Needed Devlopments in Constraints

[oldfox]

using variables in calculations to determine the lengths. This cuts out any and all rounding errors.

Yes, of course. That is the other alternative.

One measures a dimension, and puts the resulting distance in as the distance between holes. That should work since Alibre says that is the distance, and you put that same number in AS the distance in a mating part.

Then, one finds that the measured distance, when put in as the distance between holes in the mating part, DOES NOT ALLOW THAT PART TO ALIGN, according to Alibre.

That way will work ONLY if Alibre uses it's 19 places to calculate the positions and then rounds to no less than 6 places and uses THOSE numbers for the constraints. Then the standard constraints should work.

 

[oldfox]

I'll hang on to my kiss for now, thanks. You have a square root there; of course there's rounding. Also, don't forget; the length of brace 3 has already been "fixed" and its drawing published, as it is used in other structures; so using an equation to change it a little is not really kosher.

I agree that using the math *should* give you locked parts. (purposely not calling "locked", "constrained"

Its the method that is important! You can apply formulae using variables to all your problems and the rounding problem you have is a thing of the past.

See other quote above

You need to learn how to use Constraints!

Lew, teach me. I have two holes. One slightly larger than the other one. (threaded hole and clearance hole) I'me working with
chisled-in-stone tolerances and the clearance hole is within that tolerance in diameter. It is however not concentric with the
threaded hole. It is also within tolerance in position. Which constraint in the Alibre toolchest will give me a true, locked "align"?

The two parts *must* maintain angular relationship.

This is really turning out to be a fun Sunday.

(but don't forget bit wobble. No strings of 9's nor 0's)

 

[JST]

...

That way will work ONLY if Alibre uses it's 19 places to calculate the positions and then rounds to no less than 6 places and uses THOSE numbers for the constraints. Then the standard constraints should work.

So, if I understand you correctly........

YOU THINK that Alibre should NOT be able to measure a distance, and present a displayed distance that actually equals the distance that exists (as far as the Alibre internal criteria for equality are concerned)?

In other words:

YOU THINK that Alibre should use, as its definition of equality, a number of decimal places PAST what it actually displays?

YOU THINK that Alibre should be able to take two numbers which are equal to the number of decimal places allowed for numeric entry and numeric display, and DECLARE THEM TO BE UNEQUAL?

Given what you seem to believe as per the points above, HOW do you suppose the program can even BEGIN to operate coherently and consistently from the viewpoint of the operator?

HOW can a program which HIDES the true number it is using from the operator, actually be usable?

Can you EXPLAIN how it is GOOD if a program uses, for constraints, numbers which the operator cannot see, change, or input?

I CLAIM that Alibre should use 6 decimal places to determine equality, constraints, etc. AND NO MORE THAN THAT.

The remaining decimal places should be a mathematical "padding" to prevent, as much as possible, rounding etc from encroaching on the 6 visible places. They should be maintained, but not used for equality, etc.

At present it appears that this is not the case.

 

[oldfox]

The remaining decimal places should be a mathematical "padding" to prevent, as much as possible, rounding etc from encroaching on the 6 visible places. They should be maintained, but not used for equality, etc.

At present it appears that this is not the case.

Agreed.

and then rounds to no less than 6 places and uses THOSE numbers for the constraints

Agreed. And I apologize for using the word "rounds". The intended method would be "truncation".

Equality doesn't work here. How many places have the supercomputers taken PI out to so far and still haven't reached the end?

So anytime you use math within Alibre, you will be dealing with 19 places. What should happen with all of those places after 6?
I sure don't want to have to enter 19 numbers just to get two holes to "constrain".
And this still won't account for bit wobble. 4.999999 + 5.000001 = 10.000000 ******** addends are UNEQUAL.

btw. Yes , you did misunderstand me.

 

[bigseb]

Also, don't forget; the length of brace 3 has already been "fixed" and its drawing published, as it is used in other structures; so using an equation to change it a little is not really kosher.

Ok, so by manually entering a value for a dimension you get a rounding error. But using variables and calculations within Alibre you get the same value but without the rounding error. Yet instead of changing how you get your value you would rather Alibre rewrite their code.

Ok.

Not much more I can add to this thread then.

