Same Profile Form & Location & Form Only

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Same Profile Form & Location & Form Only

    I have a complex surface with a call out for Profile of a surface Form and location A/B/C/D. There is also a call out for 'Form Only' on the same surface, both dims use the exact same feature set to dimension to. My question is sometimes I will get higher 'Form Only' results and better 'Form & Location' results for the same surface. Does it make sense that the surface should be better to datums and worse to best fit form only? TIA!
    3.7mr3CAD++ / 2011mr1CAD++/2012mr1CAD++/QUINDOS7

  • #2
    Originally posted by Liambo734 View Post
    I have a complex surface with a call out for Profile of a surface Form and location A/B/C/D. There is also a call out for 'Form Only' on the same surface, both dims use the exact same feature set to dimension to. My question is sometimes I will get higher 'Form Only' results and better 'Form & Location' results for the same surface. Does it make sense that the surface should be better to datums and worse to best fit form only? TIA!
    No, it could be either or. Depends if the surface is big, but shifted in one direction, and if the surface is small and shifted the opposite direction.
    B&S One
    PC-DMIS CAD v2014

    Romer Infinity

    Comment


    • #3
      Originally posted by WolfMan View Post
      No, it could be either or. Depends if the surface is big, but shifted in one direction, and if the surface is small and shifted the opposite direction.
      Can you expand? I think I sort of get what you mean, sort of Just trying to get my head around this, thanks!
      3.7mr3CAD++ / 2011mr1CAD++/2012mr1CAD++/QUINDOS7

      Comment


      • #4
        I am not very good explaining things, I can try using as example a part I am currently working on. Imaging a part that look like a big marshroom. The marshroom hat is a Spherical Radius 3" and the location to the center of the sphere comes from the bottom of the hat, 2.000. SR has a profile .002 to AB. A is the leg of the shroom, and B is the bottom flat of the hat.

        Say the SR measures 2.950, so if you just measure a profile form Only, not to anu daums, it will be .100. .050x2. Good so far?

        Now, say the location of the sphere is 1.950 If you would
        Measure form + loc, the profile to AB form and loc would be close to zero, maybe .005,.008, not .050.

        If the location would be 2.050, the profile form and loc would be about .200. .050 radius dev+.050 loc dev x2 for the profile.

        Make sense?

        Here is a pic, dont mind the values, it was just an example
        Last edited by WolfMan; 05-12-2015, 12:54 PM.
        B&S One
        PC-DMIS CAD v2014

        Romer Infinity

        Comment


        • #5
          By worse do you mean like the number I have circled here?
          I would expect this.
          formonly.PNG

          Comment


          • #6
            Originally posted by Eric_ View Post
            By worse do you mean like the number I have circled here?
            I would expect this.
            [ATTACH=CONFIG]10426[/ATTACH]
            Like this:

            Dims_Profile.jpg

            In metric btw.
            3.7mr3CAD++ / 2011mr1CAD++/2012mr1CAD++/QUINDOS7

            Comment


            • #7
              WolfMan, I have to disagree with you. If it truly tries to "best fit" the 2.95 value to the model and no datums, shouldn't this allow it to move the point set with all 6 degrees of freedom to get the best fit? Shouldn't it try to make the deviation values from the model as small as possible? I think it would overlap the theo and actual and look something like this...

              profile example.jpg

              This shows the scenario described in your example, and for a 120° arc, the form only profile would be ~.0341... This is the way I feel like it should work, am I wrong? I feel like form only should always be better since it is not constrained by anything. Is this incorrect?

              Comment


              • #8
                No. Well, i am not talking about BF. A profile is not BF., unless you know something I dont. If its a profile without datum, the Center of the Rad will stay where it is, so if the Rad is .050 small, the profile dev is .050
                Originally posted by Mike Ruff View Post
                WolfMan, I have to disagree with you. If it truly tries to "best fit" the 2.95 value to the model and no datums, shouldn't this allow it to move the point set with all 6 degrees of freedom to get the best fit? Shouldn't it try to make the deviation values from the model as small as possible? I think it would overlap the theo and actual and look something like this...

                [ATTACH=CONFIG]10428[/ATTACH]

                This shows the scenario described in your example, and for a 120° arc, the form only profile would be ~.0341... This is the way I feel like it should work, am I wrong? I feel like form only should always be better since it is not constrained by anything. Is this incorrect?
                B&S One
                PC-DMIS CAD v2014

                Romer Infinity

                Comment


                • #9
                  Form only can never be worse than form and location if an appropriate fitting method is used. There could be situations where least squares fitting will kick a small part of a profile out in favor of a better fit on the reminder of the surface. So it could be possible that the best fitting could actually create an area with greater deviation than you would have relative to the datums. Profile is evaluated based on the worst points so this could result in a higher profile value for the form only dimension.

                  If you see something like this it is an immediate red flag that your best fitting is not working correctly. ISO requires that you use min/max fitting so you'll never have the issue described above if strictly following ISO. That said, min/max fitting is very sensitive to outliers so you need to do some filtering to make it robust for production. Pcdmis doesnt have the best selection of filtering methods and also doesn't support 3D min/max fitting so your options are pretty limited. In many situations vector lest squares or least squares are the most practical options but you need to be aware of the potential pitfalls and closely monitor the behavior.

