 years and operate as quietly as possible.  (Any comment on my previous question: why a parallel shaft is often used for the high speed stage of a wind turbine gearbox, even though the low speed stages are planetary?  This is true even for wind turbines with only one generator, ie, only one gearbox output shaft.)

If floating is not allowed, how much should the planet gears be derated for unequal load sharing?  I have heard the rule of thumb that the number of planets should be reduced by one when calculating the power rating of the stage.

Should the planets also be derated for reverse bending in the teeth, as recommended in ANSI/AGMA 2001-B88?  I have been told that the sizes of the sun and planets have something to do with this.  If the sun is relatively much smaller, then it becomes the critical member and there is no need to derate the planets for reverse bending.  But when the sun and planets are about the same size, reverse bending must be taken into account.  Derating factor is 0.7 in that case.

Would a planetary gearbox of this kind be as efficient, and as quiet, as a parallel branch type?  Thanks for your help on this.

Comments

I have never designed large drives, so cannot claim to be an expert in this field. If I were going to do this, the expert that comes to mind is Ken Gitchel, who used to work at UTS in Rockford, and was behind most of their gear software, but I believe he has passed away. But the president, who is an Indian by the name of Jack Marathe, knows a lot of people in the industry. The other expert I would consider is Ray Drago, who used to be the gear expert at Boeing Vertol, (and still may be for all I know) – but he also did consulting on the side, and probably still does. Another great expert on Epicyclics in the UK was D.B. Welbourne, a professor at Cambridge University. He has retired, but I think is still alive and might be contacted through the I. MechE.

      Here are some other comments carefully restricted to things that I know something about :

1.    When I am designing drives, I always avoid keys if at all possible, and try and use Ringfeders or some equivalent device. They eliminate fretting, simplify manufacture and assembly, and introduce a lot less stress concentration. These products have been around for a very long time. Originally, Ringfeder had just two products – tapered locking rings and a double tapered device, which they still make. This latter device I would avoid – it has very poor concentricity characteristics. The shrink disc, on the other hand, produces superb concentricity.  But bear in mind – it will slip if you exceed the rated torque. You must know what the worst case shock torque is.

2.    The main advantages of a planetary are space and weight savings. When you get to big drives, a planetary  may be the only possible solution – for instance, when the dimensions of a parallel offset drive would be so great that the gears could not be manufactured or transported. So they do tend to get more attractive for very large drives – but I don’t think you can generalize in all cases. Also, you may actually want to have an offset in the drive for some other reason – such as providing an auxiliary take off point for something.

3.    If I were designing a large drive, I would keep the number of stages to a minimum and use as many planets as necessary to get the capacity I needed. Why restrict yourself to three? Of course, with more than three planets, you cannot load share by having a floating sun or annulus. But large drives often make use of another method  - they use elastic location  of the planet journal studs.  One of my references says that with a 2:1 ratio you can have a possible maximum of eight planets. This of course also takes care of the weight problem you mentioned.

4.    I have seen the rule of thumb that you refer to about number of planets and load sharing. I’m a bit skeptical about rules of thumb unless I understand what’s behind them. Personally, I would say that the extra cost of providing for perfect load sharing was well worth it in a large and expensive planetary. If you do not do this, you should at least go through a tolerance stack up/deflection analysis to see what loads you are really getting.

5.    If you are designing a drive, the planet bearings need careful consideration. If you are going fast, you should not use bearings with a cage which rides on the rolling elements, because of centrifugal effects. Use a bearing with a land riding cage, or no cage at all. “There’s not a lot of people know that” as Michael Caine is so fond of saying.

6.    The planet teeth in your proposed arrangement will see reverse bending, as in most idler gear situations. Whether it matters or not depends on whether your gear teeth turn out to be limited by bending/fatigue strength or pitting resistance, or indeed, as you say, whether the planet pinion is even the critical gear in the train. There is a lot of general useful information in ANSI/AGMA 6123-A88 – Design Manual for Enclosed Epicyclic Gear Drives - and there are a lot of references in the Bibliographgy  about dynamic loads in planetary gear systems – perhaps that might help you.

Thanks for your excellent comments, EnglishMuffin.

For low velocity drives such as utility class wind turbines, the floating sun provides a nearly perfect load dividing mechanism. The catch is that this element must be connected to the next stage, or output shaft, which is on true center. If this is done with a single crowned spline connection, the sun will pivot slightly which must be considered in the rating.

Nearly all gearboxes of this size are using helical planetaries, due to noise constraints. Consult AGMA 6006 currently in draft form for details concerning wind turbine gearboxes.

There is a rewrite of the AGMA standard for planetaries under way at this time. The question of load sharing is in debate.

With hi accuracy gears and shaft locations, designs with floating suns are using factors of unity and this is being accepted by reviewing angencies. On the other hand, don't forget the effects on mesh derating due to the pivot action. This can amount to 30-40% unless very good design practices are acheived.