 

[oldfox]

Ok, so by manually entering a value for a dimension you get a rounding error. But using variables and calculations within Alibre you get the same value but without the rounding error. Yet instead of changing how you get your value you would rather Alibre rewrite their code.

The world is a world of pros and cons.

Your point of view:
Pro:
The coders don't have to do anything and the users don't have to wait for them. Math will get the desired effects, usually.
Con:
*Every* user would have to know what math to use and how to use it. Good luck with that one.

or

The other side of the coin:
Pro:
The coders can usually make changes that will work. It's a business efficiency/financial decision.
Con:
Yet another burden on a recuperating system. Will probably come to pass, *someday*.

So in the meantime...

 

[Lew_Merrick]

Lew, teach me. I have two holes. One slightly larger than the other one. (threaded hole and clearance hole) I'me working with chisled-in-stone tolerances and the clearance hole is within that tolerance in diameter. It is however not concentric with the threaded hole. It is also within tolerance in position. Which constraint in the Alibre toolchest will give me a true, locked "align"? The two parts *must* maintain angular relationship.

e-mail me at tangent@olympus.net. I am somewhat busy as I have equipment for a satellite that I have to have released to the "shop" by Thursday to be completed. I will, when time is available, happily have a GoToMeeting with you to clarify things.

 

[bigseb]

The world is a world of pros and cons.

Your point of view:
Pro:
The coders don't have to do anything and the users don't have to wait for them. Math will get the desired effects, usually.
Con:
*Every* user would have to know what math to use and how to use it. Good luck with that one.

or

The other side of the coin:
Pro:
The coders can usually make changes that will work. It's a business efficiency/financial decision.
Con:
Yet another burden on a recuperating system. Will probably come to pass, *someday*.

So in the meantime...

Come on, Chris. You're really gonna tell that instead of engineers getting to grips with basic mathematical principles we're gonna dumb down the software? Sorry, I'm with Lew on this one.

 

[JST]

There should BE no "bit wobble". That is for measuring systems.....as with "accurate to 0.0001 +- 1 count". But numbers are truncated, or rounded, so there can be something that is similar.

The whole point of having 19 decimal places is to shove the errors down to where that is not an issue if it exists.

Perhaps every so often, a number is at, for instance, 5.999999...........99, and it may, by a combination of rounding, etc be displayed as 6.000000............00. I would expect that to be relatively rarely seen, as there are not that many discrete numbers that will flip the 6th place with one count difference.

 

[JST]

Come on, Chris. You're really gonna tell that instead of engineers getting to grips with basic mathematical principles we're gonna dumb down the software? Sorry, I'm with Lew on this one.

SIGH..................................................................................................................

Nobody wants to dumb down the software.

There is, however, a practical consideration..... You and Lew are basically demanding that everyone should do a complete analysis of every dimension prior to starting up the software. That every part must be fully defined, before one has any right to begin with CAD.

That CAD is basically a DOCUMENTATION TOOL, used only to document in a model what has been fully designed and tolerance checked on paper, and that such is the only "PROPER" use of CAD.

Sorry, boys, but while SOME users need to do that, CAD is a very useful design tool for basic "what if" type "clean sheet" designs. I use Alibre for that.

I also have some clients who give me "engineer's drawings" that simply give some overall dimensions and structural member sizes. What I do is to "detail" those, working out the dimensions required to show a set of parts which will assemble to form the structure wanted. Drawings telling the shop where to put holes, what length to cut, etc, to make the structure as the engineer's drawing shows.

Particularly with use as a design tool, but also with detailing, some dimensions are derived, measured off the model. Yes, in many cases they could be calculated to some number of places accuracy, but that seems remarkably silly when you have the "built" structure right there, waiting to be measured.

Doing an exhaustive evaluation of the tolerances is just never done for these structures. It is not necessary, the basic allowances are KNOWN to exist in the hole oversize standards, and the fact that parts can be "drawn-into" alignment if the shop is off a bit.

What is needed to make this more do-able is the "tolerant fit" constraint that has been described.

That takes care of the dimensions that will NOT be followed to better than 1/32" (steelwork) AND the various oddities that Alibre comes up with from time to time, when it says that 4.000000 does not equal 4.000000.

