Cylinder measuring trouble

[johnnyaxle]

I'm having trouble with measuring a cylinder cut by 2 non-perpendicular planes. Can anyone tell me the mathematical fundamentals of the measurement? How can they determine thw radius with only 6 points? Thank you!

[rerun_lives]

Are just trying to get a diameter? I seem to have problems until I use three levels or rows of hits. What exactly do you mean by "cut by two non-perpendicular planes? I've not had much success using only six hits - nine minimum.

[Matthew D. Hoedeman]

I usually use 7 or 9 hits, on 2 levels, with (obviously) a differnet number of hits on each level. Works pretty good. Always better to use MORE than the minimum number of hits. The minimum number of hits will generate a 'perfect' feature and no feature is perfect. Example, 3 hits minimum to make a circle. If that is what you use, then you will always see a prefectly round circle and that just does not happen. Also, by using MORE than the minimum, if YOU take a bad hit, it will show up a lot easier than if you only did the minimum number of hits (as in roundness)

[John Kingston]

I use 3 levels, 7 hits on the upper and lower, 1 hit in between those levels.
Similar to what Matt does, but different........

[craiger_ny]

What's the scoop on using different number of hits on levels? Same with reversing direction on levels? Why is this mathematically adventagious?

Craig

[Matthew D. Hoedeman]

What's the scoop on using different number of hits on levels? Same with reversing direction on levels? Why is this mathematically adventagious?

Craig

Well, let's see (and this is just MY opinion)

If you measure 4 hits on 2 different levels, then there are 3 possible cylinder solutions based upon the possible ways to construct circles to get to the cylinder, if all the hits are evenly spaced (or even close to it). However, if you use 3 on a leve and 4 on another, the circles will not construct out the same and it pretty much forces it to a single cylinder.

[craiger_ny]

I hear you there, that one makes sense but even with say 8 hits on one level I have heard recomendations that the next level be either 7 or 9 or pretty much anything as long as it is not 8. I have also been told that the direction of measuring should be reversed on levels (PCDMIS does it that way). But I have never been given an understandable explination as to why the math comes out better that way. I'm thinking it should do the math no matter what.

Craig

[Matthew D. Hoedeman]

Now, the interesting thing, I don't do a lot of cylinders, but I do have them from time to time. I have NEVER had one 'lay-down' or flip to some whothehellknows direction. They have always stayed points in the right direction. I don't always do LONG cylinders, some are quite short and most of them I do with only 2 levels. I have to think that the different number of hits on each level MUST have something to do with that. I have seen lots of people have problems with cylinders changin direction (laying down) but I have never had the problem, in any version of Pcdmis. I also have never used the auto-feature cylinder, I have always 'learned' them.

[Jan d.]

But I have never been given an understandable explination as to why the math comes out better that way. I'm thinking it should do the math no matter what.

Craig

On the number of points: I believe that the only reason to measure several circles with different numbers of points, is so that the least amount of form error escapes. I do not think that the "math works better" that way. The math works just as good, only you will find more form issues if you measure circles with different point counts.

As for measuring in different directions, I have no idea where that comes from. Maybe some of the old machine (with lots of backlash and only rotary encoders on the motors) running with bad pieces of software (ones that do not take center of ball measurements) would be helped with this strategy. I do not think that today's machines would suffer from the same ills.


Jan.

[Matthew D. Hoedeman]

Well, Pcdmis measuring each level of points in a different direction is actually a good idea that they put into the software. What if you are only measuring 1 side of the cylinder, for whatever reason. Do you want it going all the way back around to the other side before starting again or do you want it to save time by going to the next closest point? In a way, it makes sense, what if you can't go all the way arounf the cylinder? Your start/stop angles will be in the 'clear' area, where you can move, do you want Pcdmis to try to move through the area that might be blocked? CRASH!

[craiger_ny]

I always edit my hits on cylinders (every feature really). If anything it is to put my start and end points closest to my last and next move. Execution speed is all I'm after by doing that.

What you say on form evaluaton makes sense especially in light of most form issues coming from workholding and following a pattern throughout the feature. I'm just putting the challenge out there to convince me the math is better by practicing that since I have been told that. I think your guys' explinations make a lot of sense but one would have a hard time convincing me the reason for doing it is because math is better, but you never know.

Craig

[cmmguy]

...I have also been told that the direction of measuring should be reversed on levels (PCDMIS does it that way). ...

Craig

This is done for safety reasons(anti collision) Think in terms of partial cylinder and the reversing direction will make more sense. edit: (dang it matt, go find something to do while I finish typing....)


I just had a problem with constructing a cylinder from two previously measured circles. It kept flipping the cylinder 90º to the circles axes. I had to manually go into the construction code and enter the IJKs for the cylinder(It seems that we shouldnt have access to this in the construction window ), it still kept overriding my IJKs, I had to finally create assignments for a one and zero value and put those into the IJKs for the nominals before the stupid thing would calculate this correctly.

Why does PCDMIS override my nominals? What possible setting could there be that says to change what the user enters for the nominal??? I am not talking about a datum change above and it trickling down, I am talking on the immediate features.

[John Kingston]

I was told to use 3 levels, it makes the "recognition" of the type of feature work better (as Matt pointed out). I was also told different directions. That was so long ago I don't recall why. At the time I needed to get parts checked so that wasn't too important. I don't ever recall hearing that it was due to the the math "working better".

[John Kingston]

I had to finally create assignments for a one and zero value and put those into the IJKs for the nominals before the stupid thing would calculate this correctly.

Why does PCDMIS override my nominals? What possible setting could there be that says to change what the user enters for the nominal??? I am not talking about a datum change above and it trickling down, I am talking on the immediate features.

That is an interesting solution and then a good question. Could've used it yesterday, but since mine was a short learned cylinder, I just 2 extra unevenly space hits on a third level.
The funny thing is that program worked fine last week, but didn't Monday.

[cmmguy]

I can understand three levels on a auto-feature recognition, but when you choose two circles it should be smart enough to look at the axis of the line between the circles or let me put in the IJKs.


I use quick assignments all of the time to keep PCDMIS from screwing up my programs. It is a learned response to a negative input.
At the beginning of each program, I assign a V1 for 1 a V0 for 0 and then insert that into the nominals of features that wont "stay still"