Cylinder keeps flipping on me

[Underspec]

Measuring off CAD model. When creating the cylinder it looks correct in the graphics window (along Z). But when it's measured, the **** thing flips vectors on me (along Y). Even switching to different workplanes it's doing this. What's going on?

[jmgreen]

There are 2 things I would do first. Verify the plane you are using when you measure the cylinder, do three rows instead of 2. You probably already have done this but it's all I got.

[William Johnson]

Are you using auto cylinder or manually measureing a cylinder?
If you are manually measuring a cylinder you will need to stager your points in the active workplane or measure 3 cross sections. PCDMIS will what I call invert the cylinder.

[Don Ruggieri]

Use Theo

Yes, I suspect you are using 4 hits at 2 levels. There are actually multiple solutions, and it is picking the wrong one.

As stated, changing to more levels, or even more hits per level should solve it. If you merely "cocked" one set of hits to skew it from the other it should also solve it.

There is a setting in the Auto Cylinder called (I think) 'use Theo' which compares the possible solutions to the theoretical vectors and eliminates the others.

[Wingman]

Cylinder is a 3d feature and the direction should be determined by the inspection order. Are you doing sets of circles? Hitting just anywhere? Is it a really short cylinder where the spacing between the hits of a particular band are as great or greater than the spacing between bands?
Sometimes I have had this happen with PC DMIS and it sux. Even once you get it right it will still capture it wrong on occasion. I have never figured out why so I am very interested in this.
Are you using points, circles, or a cylinder to measure it?
Regards,
Bill

[Underspec]

Manual cylinder. What do you mean 3 cross sections? For some reason it created a correct cylinder after the 6th time lol.

[Matthew D. Hoedeman]

Definately use an odd number of hits per level and an odd number of levels. This will give ONLY one solution.

[William Johnson]

Just measure it in 3 levels of points.

[Wingman]

Just measure it in 3 levels of points.

I have had three levels of points report incorrectly too on occasion. hmm

I have not tried the odd number voodoo though. Interesting.

[Underspec]

I'm just taking points off the CAD model due to the strange nature of the cylinder. For some reason my scans are coming up incorrect when I construct a cylinder so I am having to do this until I figure out what's going on.

[#2#]

As stated before - several times - use three levels of hits. Even as little as one hit in the middle level will set it straight. If you are doing 'learned' features measure three levels to get the cylinder. Then hit F9 and edit the number of hits you want, click the 'regenerate hits' box and click ok. It will ask if it is ok to equally space the hits. Click 'ok'. Now, if you do a cone like this it will ask how many levels you want. For some unknown reason it doesn't do this with a cylinder. It would make sense to ask this for a cylinder but it doesn't. So, it just assumes you want two rows. Ok, back to the task at hand. Lets say you specified 10 hits. That will give you 5 hits per level. If the cylinder's vector is 0,0,1 you will have 5 pairs of hits. Each pair will have the same x and y values but the z value will be different. Find the middle pair (where the z value changes). Highlight one of those hits, copy and paste it in between the two where the z value changes. Now, edit the z value to be something in between the other two z values. Move the cursor outside the feature. It will prompt you 'update feature nominals' or some such. Click yes. Now, you will have 5 hits on one level, one hit on another level and 5 hits on the last level.

If you want 5 hits each on 3 levels, copy the first five hits and paste them in between the original hits and edit the z values.

Another option is to measure two circles and construct the cylinder from them.

PC-DMIS is just guessing which way the cylinder is oriented. Sometimes it guesses wrong.

[Rob_K]

Increasing number of hits/rows solved my fliping cylinder

[cmmguy]

Someone tell me why, if you put the correct vectors in for the nominals, PCDMIS would not use that info to make sure that it calculates the feature correctly? The software should be using the nominals as part of the input for calculating the feature correctly and not just "guessing" what the correct vector is.

I can understand the problem happening with unknown features but not defined features.

[Wingman]

As stated before - several times - use three levels of hits. Even as little as one hit in the middle level will set it straight. If you are doing 'learned' features measure three levels to get the cylinder. Then hit F9 and edit the number of hits you want, click the 'regenerate hits' box and click ok. It will ask if it is ok to equally space the hits. Click 'ok'. Now, if you do a cone like this it will ask how many levels you want. For some unknown reason it doesn't do this with a cylinder. It would make sense to ask this for a cylinder but it doesn't. So, it just assumes you want two rows. Ok, back to the task at hand. Lets say you specified 10 hits. That will give you 5 hits per level. If the cylinder's vector is 0,0,1 you will have 5 pairs of hits. Each pair will have the same x and y values but the z value will be different. Find the middle pair (where the z value changes). Highlight one of those hits, copy and paste it in between the two where the z value changes. Now, edit the z value to be something in between the other two z values. Move the cursor outside the feature. It will prompt you 'update feature nominals' or some such. Click yes. Now, you will have 5 hits on one level, one hit on another level and 5 hits on the last level.

If you want 5 hits each on 3 levels, copy the first five hits and paste them in between the original hits and edit the z values.

Another option is to measure two circles and construct the cylinder from them.

PC-DMIS is just guessing which way the cylinder is oriented. Sometimes it guesses wrong.

But isn't this generic? I mean, we TELL PC DMIS which way the Cylinder is oriented. It should at least TRY to fit the points in that particular orientation first. I mean we ARE looking at the thing, you know.

There have been times that I have checked a roughly 5mm Diameter Cylinder over a length of 20mm and PC DMIS still got it wrong. More than math error going on in that case I would think. This even happens when the points are captured in DCC and the approach vectors have been corrected. It is mathmatically impossible (at least very highly improbable) for the orientation (and size) to be where PC DMIS gusses it to be given the particular approach taken.

BTW, I am pretty sure 3 levels is the first thing I tried when I encountered this situation and still had occasional problems with Short Cylinders. But maybe not. hmm I will keep an eye on that.

Regards,
Bill

[Underspec]

As stated before - several times - use three levels of hits. Even as little as one hit in the middle level will set it straight. If you are doing 'learned' features measure three levels to get the cylinder. Then hit F9 and edit the number of hits you want, click the 'regenerate hits' box and click ok. It will ask if it is ok to equally space the hits. Click 'ok'. Now, if you do a cone like this it will ask how many levels you want. For some unknown reason it doesn't do this with a cylinder. It would make sense to ask this for a cylinder but it doesn't. So, it just assumes you want two rows. Ok, back to the task at hand. Lets say you specified 10 hits. That will give you 5 hits per level. If the cylinder's vector is 0,0,1 you will have 5 pairs of hits. Each pair will have the same x and y values but the z value will be different. Find the middle pair (where the z value changes). Highlight one of those hits, copy and paste it in between the two where the z value changes. Now, edit the z value to be something in between the other two z values. Move the cursor outside the feature. It will prompt you 'update feature nominals' or some such. Click yes. Now, you will have 5 hits on one level, one hit on another level and 5 hits on the last level.

If you want 5 hits each on 3 levels, copy the first five hits and paste them in between the original hits and edit the z values.

Another option is to measure two circles and construct the cylinder from them.

PC-DMIS is just guessing which way the cylinder is oriented. Sometimes it guesses wrong.

I should of explained a little better. Imagine a potato chip sitting with it's flaps facing upwards. The points I have to take run along the outer edge of the top surface (Z). Now the cylinder I want is supposed to run along the X axis using these points. The potato chip would sit nicely on top of the cylinder if correctly constructed.