I will point out that this is not "dumbing down", it actually is allowing actual effective use of the tolerances which the design allows. Enter the tolerance in the "tolerant fit", and you get a check on whether you are in fact hitting the tolerances with the parts you have defined in CAD.

 

[bigseb]

You and Lew are basically demanding that everyone should do a complete analysis of every dimension prior to starting up the software. That every part must be fully defined, before one has any right to begin with CAD.

Its good practice. Do it or don't, up to you. If you don't then consider there are other constraints that can be used beyond hole align.

 

[JST]

Its good practice. Do it or don't, up to you. If you don't then consider there are other constraints that can be used beyond hole align.

Of course, hole align is what is WANTED.

And the "tolerant constraint" would work perfectly for this, as well as NOT being a "cheat", but a useful tool. Using other constraints is a "cheat", another WORKAROUND (I know you do not accept that term).

Not to mention the cases where Alibre gets into a snit and insists that there is a difference in the 14th decimal, (which you have no control over, whether you enter a calculated number or not) so that the constraint fails.

And the case where you are NOT GOING TO USE the calculated number, because the part will be made on a 1/32" grid, and so you MODEL it done to a 1/32 inch grid.

Academic purity is fine, but not when it means we all have to go use a different software because some refuse to accept anything other than perfectly proper academic design (according to their personal definition of that).

A complete and exhaustive tolerance analysis of every possible combination of limit conditions, over a structure of a hundred or more parts, is a large task. It is simply not done in the usual case, no matter how one insists that it is "good practice". I am by no means the only one who does not do it, but I may be one of a few who ADMIT it.

Accepting that many designs do not need to be calculated to 6 decimal places is really a necessary condition. Admitting that there is a tolerance is pretty useful for the real world. And the "tolerant constraint" is actually USING the tolerances in a positive way.

Plus, not every model is intended for creating drawings. It may be used as input to a rendering program, but that still does not mean that one can necessarily cheat the constraints, or ignore them.

Remember, CAD takes NO NOTICE of any tolerances. Tolerances DO NOT EXIST in the world of CAD. Every part is shown as one set of dimensions, presumably "down the middle of the tolerances". That set of dimensions either works or it does not. If not, CAD insists that they are wrong, it has "zero tolerance".

NOBODY is trying to force YOU to USE any of the tolerant constraints if that offends your sense of how CAD "should be used". "Alles muss in Ordnung sein", and all that is fine if you want to insist on it for yourself. I am not quite sure of the motivation for refusing to consider them as part of the "toolbox" of things within the program.

 

[HaroldL]

This is really an interesting thread.

JST, if I understand your situation, you have a beam that needs to be constrained diagonally from one vertical column to another at the opposite corner. You are attempting to place concentric constraints on the holes at both ends and Alibre won't allow this because of its rounding error. Why not use Tangent Inside constraints at each end between the holes and the bolts and a Parallel on the beam to hold it level in the assembly, if in fact it is and Alibre would allow it. Wouldn't that reflect how the actual beam is place in the construction? Maybe the Parallel isn't needed as long as the holes align with the bolts installed.

I don't think that the holes are going to remain concentric in the assembly. Even though that may be how they are initially assembled by the time all the bolts and nuts get tightened to the proper torque the holes will likely end up being tangent to the bolt at some point.

I could be all wrong on this but thought I'd put it out there as solution to try.

 

[JST]

It may NOT be a result of rounding. It may be inherent due to the shop working only to the nearest 1/32", so all locations and lengths etc must be on a 1/32" (or other dimension) "grid". That means that any fancy calculations are of no help, since the nearest grid location will be used by the shop, and so should be used in the model.

Tangent has a locus of positions that is circular. All length mis-positions are accompanied by a transverse mis-position, inherently. It follows the circle. To the original mis-position, is added a second one imposed by the "fix". I believe "the cure is as bad as the disease".

So, a floor beam would be moved out of parallel with other floor beams, potentially causing a cascade of other problems.

A standard align, since we are discussing that at the moment, requires that the centers of the holes shall be perfectly coincident, to some number of decimal places (which appears to be greater than 6).

The proposed "tolerant" constraint in this case would provide a selectable "area" , around the center of the other hole, in which the center of the first hole could be located and still be accepted as aligned. ANY location within that area would be OK.