Planets must be derated 30% for reverse bending if present.

Everything above is based on 3 planets. As soon as more than 3 are used, load dividing becomes a spring-mass displacment problem. In 4 planet and greater systems, there is always a derating factor and the analysis is complex.

GearmanPE, thanks for your valuable comments.  I have asked AGMA when the 6006 standard will be available.  If planetary gearboxes must be severely derated for problems inherent in their design--load imbalance or pivoting sun gears, and reverse bending of the teeth--and if the highest power density versions are difficult to design, wouldn't a multi-branch parallel shaft gearbox be better, that is, lower cost and easier to manufacture?

I am still hoping that someone will comment on my question concerning the high-speed stage in utility-scale wind turbines.  It is almost always a parallel shaft, while the low speed stages are always planetary.  Why not a planetary all the way through?

Also, if single helical gears are used in a planetary gearbox, how hard is it to handle the thrust?

I wouldn't say that planetary boxes have to be "severely" derated. I also believe it is possible to have a design with elastically straddle mounted planets that eliminates the tilting problem, although these need to be derated somewhat for load sharing. Also, the cost of a gear is not necessarily directly proportional to its strength - although it depends on the volume of manufacture somewhat.
The difficulty of handling the thrust in a planetary with helicals is basically the same as handling the thrust in a conventional box. It is most commonly done with angular contact ball bearings. But, on the planet pinions, for the type of box you are envisioning, the axial thrust is very low. As I think I mentioned before, if you have significant centrifugal loads on the planets, don't use a cage which rides on the rolling elements. Sometimes, in small planetaries, people just use plain thrust washers.
I suspect the answer to the parallel shaft question may have to do with efficiency and cost, but I could be wrong. As I said before, the main advantages of a planetary are size and weight savings - if you are not getting much of those, there's no particular advantage - unless you have to have an in-line drive for some reason. (The high speed stage has much smaller gears presumably).

EnglishMuffin, maybe I do not understand these derating factors well enough.  Suppose the planet gears are derated 30% for reverse bending according to the AGMA standard.  If the sun gear is allowed to float by pivoting as described by GearmanPE, let it be derated 30%.  (If elastic straddle mounting of the planets eliminates the need for a floating sun gear, what would be the derating factor for unequal load sharing?  I suppose it would be in addition to the derate for reverse bending.)  Now, if the aspect ratio of the gears is held constant, then by a rough calculation I find that for a given rating, the gear diameters must be about 15% greater because of the 30% deratings.  This will increase the diameter of the gearbox by 15%, and the length by maybe half of that.  The volume of the gearbox would then be about 40% greater than it would be if the gears did not require derating.  I suppose the weight would also increase by nearly this much.  This increase in size and weight (severe?) is due solely to the problems inherent in the planetary design, so far as I can tell.  These problems are presumably not present in a multi-branch parallel shaft design.  Wouldn't it be worth looking at here?  It also has in-line input and output shafts.  I find that a six-branch parallel shaft gearbox would be about 30% smaller and lighter than a three-planet planetary, and it should be easier and cheaper to manufacture.  If the number of planets is increased and the the load sharing becomes harder to determine, and a floating sun cannot be used to counteract the problem--too many planets as you say--then there might not be much advantage in using more than three planets, although this apparently has not deterred the Niigata company.  They are using as many as eight planets, see reference in my earlier comments above, although I have been told that their flexible pin mounting might not be a complete solution to the load sharing problem.  Is that what you meant by elastic straddle mounting?  There are two reservations about flexible pins that I know of--fatigue of the pin and gear misalignment, although Niigata says they have mostly eliminated the latter problem.

Just a quick response - as I read him, GermanPE does not say you must derate by 30% for pivoting - only that you might have to. It all depends on accuracy and design details - in a lot of cases I don't believe that you would have to derate this much. Its impossible to say exactly without a detailed design to study. If you did what I am suggesting, with the straddle mounted flexible planets, you could use extra planets without having to increase the length of gears or the box, although this type of design is usually only seen in very large planetaries. With this system, the derating amount would depend on the accuracy of planet location, from which you could compute the worst case load sharing from the spring stiffness of the mountings. Usually, when you want to increase the strength of a gear, you first look at making the face width wider - not increasing the diameter. That is why I said that gears don't necessarily get more expensive directly in proportion to their capacity - it depends on set-up versus machining time. As far as reverse bending goes, bear in mind that when you design gears, you have to run two calculations - one for bending strength and one for pitting resistance. Depending on which dominates, you may not need to derate for reverse bending at all. It is also sometimes the case with helical gears that the face width is determined from considerations of tooth overlap, so there may be headroom anyway. The whole subject is far too complicated to generalize about in simple terms - you might have to rough out two designs in detail and see how they compare.

One other thing - exactly what do you mean by a six branch parallel shaft box ? Does this also involve load sharing ?

EnglishMuffin, please contact me at georgetf@sssnet.com, and I will send an Autocad drawing of the six-branch parallel shaft design.

