hexagon logo

Issues w/ Profile of a Surface

Hi,

I'm having an issue with profile of a surface using Xact Dimensioning on a Romer arm If there's not datum called out in FCF the part passes no problem, if there's one Datum called out the part passes no problem but multiple (two datum) and my part starts to fail by huge margins (+- 0.300"). I've tried to correct this issue on multiple, unique parts but have had the same issue with the dimension failing, considerably (the dimensions are very passable, FCF tolerance is .060 for a laser cut part).

I'm wondering if there's something (most likely) that I did in my programming, a rookie mistake, or something that I'm missing. I align, define datums and take points using a CAD model.

Is it possible I'm having issues with the vectors being controlled by the second Datum?

I will happily post code, pictures, etc. if that will help.

thank you for your support!
Parents
  • #1. You're building features with the minimum amount of hits, maximizing your repeatability due to variance. This is bad. Use more than the minimum amount of hits. I can't tell you how many to use, as I have no idea of how much form error is inherent in your manufacturing process, but i would only use the minimum amount of hits on a high tech 7-axis CNC-milled aluminum part. If this were a plasma cut part, i might use anywhere from 7-25 per circle, depending on the size of the circle.

    #2. in regards to #1: you shouldn't be using circles and lines, collected into a set, to approximate a scan. Here are the two options you should use, per our training:

    2A: Activate FindNominals with an appropriate tolerance, and use a Manual Scan or a Body Axis Scan to capture measured data (during an arm scan) and nominal data (due to FindNominals being active) simultaneously. Scan the entire part, or whatever portion you think is appropriate, and build your FCF accordingly.

    Pros: easy to run
    Cons: harder to replicate. It is literally impossible to take the same manual scan twice. It's much closer with a Body Axis Scan, though.

    2B: Create Auto Vector Points offline with the model (NOT MEASURED POINTS, THEY ARE DIFFERENT; AV points capture the IJK vector in a way that Measured Points do not.) Use Point AutoTrigger to accurately and repeatably shoot those points. Capture all points in a Constructed Feature Set. Profile that Set per the drawing.

    Pro: extremely repeatable, as you're taking nearly the same hits every time.
    Cons: more annoying to execute. more difficult to program. Less data taken than in a scan, unless you really go crazy with the hits.


    ----------------------------------------------------

    cliff's notes:

    - your program isn't working as expected because of your features, not because of the FCF.
    - your GD&T callout doesn't constrain all 6 degrees of freedom, so the scan still has room to slide in an attempt to fit. I don't know if that's what the engineer designing the drawing wants, but it's on the drawing, so that's what we're inspecting to.
Reply
  • #1. You're building features with the minimum amount of hits, maximizing your repeatability due to variance. This is bad. Use more than the minimum amount of hits. I can't tell you how many to use, as I have no idea of how much form error is inherent in your manufacturing process, but i would only use the minimum amount of hits on a high tech 7-axis CNC-milled aluminum part. If this were a plasma cut part, i might use anywhere from 7-25 per circle, depending on the size of the circle.

    #2. in regards to #1: you shouldn't be using circles and lines, collected into a set, to approximate a scan. Here are the two options you should use, per our training:

    2A: Activate FindNominals with an appropriate tolerance, and use a Manual Scan or a Body Axis Scan to capture measured data (during an arm scan) and nominal data (due to FindNominals being active) simultaneously. Scan the entire part, or whatever portion you think is appropriate, and build your FCF accordingly.

    Pros: easy to run
    Cons: harder to replicate. It is literally impossible to take the same manual scan twice. It's much closer with a Body Axis Scan, though.

    2B: Create Auto Vector Points offline with the model (NOT MEASURED POINTS, THEY ARE DIFFERENT; AV points capture the IJK vector in a way that Measured Points do not.) Use Point AutoTrigger to accurately and repeatably shoot those points. Capture all points in a Constructed Feature Set. Profile that Set per the drawing.

    Pro: extremely repeatable, as you're taking nearly the same hits every time.
    Cons: more annoying to execute. more difficult to program. Less data taken than in a scan, unless you really go crazy with the hits.


    ----------------------------------------------------

    cliff's notes:

    - your program isn't working as expected because of your features, not because of the FCF.
    - your GD&T callout doesn't constrain all 6 degrees of freedom, so the scan still has room to slide in an attempt to fit. I don't know if that's what the engineer designing the drawing wants, but it's on the drawing, so that's what we're inspecting to.
Children
No Data