In that way, our floor beam can stay parallel, while having a lengthwise position error off of exact co-incidence. If that error does not move it out of the "tolerance area", it is not regarded as a failed constraint.

You are correct, the holes may be concentric while the spud wrench is in, but over time the structure will settle into a different set of positions. However, by then, the thing will be assembled, which is the immediate goal... we want to assure it can be assembled without springing structural members.

 

[VoltsAndBolts]

In most cases when I am dimensioning a part the tolerances are

2 places (+/- .01)

3 places (+/- .005)

Shouldn’t you be able to set the constraints to follow that as well?

 

[JST]

In most cases when I am dimensioning a part the tolerances are

2 places (+/- .01)

3 places (+/- .005)

Shouldn’t you be able to set the constraints to follow that as well?

Yes you should.

Making what is essentially a prissy-fussy argument that all dimensions MUST be calculated before starting is not helpful whan you KNOW that the resulting hole spacings will NOT match every location on the structure, due to the shop measuring increments.

And Alibre is totally inconsistent in its code ANYWAY, and it NEEDs to be re-written.

Just had another one today.... distance is 47.62500 inches on the calculator (which is what was expected). AND it measures as 47.625000 inches. So I set my spacing to 47.625000 inches.

HAH....! Alibre will NOT allow the alignment. There is some tiny error that I have no means of seeing.

Even if you calculated the distance, and it came out to 47.625000073", what would you DO about that? You have NO MEANS to enter that extra 0.000000073" that is required.

So, you say, easy-peasy, just CHANGE that distance to be SET TO 47.625000 inches. Seems reasonable, but maybe not......

What you have done is to push the error elsewhere, with unknown but probably bad effects on the rest of the structure. Geometry will not be got round, the distance is the distance.

Alibre gives you no tools to fix this. In fact the only way it is even discoverable is that the constraint has an error.

THIS. IS. COMPLETELY. WRONG.

No matter WHAT your argument about global variables is, IT IS WRONG. Alibre is saying , in essence:

"There is a tiny error in your distances which we WILL NOT ALLOW you to see, and WILL NOT ALLOW you to fix, but we also WILL NOT ALLOW you to make the alignment, basically you are screwed and we are laughing at the jam we have gotten you into....have a nice day.....".

The software should accept agreement to the 6 displayable decimal places as equality sufficient to allow the alignment.

Agreement to the full displayable number of decimal places is obviously within any tolerance that is meaningful in Alibre. Or anywhere else, since the error is detectable only in a very advanced metrology lab.

The tolerant alignment would permit this alignment, AND it would alert you (with an error just as happens now) if the misalignment exceeds the tolerance you set. Therefore it would be a tool, and not a dumbing down of the software. There is no obligation to use it, so I have no clue why it is being attacked as a "dumbing down".

 

[oldfox]

This reply is directed to no one in particular and everyone in general. I know I have stumbled on real simple hole aligns that no matter
what constrain I tried, it wouldn't work. I have since learned to forget the concentric alignment and *first* do an axial constraint with what ever offset exists. Then I can do the concentric align constraint and it works.
Problem? One of the pieces is rotated along it's perpendicular axis to the hole. Just thought I'd throw that out there for those of you who
may be having difficulties.

JST, I realize that is not your issue. Don't take it personal.

 

[JST]

Not taking it personally...... (Didn't the other time either it was a rather heavy handed reductio ad absurdum on my part, sorry)

But, in my experience, the hole align should do everything, it the part is not currently set with the holes parallel, the constraint will rotate them to make it so (unless that conflicts with another constraint).

There is also the "not consistent" error, and that is often cured by moving parts to a location from which there is exactly one shortest movement to get to the intended position.

 

[oldfox]

No problem. And I agree with you that the hole constraints really need some help.

the constraint will rotate them to make it so (unless that conflicts with another constraint).

Reminds me of the times when I have tried to line up something by using the "orient" constraint. OOPS, I picked the wrong face. It just
flopped 180+- degrees! Nothing like having to do something twice.

 

[Lew_Merrick]

Yes you should.

[Text removed by Moderator] You demonstrate no conept whatsoever as to how Constrints function!

What ypu are talking about is a "Tolerance Variation Analysis" -- sort of like the "Interference Analysis" tool.