Profile verification

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

  • Profile verification

    Hi. I am at a new company and I have been tasked with confirming a customers complaint of profile exceeding the ".010 to ABC all over ". this was an existing complaint, and it turns out my predecessor neglected to measure it. I have now measured one part on each surface, using auto features and scanning. my results are way better than the results the customer provided, however there are some areas that while they do show out above .010 for form and location, it is only a few thousandths ( .004 max above). i am looking for way to eliminate outliers and any possible machine "noise" to see if that corrects any of the deviation or at least lets me know i am accurate in my reporting. the print is very basic in that these surfaces have no location drawn on the print for them, it just says use the CAD model for location.
    Any help is always appreciated. Unfortunately I cannot share prints due to ITAR.

  • #2
    Make sure you guys have your alignments & probing methodologies between the two programs nailed down & both yielding correlating/repeatable results. Once you can say you have achieved that..and you still have bad points ANYWHERE on your "all over" surface relative to ABC, then your part is out of tolerance.

    Your requirement is "0.010 to ABC all over", not "0.010 to ABC in the spots where it measures good!"
    Last edited by DAN_M; 06-01-2020, 09:26 AM.

    Comment


    • #3
      If you have areas that exceed the boundary, then it's out.

      Comment


      • #4
        thanks for your responses i understand both of these comments, but how i do i determine if its an outlier causing the deviation, for instance the first time i ran the part it was out .008 in one corner, then i cleaned it with alcohol and reran it and it was good in that area. So how do filter the outliers or other possible machine/software causes? in some areas, if i turn on graphing, there is just one or maybe 2 points that are out.
        as for the correlation to the customers machine it may be impossible, they are using calypso to my Pcdmis. And DAN M, your point
        "Your requirement is "0.010 to ABC all over", not "0.010 to ABC in the spots where it measures good!" is exactly why im trying sort this out, ultimately if its bad thats fine ( sort of ) but i want to be able to without a doubt say that.

        Comment


        • #5

          JEFMAN can this be done using some type of array filtering? Measure all of the profile points, collect all of their ".Z actual values" in an array, and do some of your fancy-robot math to see if there are outliers that are greater than 0.002" different from the average array value, automagically subtract those outliers, and reconstruct a feature without the outliers to then report?
          Last edited by DAN_M; 06-01-2020, 02:11 PM.

          Comment


          • #6
            that would be awesome if thats possible. i knew i went to ask the right people

            Comment


            • DAN_M
              DAN_M commented
              Editing a comment
              This is possible, just gotta get the right people awake on the forum

            • charliejoe
              charliejoe commented
              Editing a comment
              yeah i figured, no rush, i have a meeting tomorrow to discuss progress on the findings. but otherwise no one is asking me to have it done yet.

            • charliejoe
              charliejoe commented
              Editing a comment
              any updates, or any way to contact the "right people" im starting to run out of time.

          • #7
            Originally posted by charliejoe View Post
            thanks for your responses i understand both of these comments, but how i do i determine if its an outlier causing the deviation, for instance the first time i ran the part it was out .008 in one corner, then i cleaned it with alcohol and reran it and it was good in that area. So how do filter the outliers or other possible machine/software causes? in some areas, if i turn on graphing, there is just one or maybe 2 points that are out.
            as for the correlation to the customers machine it may be impossible, they are using calypso to my Pcdmis. And DAN M, your point
            "Your requirement is "0.010 to ABC all over", not "0.010 to ABC in the spots where it measures good!" is exactly why im trying sort this out, ultimately if its bad thats fine ( sort of ) but i want to be able to without a doubt say that.
            I would ask what type of evaluation settings your customer is using on Calypso. I'm very much a novice when it comes to PC-DMIS, but in Calypso there is a LOT of different filtering options.

            Calypso EVAL settings.PNG

            That's not including outlier elimination either.
            Attached Files

            Comment


            • #8
              im not sure what they have that set at, it is definitely worth asking, i have a calypso background too, and i agree the filtering seems more developed, but PCDMIS has grown on me .

              Comment


              • #9
                I would ask the customer to provide or see if you can find a report that specifically defines area of failure so that you better try to correlate results. What specifically is failing? Feature , Basic where? All Around could be any where. What methods were used to define the error for rejection? How many parts were shipped? Inspected? You need some information and detail to understand the issue here to deal with a complaint like that.

                Comment


                • #10
                  i have a calypso printout. i have a meeting today to discus the part, hopefully i get more info as to what they are asking for, and then if i can get more info on this thread, maybe i can finish this project up. thanks for the help so far everyone.

                  Comment


                  • #11
                    customer will be here this morning, thanks to everyone tha tried to help, if any answers are still available, id like to learn for future use.

                    Comment

                    Related Topics

                    Collapse

                    Working...
                    X