iSixSigma

help defining a subjective defect

Six Sigma – iSixSigma Forums General Forums Methodology help defining a subjective defect

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
    Posts
  • #53680

    Smith
    Participant

    does anyone have experience/suggestion as to how to define a subjective attribute defect?
    We make painted speaker grills with many visual requirements such as “no heavies” or “no slugs”. When a part with thousands of speaker grill holes has a small amount of paint covering or partially covering a hole, many different people consider the part good or bad and there is no adequate standard to define good or bad. Therefore, my attribute gage R&R is poor and difficult to perform since I have no standard.
    Any input would be appreciated
    Thx

    0
    #191072

    roarty
    Participant

    Have you considered ‘automated visual inspection’ – same technology as bar code reader. You can train the device to classify as pass or fail

    0
    #191074

    Hannabarger
    Participant

    General suggestions:

    1: Define and quantify “heavies” and “slugs” (go to CTC / CTQ criteria for the spec)
    2. Train process workers and inspectors on the new standards (hint: pictures of acceptable and not-acceptable conditions help)
    3. Identify those who will be making the accept / reject decision – put all of them into your study
    4. Design and run your MSA study
    5. Improve as necessary to get acceptable results

    Without well defined and understood standards, you cannot run a MSA study

    0
    #191075

    MBBinWI
    Participant

    Mike: While the other replies may be helpful, I don’t think that they actually answer your query.

    You think that you don’t have a standard, but you really do, it just needs to be codified and published. Somebody is making this evaluation and perhaps somebody else (quality perhaps?) has responsibility for setting the standard. One of these entities needs to be considered the “gold standard.” The evaluation of this individual/group identifies what is good and bad. All other evaluations done by through an attribute MSA are compared to the “gold standard” evaluation.

    As an aside, I had a recent experience similar to this. The “gold standard” made the evaluation and then subsequent analysis showed that even the individuals considered the “gold standard” were not accurate to what was really acceptable to the customer. So we needed to re-calibrate the “gold standard” to what was actually acceptable to the customer.

    BTW – attribute MSA’s are notoriously poor due to the subjectivity of the measurement device (humans usually). I would suggest that you search out a means of turning this into an objective metric (as suggested, a camera system that can evaluate each distinctive location – it doesn’t blink, get tired, come to work with a hangover, etc.).

    0
    #191076

    Smith
    Participant

    Thanks to all for the info. Yes MBBinWI, your recent experience is my current experience. Any vendor suggestions for automatic visual inspection?
    I can search on the web but proven, successful experience would be helpful.
    Ultimately, it is difficult to “recalibrate” the “gold standard” because the true “expert” is really the customer and whether he/she would reject a part based on their subjectivity.
    If I was to ask the customer to create the “gold standard” they will certainly err on the safe side which would result in us scrapping good parts.
    If I was not to ask the customer to create the “gold standard”, I’m sure they would tend to accept our parts moreso.

    0
    #191077

    Smith
    Participant

    Any vendor suggestions for automatic visual inspection?
    Thx for the input.

    0
    #191078

    roarty
    Participant

    A previous employer of mine used automated visual inspection to check that all components had been placed in packageing before being despatched from the warehouse. My role is in the R&D laboratory so I had nothing to do with the implementation of the system and cannot advise on vendors.

    These systems are based on neural network software. This technology is well established so I would suggest that most vendors can supply a quality product. Your decision should therefore be based on the level of support during implementation and the after sales service.

    Remember, because it is a neural network the system will have to be trained by a human expert and you will probably need the vendor to guide you in this implementation phase.

    As always, it costs nothing to invite a sales person in to demonstrate and advise on implementation.

    0
    #191083

    Hannabarger
    Participant

    We have no spec that we are aware of (other than some vague terminology), we don’t trust our customer enough to work with them to develop an agreed upon spec and inspection process, our workers have no idea how to inspect because they don’t know what they are looking for, and we’ve somehow jumped to a solution of vision systems to do the inspection for us. Seems this brings us back to square one: we have no agreed upon spec and inspection process with the customer.

    All of the wastes of end-of-line inspection systems are coming into play with this issue, in my opinion. My lean sensei once cautioned me about the wastes of automated systems that remove us from our intimate knowledge and understanding of the process. You might want to follow Deming’s advice and get some “profound knowledge” about the process before jumping to solutions.

    0
    #191123

    Andell
    Participant

    I think the solution may be to develop a standard that reduces subjectivity. This could be done in a team environment, perhaps even with inputs from a customer panel. Once that is done, you have the option of creating “standard” units for visual comparisons. With or without that, you should be able to train your inspectors in the consensus standard and repeat your attribute R&R.

    0
Viewing 9 posts - 1 through 9 (of 9 total)

You must be logged in to reply to this topic.