To Exist or Not to Exist

I don't think this is something which should be governed by blind imposition of policy. The designer should use their judgement to decide whether the feature is significant, or whether it can be simplified without negatively impacting the value & utility of the model.

Designers who consitently exhibit good judgement should be promoted.

High fidelity or simplified geometry? Isn't asking for a one-size-fits-all answer being a little simplistic? What serves the practical purpose(s) of the model?

Term paper?

Oops, passive attribution ('it has been proposed to me'). Naturally, we'd all like to know who the "proposer" is; if it be one of us, feel free to say so. I've had my go rounds with Mark on this issue. Maybe I'm the "proposer" or maybe Mark was just reaidng my mind as I just thought, today, that getting the issues we were discussing into a larger forum would be beneficial. Doesn't seem likely, though, does it?

Good idea to broaden and widen the forum but can't find the discussion/topic on any of these. Maybe you could be more precise in referencing these locations, IOW, the locations where you actually posted the discussion. Why would you make us hunt for it if you really wanted us to participate?

Sorry, no 'moderators' here; stay posted

Yes, they exist, they are very commonly used instead of the more realistic looking helical swept features. Shorter regen times are a strong factor in their use; a stonger one is that those of us who have, at one time or another, used these advanced modelling techniques, have discovered the WTFC zone, that place where everyone, from users to in touch management, says 'what does this buy us' and finds a void.

I think the audience here is reality based, experienced and quite practical. The sentiment is against absolutist rules. It favors the contextual, the local, the skilled user and the application of a broad brush to particular applications. In this case, (gearing) the application is so exclusive, that outside input (from Pro/e or any solid modelling software) is difficult.

I think I'll repond a little later on the issue of substance. But, hey, thanks for getting the ball rolling. You read my mind.

David Janes

If it truly serves no purpose, I see no reason to accurately model a component. But it often does serve a purpose. If the threads on a screw make it into a work instruction illustration, or the hex recess on a SHCS into a presentation rendering then they're useful because they prevent confusion when people view these things. A BHCS with no details may make the mass properties work out, or the BOM correct, but they look a lot like a rivet. But accurately modeling a gear, when a simpler tooth profile would look basically the same is probably a waste of time. Unless you're using it in an analysis or manufacturing it.

Threads modeled as a simple revolve (no helix) look good enough most of the time.

Sometimes I'll make hollow parts solid, like blow molded or rotationaly molded parts, to keep assembly complexity down.

On the flip side, sometimes I'll temporarily add small fillets to components when, in reality, they're sharp. This creates an edge highlight in presentation renderings that helps the picture read.

Is it just me, or has the power of computers over the last few years not kept up with the power needs of Pro? Feels slower than it used to...

When you visit the zoo, "Don't feed the animals." When you vist the newgroup, "Don't feed the egotist."

One doanhaftabe campy ta go campin' GA

Better ~ hobby, gear design theorist striving for the perfect tooth form. And, apparently, without ever considering AGMA standard tooth crowning, the sine qua non of gear making. So, perfection in the abstract. Perfection without ever touching, designing, making, or testing a real gear.

David Janes

Put in this reasonable way, it doesn't seem controversial at all. Yet, we've been having one over, not so much the general issue of "blind imposition" vs "judgment", but over which products or geometries qualify for "simplification" or being made less than perfect.

Two recent examples: screw threads and gear teeth. The latter sparked Mark to start this thread. Realism was the test in both cases. The thread issue, briefly, was about a more realistic way to terminate a blind thread on a screw as it reached the head, "like single pointing" was the test of realism. I pointed out a bunch of different manufacturing methods, some of which I'd done myself, e.g., that I had chased threads on a screw machine, single pointed them on a variety of lathes, milled them and ground them. Of these, only the screw machine could be considered a mass production method. Yet 99% of fasteners have rolled threads. And none of these different methods of thread production terminate with the same geometry. Of the 1% produced by some form of thread cutting, not one-thousandth are produced by single pointing. So much for the realism of the "single point threading" test. But it does point out the difficulty of having discussions of modelling and feature creation in a design vacuum.