                  The first thing to do when you see something like you are seeing is to plot the profiles in an analysis view so you can see what it is doing. I'll bet that you'll see that the fit is doing something really funky with the profile.
                  2013MR1 SP6
                  Global Frames, Tesastar-M Heads, LSP-X1M/H Probes

                  Comment


                  • #10
                    Originally posted by DaSalo View Post
                    Form only can never be worse than form and location if an appropriate fitting method is used. There could be situations where least squares fitting will kick a small part of a profile out in favor of a better fit on the reminder of the surface. So it could be possible that the best fitting could actually create an area with greater deviation than you would have relative to the datums. Profile is evaluated based on the worst points so this could result in a higher profile value for the form only dimension.

                    If you see something like this it is an immediate red flag that your best fitting is not working correctly. ISO requires that you use min/max fitting so you'll never have the issue described above if strictly following ISO. That said, min/max fitting is very sensitive to outliers so you need to do some filtering to make it robust for production. Pcdmis doesnt have the best selection of filtering methods and also doesn't support 3D min/max fitting so your options are pretty limited. In many situations vector lest squares or least squares are the most practical options but you need to be aware of the potential pitfalls and closely monitor the behavior.

                    The first thing to do when you see something like you are seeing is to plot the profiles in an analysis view so you can see what it is doing. I'll bet that you'll see that the fit is doing something really funky with the profile.
                    Hi DaSalo, thank you - What you have described is basically what is happening with this surface. To simplify, Imagine looking down on a semi circle shape. After I dimension profile of a surface to a fully constrained FCF, I have larger deviations on the left of the circle say +0.600mm material on, but only in a small area on the left side, near the edge of the circle if you like (Center), now on the right of the circle I have plus material on as well say 0.300mm, but in a more controlled and larger area. Now that all looks OK with form and location, and I believe the part actually is like this, with material on both sides.

                    Now when I dimension 'Form Only' and apply a Vector or least squares best fit algorithm, it seems to favor and best fit the rest of the surface on the right and push out the small area with the larger deviations and thus pushes the 0.600mm plus material to say 0.800mm and then report the profile.

                    If I apply 'No Fit'. Form and Form & Location are the same.
                    3.7mr3CAD++ / 2011mr1CAD++/2012mr1CAD++/QUINDOS7

                    Comment


                    • #11
                      Originally posted by WolfMan View Post
                      No. Well, i am not talking about BF. A profile is not BF., unless you know something I dont. If its a profile without datum, the Center of the Rad will stay where it is, so if the Rad is .050 small, the profile dev is .050
                      A profile with no datums is BF. The center of the radius is meaningless for a profile without a DRF. Mike Ruff is correct.
                      "This is my word... and as such is beyond contestation."

                      Comment


                      • #12
                        It's a tricky problem. There are software packages on the market, like "SmartProfile" from Kotem, that will handle 3D min/max fitting and would likely give you a better result.

                        If you can't justify that you could consider bringing the point cloud into a CAD software and manually adjusting the fit to achieve best min/max condition and then documenting your results in a supplemental report.

                        Finally, you could create a series of 2D min/max fits covering the surface in sections and then show the results and differences in transformation between these fits. This is typically how complex shapes are handled: a series of 2D sections at defined locations and then some additional requirements that prevent the sections from being too far shifted or rotated from each other.

                        Least squares fitting is what it is and if it isn't giving a good representation of true surface condition you need to work out another method using a different type of fitting.
                        2013MR1 SP6
                        Global Frames, Tesastar-M Heads, LSP-X1M/H Probes

                        Comment


                        • #13
                          If you are using "feature set" for this profile, you are measuring "form and size" and not just "form"
                          Form only option in PcDmis would evaluate form or form and size depending if you are using plane (form only) or feature set (form and size)
                          sigpic...engineering

                          Comment


                          • #14
                            Originally posted by Obelix View Post
                            If you are using "feature set" for this profile, you are measuring "form and size" and not just "form"
                            Form only option in PcDmis would evaluate form or form and size depending if you are using plane (form only) or feature set (form and size)
                            I'm using a feature set made up of defined scans on the surface in question. I can see that it is using 'Size' even in the calculation of 'Form Only' as it splits the tolerance into two sides, even though it shows just the Plus tolerance in the dimension. I'm using Xactmeasure in 2012mr1.

                            My understanding was to turn of the use of Size in profile form only, I would have to turn off 'Use Size For Profile Dimensions' in the settings editor and use a legacy 'Form Only' Dimension instead of Xact.

                            Cheers!
                            3.7mr3CAD++ / 2011mr1CAD++/2012mr1CAD++/QUINDOS7

                            Comment


                            • #15
                              Originally posted by VinniUSMC View Post
                              A profile with no datums is BF. The center of the radius is meaningless for a profile without a DRF. Mike Ruff is correct.
                              I guess I was wrong! I was not aware that a Profile works like BF profile. If thats the case, then yeah profile form only can never ne larger than the form and loc. I guess it can be in some cases in the corneres, as stated. I am not sure calypos evaluation is the same way. I had a part where I had location with in .001, but size was .025 small, the profile form only shows 0max, .025 min, which is .050 profile. If it was doing BF it would be half of that.
                              Last edited by WolfMan; 05-13-2015, 07:54 AM.
                              B&S One
                              PC-DMIS CAD v2014

                              Romer Infinity

                              Comment

                              Related Topics

                              Collapse

                              Working...
                              X