Well, it looks as though your six branch train is certainly much more compact and lighter than a conventional train, which is no surprise. I am not sure how you calculated the weights though - did you have webbed gears or solid ones ? I don't think you are the first to propose or use such a multi-branch design for a wind turbine. The saving in weight will have cost saving implications for the rest of the structure of course. I would not like to say off the cuff whether it is any cheaper in itself versus a conventional train - you do have the extra cost of the load sharing mechanisms. One way you can do this is with helicals having slightly different leads, as I indicated in one of my earlier posts, but you also have the timing issue to contend with, which can add some cost. You may get the economics advantage of larger quantities of smaller gears - large gears are expensive and there are heat treat issues. I don't disagree with you about multibranch designs versus planetaries - I think they frequently are cheaper, although I don't think they are always quite so compact - I believe you can get a bigger ratio into a slightly smaller space with a planetary, in spite of what you say, and you don't necessarily have the timing issue to contend with, so in something like aerospace I believe they would usually win out. But in your case, ultra compactness is not so important. Since this is a speed increasing drive, the gear tooth design needs special consideration. I also believe you should try and balance the design so that the speed-up ratio in any one mesh is as small as possible.

Regarding the question concerning the high-speed stage in utility-scale wind turbines being almost always a parallel shaft, whilst the low speed stages are always planetary.  Why not a planetary all the way through?
The dominant Danish designed wind turbines require a lead through the gearbox into the rotor to control the pitch of the blades etc. The last stage of gearing is often offset so that these leads do not have to pass through the generator.

Frostie - sounds like a good point - wish I knew more about wind turbines.

Frostie is correct, although concentric stages can also allow a conduit through the sun pinion. In any case, the complexity of planetaries pays for itself as torque increases. They provide less cost advantage in HS stages and are difficult to produce with high helix angles.

Noise is a huge concern, so the prefered configuration uses helical stages in the higher speed, lower torque sections. These helicals have sufficent tooth overlaps and extreme accuracy to keep them quiet.

As the size increases, the use of 2 planetary stages or compound stages starts to look good. Other split power path designs have been used sucessfully.

My comments regarding tilt refer to the derating factor for imperfect face contact. If manufacturing variation is to be included, it is difficult to justify KHbeta factors below 1.25. A proper analysis requires 3D mesh analysis.

Concepts involving floating pins or elastic mounts have been around for years. With a life requirment of 200,000 trouble free hours, these solutions are greeted with due skeptism. To be comercially viable, new designs require a lot of testing and field experience.

Thanks for the comments on why wind turbine gearboxes are not planetary all the way through.  I could not find this information anywhere else.  All other comments are also good and hard to obtain elsewhere.

Suppose a two stage planetary gear with three planets per stage has fifteen moving parts - bearings and gears.  Would it be substantially less reliable than an equivalent single branch parallel shaft type with only ten moving parts?

Is accelerated testing of gearboxes an acceptable practice?  I mean overloading them, and maybe operating them at higher than design speed, to cause quick failure.

An answer to question 1 : It's impossible to predict reliability based on number of parts alone, athough the part count is often used, rightly or wrongly, as a measure of design quality. This brings to mind a comment once made by Werner Von Braun, when he was asked how something as complex as the moon rocket for the Apollo mission could possibly be reliable with so many parts involved. He replied that in his garage, he had a car and a lawn mower. The lawn mower was much simpler than the car, but the car started every time he pulled the starter, and the lawn mower didn't.
An answer to question 2 : It all depends on how much money you have available !

I have designed both "multiple path" and planetary gearing for turboprop gearboxes.  My guess is that simple gearing is used for the first stage or two, and the planetary used for the final stage, based on gears being sized by torque. The planetary is used for the high torque part of the geartrain, and standard parallel axis gears are used for the first stages where torque is low, and consequently size is low. Most of the weight and size savings occurs at the high torque part of the geartrain.
For interrupted and very low speed operation, rolling element bearings would be a good choice for the planetary. The choices for gearing would then be high contact ratio spur (noise reduction) or single helical gearing. I have not found anyone happy with double helical gears/rolling element planetaries (two many floating members for load sharing creates dynamic problems)(double helical/journal bearings seems to work well). As the axial loads from the sun and ring mesh for single helical planet would be equal and opposite, a pair of cylindrical bearings would be used to react the resulting moment. A nice feature of the planetary is that there are no resulting radial or tangential gear reaction loads on the bearings on the input or output shafts.
With precision ground gearing, and correct means of load sharing, there really is no reason to derate a planetary by more than a few percent. Usually the planet bearings will size the unit, not the reverse bending of the planet gear. Clearly the planetary is smaller and lighter.  
Cincinati Gear was developing a unit for Enron that used a planetary. Their Chief Engineer, Octave Labath, is now a consultant. He is listed on the AGMA website, I believe he could save you many years of development problems.