hexagon logo
Register Sign in
  • Replies 30 replies
  • Subscribers 222 subscribers
  • Views 876 views
  • Users 0 members are here
Options
Related

Question about the small arc method

constadin
Hello all!

Please let me know if I am doing something wrong following the small arc method.

Btw, I am using vector points and auto circles with NO CAD always in DCC mode. Let's also assume that I am measuring in Z plane.

1) I am measuring the small arc with a fixed radius auto circle.
2) I am setting my origin XY to this circle.
3) Just to be sure I am repeating with a new auto circle fixed radius and realign to it.
4) I am taking 3 vector points in polar coordinates to be precise.
5) I am reporting the PR of the points.

This morning I had a 80deg on an arc R2.8mm to measure. I used this method and I was getting repeatable results at 2.80. The issue is that the radius has been measured to be 2.75mm (verified with 2 other methods also).

I also tried to change the fixed rad value based on the R measurement I had performed before with least squares (around 2.75 non repeatable value) and I was getting results close to that value (the least square value).

With this method, does the fixed Radius value provided alter the centre that is being found? Doesn't this beat the purpose of using this method?

Thanks in advance!
  • in reply to constadin
    Why do 2 of your 3 points have a K vector value?
    Is your cylinder actually a cone?
    Is it non-perpendicular to your level plane?

    -.34 = ~40 degree of cosine error. This could absolutely produce half a mm of resultant error.
  • Coupe of observations/questions:

    You say the rad has been verified at 2.75 by two other methods - what methods are these? Any system that calculates the radius from points would be susceptible to the same issue (this is not a PC-Dmis issue, it's a math issue).

    The point of using the fixed rad method is that, by calculating a circle from points over an insufficient degree of arc, can give an inaccurate and non-repeatable centre point, giving bogus radius values.

  • in reply to constadin
    The circles are at the J plane that's why (J remains 0) Slight smile
    Before giving the code I wanted to simplify the request and just focus on the method and identify errors that I potentially did. This is why I said " Let's also assume that I am measuring in Z plane"

    Also the part is really small and does not really have a plane normal to the cylinder. It is a hollow tube of a few millimeters of diameter that has a offaxis semi cylindrical area cut off from it normal to its main axis. Imagine an extruded donut, grabing 4 points every 90 degs and pulling the north and south point while pushing the other 2.

    In any case, the method is what is I am after for, for future references also Slight smile
  • Coupe of observations/questions:

    You say the rad has been verified at 2.75 by two other methods - what methods are these? Any system that calculates the radius from points would be susceptible to the same issue (this is not a PC-Dmis issue, it's a math issue).

    The point of using the fixed rad method is that, by calculating a circle from points over an insufficient degree of arc, can give an inaccurate and non-repeatable centre point, giving bogus radius values.



    So 1rst method is giving it to an external metrology laboratory. It came up at 2.75 something. Second method is a mixture of results such as: Isomat projector (2d visual measurement), CMM simple measurement with least squares and finally testing fixed dia pins and seeing how they best fit on the rad cutoff.

    To tell you the truth, I am disappointed of this method because it is dependent on the value the fxd rad we initially give. As said, I tried 2.8 (nominal) and got repeatable close to 2.8 results and then I tried using...

    CYL_2_8.D (a least square circle that was taken before that gives 2.75..mm) as the nominal D and the results I was getting were repeatable 2.75 something.

    The repeatability is there but not the accuracy.
  • in reply to constadin
    are you sure these vector values accurately represent being tangent to the circle, when it takes the hit? Your vector says you are measuring those two hits at 40 degrees from the middle hit, however your PA's are only spaced 20 degrees from the middle hit at 180. Vector error is what this method should be mitigating.

    You can reproduce the first and third points by doing a copy/paste with pattern with an angle of +20 then an angle of -20, of the middle hit.
    or you can rotate your alignment 20 degrees each way and keep middle point's coords and normal (zero) vector.
  • in reply to constadin
    I am not sure what you mean by PA being different in pnt1 and pnt3 than my vectors. ACOS 0.9396926 is 20 degs. I am taking the points at -20 degs, 0 degs and 20 degs. The method of doing that (If I remember correctly) was with c/p pattern as you suggested. AFAIAC, there is no error here Slight smile
  • Do you experience using this method? If so... was is as accurate as it was repeatable? Was it not dependent of the value given for the fixed rad?


  • So 1rst method is giving it to an external metrology laboratory. It came up at 2.75 something. Second method is a mixture of results such as: Isomat projector (2d visual measurement), CMM simple measurement with least squares and finally testing fixed dia pins and seeing how they best fit on the rad cutoff.

    To tell you the truth, I am disappointed of this method because it is dependent on the value the fxd rad we initially give. As said, I tried 2.8 (nominal) and got repeatable close to 2.8 results and then I tried using...

    CYL_2_8.D (a least square circle that was taken before that gives 2.75..mm) as the nominal D and the results I was getting were repeatable 2.75 something.

    The repeatability is there but not the accuracy.

    Normally when i use this method i use the fixed radius and then change back to the original alignment and report out T values of each point. It doesnt give a radial value but it gives nominal surface deviation from the intended surface.


  • Normally when i use this method i use the fixed radius and then change back to the original alignment and report out T values of each point. It doesnt give a radial value but it gives nominal surface deviation from the intended surface.

    I am not sure I follow. So you use the fix rad method to find a center or not? Not really sure how this can help in the issue I have with the accuracy of this method Disappointed
    If there is a "repeatable and accurate" way to measure small arcs other than going the least square methos, please someone list the steps because I am clearly missing something with the fix rad method.
© 2024 Hexagon AB and/or its subsidiaries. Privacy Policy | Cloud Services Agreement