iSixSigma

GR&R on Two New Visual Inspection Systems

Six Sigma – iSixSigma Forums General Forums Tools & Templates GR&R on Two New Visual Inspection Systems

This topic contains 4 replies, has 3 voices, and was last updated by  Joel Mason 1 day, 3 hours ago.

Viewing 5 posts - 1 through 5 (of 5 total)
  • Author
    Posts
  • #236160

    TNH1
    Participant

    Hello,

    I am working on two new visual inspection systems that are supposed to be functionally identical. They are to inspect parts on a board. Each part yields a pass or fail. The final result passes if all parts passed and fails if at least one part fails.

    Should I run a gauge R&R on each part tested (almost 100 parts) or is the final test result is sufficient?

    From what I’ve been reading, these are attribute data and that an attribute agreement analysis should be performed.

    How do I make the correlation between the two test systems? I read about the Bland-Altman approach, but I am not sure if it is applicable to my application.

    Please provide me with some guidance as to what the best approach is in performing Gauge R&R for these new test systems.

    Thank you and best regards…

    0
    #236198

    Mike Carnell
    Participant

    @tnh1 I am just going to give you what I would do or what I am thinking I would do. This is going to be what we call Jerry Garcia Theory (stay as high as you can for as long as you can except this is high level view not drugs). There is no reason to dive into this deeply until you have a reason to do so.

    If you do the attribute agreement study you are going to get percentages as outputs. Nothing specific. I am also making the assumption this is component inspection on a PCB? With the visual inspection you should be fixtured so everything is in the same spot all the time which means there is no operator intervention so reproducibility doesn’t mean much. You might want to take a fairly dense board and run it the way it normally runs and verify. Reproducibility should be negligible. Based on that I would separate the 100 boards onto groups and run the Attribute R&R substituting the board type for operator.

    If your vision system doesn’t change much from one group to another then you can probably assume they are the same (it can be set up as a hypothesis test).

    I would start there before I ran 100 R&R studies.

    0
    #236242

    TNH1
    Participant

    Hi Mike,

    Thank you for the response. Yes, this is component inspection on a pcb.

    For clarification, I am inspecting over 100 components on a single board.

    What do you recommend the minimum number of board per group?

    0
    #236260

    Mike Carnell
    Participant

    @tnh1 I wouldn’t think the number of components on the board would be as big an issue as the density of the components. This is kind of a pretty basic test. It has been 30 years since I did this specifically on a PCB so our visual systems were pretty rudimentary. The concept doesn’t differ.

    You are using Attribute data so sample sizes can be pretty high. It you plan to use a Hypothesis test specifically I would select the sample the same way you do any other time you use those test. If you are just going to eyeball it which has some risk attached to it then I would run it around a 25 piece test. For Attribute data that doesn’t give you a huge amount of confidence but you should be able to see a large difference. At the end of the day you can run it multiple times and also check to see if you have a time to time difference.

    Remember if you run this over multiple days you need to run some of every board every day other wise you will confound board type with time.

    I normally don’t like to run this loose but right now you are just looking for direction befor you commit a large amount of time.

    Good luck.

    0
    #236435

    Joel Mason
    Participant

    @tnh1,

    Thomas Rust of Autoliv gave a great presentation on attribute MSA at the 2016 ASQ Fall Technical Conference. To get the recordings, you can contact ASQ. I’m not sure what the charge would be, but I’m guessing it will be marginal. I have found their costs to be very reasonable. Since you have a vision system, you more than likely have an underlying variable characteristic even though it turns that information into a pass-fail judgment. In his presentation, I thought Thomas did a nice job communicating how you can leverage underlying variable characteristics in an attribute MSA. For example, the vision system is more than likely calculating a count of pixel matches. With the resolution now available with these kinds of systems, that kind of data approaches a continuous measure. Leveraging that can get you around the sample size problems Mike mentioned in an attribute agreement analysis.

    Joel

    0
Viewing 5 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic.