Using Recall Points Boxes to analyze 2D Profiles

[[Th...]]

Dear all,

I have been struggling with the recall points box option for a while now, avoiding what seems to be an issue with a doubtful technique... I now have a slightly different application for which I cannot use the same way, so I will try to expose both my needs and problems.

A typical example of the usual need is to get angle and height/widht of a chamfer, which can be positioned at a varying position.
[img]https://i.ibb.co/5c6nmvc/Capture-d-cran ... 162842.png[/img]
We first scan a 2D curve using unkown contour, with margin on each side of the chamfer. From the measured profile, we get the two summits of the chamfer, using the systems shown on the right and the minimum coordinate feature (+22.5° and -22.5°).
The idea is to recall these two points in the base system, and use their coordinate to build adjusted boxes around the measured summits (their limits are fixed using formulas based on the previous two points, like [left summit.x+0.1;rightSummit.x-0.1] for the 45° line) (because these "measured summits" are not always precise, in case of small burrs, probing defect, ...), and get three lines, from which we recalculate two precise intersections to obtain the length of the chamfer. This construction is roughly the same than what we would do on a contourograph. But in this case, we need to "calculate" the boxes in a mathematical way which corresponds to what we do manually.
But it is not optimal to have boxes with fixed coordinate, especially with small chamfers (as you need in that case to have a small margin, implying a risk to recall the wrong points for the 45° line for ex). This explains the need to calculate the boundaries of the boxes from the measured summits.
[img]https://i.ibb.co/SQbTzk1/Capture-d-cran ... 162842.png[/img]
What goes wrong when using boxes with coordinates without formulas (horizontal line will not be calculated as horizontal as some chamfer points are within the box) :
[img]https://i.ibb.co/QX1vsRS/Capture-d-cran ... 162842.png[/img]

[[Th...]]

Now this way of doing things seems OK, but there is a glitch when put in application. From what I understand, the box is not applied to the evaluated point (which are compensated in my case using the measured points and not the nominal in the evaluation box).
As an example, i try to extract the radius part of this profile to evaluate it as a form defect :
[img]https://i.ibb.co/s5qW9v6/Capture-d-cran ... 162842.png[/img]
I come up with this :
[img]https://i.ibb.co/Xk6zvDy/prof.jpg[/img]
It seems that Calypso extends the “measured vectors” in green, get something like the intersection with the nominal profile (I do not know what are the cross symbols on the picture) and keeps the center ball points whose intersection with nominal profile is in the box. As the vectors are not perpendicular to the profile, the center ball points filtered are not the ones we want, and the extracted profile which is computed again from these center ball points is not located where I want.
What I would need is a simple filter on the evaluated points, and not a filter of the center ball points based on a vector which is not linked to the measured geometry…

I get to the result I want by saving the evaluated points into a file, recalling them in another feature from this file, and the box works correctly as the ball radius is already compensated.

However, I need to do this again, but on a profile which I do not want to evaluate (keep the profile of the center of the probe), and even with this trick, I face the same issue shown before, the box relies on measured vectors istead of simply filtering the evaluated points (in this case, ball center).

This is the way I usually use when I have to program on Mitutoyo, which has a different way to handle the curves… If any of you already faced this problem and has a solution, I would gladly try it ! Or another way to do this !

Thank you all in advance for your help 💡

Theo from Alpen’Tech, France

[[Ma...]]

First of all - "x" marks are points
Then - you have wrong vectors for curve

I use boxes sporadically - only when there are many paths but you need only portion of some of them.
For one 2D curve you can use selection from - to.

For boxes - measured points are shown always from base alignment and boxes don't move with them.

Best what you can do is make alignment from curve, this will give you alignment very close to nominals. Then copy this curve, recall points and select feature alignment from curve. Then again copy original curve and recall box from copied curve - this will make box move as alignment from curve changes.

Play with it.

[[Th...]]

Hello Martin,

And thank you for your reply.