Advocating in favor of simplification (cosmetics on screws, hole feature with cosmetic threads), I pointed out that most people didn't have time to mess with a helical sweep or the huge regeneration time or debugging the failures, especially not when you get hole data, useable in a callout, virtually for free, plus 95% accurate geometry. And you can waste (my prejudice) 90% of your time trying to get that last 5% "perfect". I put perfect in quotes because of the flaws in the reality tests. The "reality" being tested to was NOT that realistic. But I think this didn't settle the question as Bruin presented a 3 point list of reasons to use helical swept threads. I didn't reply. Still, Bruin's remarks addressed some exceptional circumstances, not the common engineering one: why does one need fasteners in the assembly at all? Is it something fancy, such as, to do stress analysis, Failure Mode and Effects Analysis, some kind of analysis on the assembly? No, it's to principally get part numbers and quantities into a BOM for ordering; secondarily, to get these items into exploded views for assembly instruction (where do which fasteners go); thirdly, to do interference checking, especially on lengths of fasteners and potentially, on body size, the latter being made impossible by the inclusion of cut threads in hole/fastener (no, please don't anyone tell me they actually screw the fastener into the hole to assemble, the ONLY condition for their NOT interfering.) In the end, Jeff commented that, as the power of computers grows, we seem to be expecting more out of this process and invent new demands on the models. Anyone remember the drafting board/ACAD days when a screw thread was represented by a cartoon, a thread symbol (alternating, parallel long and short lines)? BTW, if you are seriously worried by an accurate weight of your fastener, DON'T depend on the accuracy of your geometry or the density of whatever convenient material you could find. Get a thousand screws, weigh them, divide by 1000 and assign this MEASURED weight in your MASS PROPS dialogue. This is the only really SERIOUS way to do it. Then compare it to the calculated one. And multiply the difference by a million such fasteners in a Boeing 787. Aren't we talking about how engineering gets done? isn't this how engineers work? aren't the models just rough, more or less accurate representations? and doesn't the level of disappointment over the difference between the model and the real thing depend on the EXPECTATION of absolute fidelity, unwarranted though it may be? Well, that's my position, anyway.

On to the gearing controversy or does the geometry of gear teeth need to be really accurate? The question that come to my mind anyway is why, what drives this issue, what need exists for very accurate tooth geometry? The real difficulty in discussing this issue is that, though people have found uses for solid models of gears, the gearing industry (design, analysis, testing, production of gears) has no use for those models. Take bevels, for instance, which I know fairly well since I was in charge of the bevels department at Allied Gear in Chicago for 5 years. Good gears depend on the setup of Gleason bevel gear machines. The parameters for setting up these machines are generated by specialized gearmaking software. The design of those gears depends also on highly specialized design software that can take some general torque, load and time numbers as well as some other variables, such as how smooth the transmission of power ought to be, and give recommendations on diametral pitch (smaller is better for smoothness), gear ratios, and come up with diametral pitch, pitch diameter, pitch cone angle (bevels), FA and RA. And, as an extra added bonus, you get the mounting distance of each member of the bevel gear set. The output of this software, given a Gleason, is gear ratios for cradle roll, workpiece roll, mounting distance, pitch angle and tool stroke length. The Gleason tables also give you a rough number for crowning length/location ~ it can be lengthened/shortened and moved up/down the tooth face. No model, however theoretically perfect, has anything to do with this setup which depends heavily on lookup tables based on experience gained over decades of use. The models in Pro/e are interesting visual references; they provide the WOW factor. But nothing in the manufacture of gears depends on them. No real gear will be more or less accurately made because of the more or less accurately modeled Pro/e gear. In fact, no practical use of the perfectly modeled Pro/e gear has yet been demonstrated. And, if one took into account manufacturing inaccuracies, the current methods of gear modelling are already about as accurate (less tooth crowning) as the manufactured ones.

Finally, the "who cares" factor: I have the perfectly modeled gear to put in an assembly. I try to do animation or even, mechanism design in Pro/e ~ how does good or bad gear geometry come into play? Sorry, no, it doesn't. No dynamic analysis in Pro/e. In fact, the way you set up gears to run together in MD is to pretend they are rollers, with pitch cone geometry, rolling together, cone face to cone face. No tooth accuracy required here. How about stress analysis, also highly unlikely in a generic program like Pro/e. Well, it is static, depends on what you can define as contact surfaces. Certainly, the usual deflection will not enter into the calculation. And there's some other reasons not to trust its calculations: no tooth crowning as in the usual bevel gear so possibility of appearance of interference at toe/heel and so analysis of likely failure. Not so in real crowned tooth bevel gear sets which cancel the effects of deflection. Conclusion: one does not wish to do stress analysis, static or dynamic, with the "best" Pro/e models. One should do this with the specialized software to this purpose, invented by the gear manufacturers. (Ask AGMA). IOW, I don't give a crap about a "perfectly" modeled Pro/e gear, it is a worthless, pretentious waste of time.

Here's hoping that I've stated my position clearly and unambiguously. Especially since no one seems to want to acknowledge there exists a debate or a debatable issue.

David Janes

Interesting. You should repost under a meaningful subject line (i.e. Essay on the Practical Considerations of Modeling Detail) instead of this Existential poohpooh one. Someone may stumble across it and find it enlightening.

It's been kind of a trial recreating this discussion. Maybe you could follow it, if your were in at the start. Hard to say how anyone else would take it. Especially in the "one liners" world of NGs. Where might I post a summary that might not be blown, uncerimoniously, out of the water. In spite of opposing the formation of new groups (balkanization), I do think the general issues deserve more discussion.

David Janes

snip

Wow. You know a lot about gears too.