What was disturbing me with the "X" marks is that they are greatly different from the shown deviations (small green dots), and the deviation profile makes much more sense than these X marks. In the screen shot of calypso in my first message, you can see that the "X" marks which are filtered corresponds to measured points of a part of the profile which is out of the box ! Meaning that when I evaluate this filtered section (when probe radius is compensated), I end up with deviation dots outside of the box (and by much).

You point out the fact that I have "wrong vectors for curve", and indeed they are not perpendicular to the measured profile. For my comprehension it is not a problem (the center ball points should be sufficient, as the direction chosen to compensate the probe radius comes from the points before and after). Maybe I am wrong from this point, but then the question is how do I get measured vectors perpendicular to the effective profile ? From what I know, I do not control the direction of the green arrows, and the evaluation of the profile (green dots when you check "show deviations") is not dependant on them.

Regarding boxes, I always work in base alignment with them, as I understood it is the way they work. I base their boundaries coordinates on the coordinates of other features in the base alignment.

Problem with the from-to filter is that in my sense it is neither precise nor adaptative to the profile. For example, the idea is to be able to give a consistent value for a chamfer of length 0.5±0.3, and in this type of tolerances, the form itself varies quite a lot. Even aligning directly on the curve (using best-fit I assume) will not be enough to get consistent values.

It is a bit unfortunate as from my perspective the box system is quite powerful, combined with formula based boundaries, and could allow us to perform consistent and advanced profile analysis...

Moreover, the thing is that with the trick of recalling points from a file, I get rid of the measured vectors and of the radius compensation itself, and then the box correctly applies to de deviations (evaluated profile directly) and I get my adaptative box package which works perfectly, and get me access to the measurement of nearly all characteristics on our drawings, including chamfers, radiuses, etc... And this is what I do not get : is this intended that the box should work like this, because it makes more sense to me to be able to filter the evaluated points (i.e. filter after radius compensation).

Or maybe the problem comes from another thing that I do not get... I learnt Calypso by myself which is not helping me, but I went through both general and curve manuals without finding the solution 😭

I hope that my explanations are clear enough 🙄

Thank you in advance for your answers 💡

Théo

[[Ma...]]

With this you can have precise boxes, which moves with part. Thats only solution i found to to make exact selection box.

Used:

• Resources -> Utilites -> Alignment from several curves
• Features -> Circle
• Features -> Curve

For better bestfit you can exclude some points from original curve ( or which is used in Alignment from curves )
Feature -> Evaluation -> Limit evaluation -> X Norm. Vector. Direction -> Parameters ( selected will be excluded )

[[Th...]]

Thank you again Martin for your explanations.

In fact I do not have any real problem positioning the box, as I make it move through formulas (base on coordinates of other measured features in the basis system). The issue is linked to the way boxes work, more precisely to the green measured vectors not being perpendicular to the profile, causing the fact that the evaluated points after box filtering are completely outside of the box (last picture of my two first posts).
I still do not get your remark saying that I have "wrong vectors for the curve" : from my understanding, the green vectors direction is not relevant to compensate the probe radius. In my opinion it should not be relevant for anything, however I assume from my tests that it plays a role in the box filtering action.

So either this is the intended way boxes should work, then I do not understand the advantages of programming things this way, in the other case this would be an issue to be corrected.

Am I completely missing something ? If there is no solution to my problem, I'll find a way around or will just abandon these kind of measurements.

Thank you again,

Théo

[[Ma...]]

Alright - what i saw is that your curve vectors are in material - invert them. It's like inside/outside - perhaps this will help you with something.

[[Ri...]]

Please sign in to view this quote.

Same with the Circle. It should be an Outside material Circle.

[[Ma...]]

Please sign in to view this quote.

I just picked some elements for demonstration purpose.

When he will try to bestfit with wrong vectors, then it will compensate whole touch diameter. Once vectors are corrected, then green arrows goes close to nominals, i think.