iSixSigma

Which is the true value?

Six Sigma – iSixSigma Forums Old Forums General Which is the true value?

Viewing 19 posts - 1 through 19 (of 19 total)
  • Author
    Posts
  • #34051

    Jackey
    Participant

    I feel I am in the jam of measurement. There is one optical measurement equipment on our production line which will check every finished parts and seperate them as “good parts” and “bad parts”. We selected some parts and measured them by optical measurement equipment, Smartscope ( an optical 3D measurement equipment) and Profile projector again. But unfortunatelly, we got discrepant measurement results from the different measurement equipment. Now I am puzzled which one is the correct result? And how can I approach the “true value”? Any ideas from you are all welcome!

    0
    #93357

    bitoy
    Participant

    Jackey,
    Calibrate you optical measurement gadgets.
    Bitoy

    0
    #93358

    Jackey
    Participant

    Hi, Bitoy,
    Thank you for your faster response! But all of three measurement equipements were all calibrated and the result were all OK. Why is there still so large deviation between these three equipments with the same parts?
    Jackey

    0
    #93360

    Arend
    Participant

    Your first concern may not be about the ‘true value’, but which measurement is the better measurement. Calibrated equipment does not imply that the measurement result is reliable per se. Calibration is a property of the equipment but not of the measurement. The measurement also involves the operator (if it isn’t autometed), the parts, environmental conditions and many more issues that are not covered by calibration.
    You might want to do an R&R study, to check if the results are sufficiently reproducible and repeatable. There is some good literature about this, a so I assume you’ll find your way. If not, just let me know.
    kind regards,
    Arend
    Senior Development Engineer

    0
    #93362

    bitoy
    Participant

    Jackey,
    Is it calibrated for both zero ( mean) and span(variation) adjustment? If yes, try checking the operators by attribute Gage R&R.
    Bitoy
     

    0
    #93363

    Jackey
    Participant

    Do you think it’s a better way that do Gage R&R to three measurement equipments respectively? Then the result of measurement equipment with high Gage R&R score is approach the “true value”.

    0
    #93377

    Craig D
    Participant

    If you are obtaining different results from 3 supposedly calibrated machines then you have an issue with Gauge R&R.However, resolving this will not necessarily give you the ‘true’ value. Every measurement system will have a tolerance and will be subject to common cause variation. Depending on how small the differences are you are trying to detect it might virtually impossible to get an exact measurement and you may have to consider such things as the affect of ambient temperatures on your measurements, both for the part being measured and the measuring equipment.The test of your measuring system is whether it is allowing bad parts through to the next stage of the process (your customer) or whether it is incorrectly identifying good parts as scrap (false negatives and false positives). Don’t get hung up on measuring the exact value ‘because you can’ instead concentrate on getting the spec limits correct and having a measurement system that consistently passes good parts and fails bad ones.Apologies if I have misinterpreted your search for the ‘True Value’ but in any case you have problem with your measuring system.

    0
    #93378

    Mike Carnell
    Participant

    Jackey,
    There are several different possibilities you need to investigate when you get this type of result. Remember there are 5 different types of issues around gages: bias (calibration), linearity, stability (probably not your issue in this case), repeatability and reproducibility.
    The data you have already says you have a reproducibility issue. If you were to run a R&R type study using one operator and substitute the 3 different systems for what is normally the operator catagory you can quantify the problem. That will only confirm what you already know. The issue then become technique or gage. You can separate that with a DOE (different gages and different operators).
    It can actually be around the level of discrimination. That you can get from a % Tolerance calculation by doing a normal R&R study (multiple operators on one gage – I am assuming you have continuous data). They could easily have different levels of discrimination.
    You can also have a linearity problem. Depending on how it is calibrated they may not pick this up during calibration. If you select your samples properly you can run linearity at the same time you run R&R.
    As far as a true value – that is by definition an accuracy (calibration) issue. You should be able to feed them parts with known sizes and see who gets it right.
    The accuracy problem can get confounded with optical equipment by geometry. Straight lines are easy to measure, curves can cause problems. Stratify some of your data and see if the disagreement between the systems is indigenous to a particular geometry.
    This is no different than any other SS project. You have a Y (disagreement between systems) – you need to drill down just as you would any other problem.
    Good luck.

    0
    #93380

    Marc Richardson
    Participant

    Jackey,
    Have you tried running the same parts (say 20) through your production inspection system at least 3 times? The parts you select to run should represent the range of production, in other words, make sure you have parts that both passed and failed on the first run through.
    You have an immediate issue that you may be passing nonconforming product to your customer. I would want to straighten that problem out before i worried about the accuracy of your other measuring equipment.
    Marc Richardson
    Sr. Q.A. Eng
     

    0
    #93394

    Thomas C. Trible
    Member

    Jackey:
    Your quandry is not at all unusual.  First, understand that there is no such thing as a “true” value.  Is there a “true value” for the speed of light?  Not really.  There is only the outcome of a measurement process.  Change the measurement process, obtain a different answer. Change the measurement equipment, get a different answer.  Within the limitations of the measurement equipment, answers from “A” and “B” may be correct.
    Measurement data collected from equipment A and B will have a distribution.  “A” may have better precision and have a narrower distribution.  On the other hand, “B” may be operating in a superior state of calibration, so that it’s producing data that is closer to the accepted mean.
    The place to begin to solve your quandry is to understand the accuracy and precision limitations of equipment “A” and “B”.  The place to start to solve your problem is with recalibration.  Once calibrated, measure an artifact, or reference standard with “A” and “B” and evaluate the difference beween the two measurement instruments.  Are the differences within the manufacturer’s stated tolerance?  If yes, then there is no problem with respect to defective equipment.  If yes, one instrument may be more precise, or more accurate than the other.  However, if the differences are not within the stated tolerences, then look to an equipment malfunction first, and to other factors that influence accuracy or precision..
    Good luck.
    TC Trible
     
     
    With

    0
    #93395

    Arend
    Participant

    Dear Jackey,I guess from your description that there is one type of testing equipment that is used for in-line measuring, and the others are instruments for other purposes like laboratory measurements. Is that correct? In that case, the requirements for these testers probably are different:
    -a tester that is used in comparing products against specifications must have a good R&R compared to the specification tolerances.
    -a tester that is used in a laboratory probably is used for comparing products against eachother. If that is the case, the R&R must be good against the spread between the products. This is a more strict requirement.Your first concern is the production measurement, and the question if it can be used for rejecting and accepting products. In first instance I would focus on this measurement.You mention that the final test result is a good/bad judgment. It is possible to do an R&R study in such cases, but it doesn’t have the same strength of an R&R study on a continuous parameter. In many cases, the good/bad judgment is generated by a computer after after measuring the part and comparing to the spec. Is that also the case here? If that is so, my advise would be to ‘forget’ about the computer judgment and focus on the actual measurement in the R&R study.

    0
    #93396

    bitoy
    Participant

    Jackey,
    Since all your gage are “correctly” calibrated, then, you try attribute G R&R using one of the gages with three operators and 30 samples. If you don’t know how to conduct attribute  G R&R, just post your e-mail and I will send you an excell file and instruction.
    If all operators agree with each other and to the randomly listed samples, make a chi-square test to your gages so that you would know if one or more of the gages are different.
    If there is a difference, pick the best gage that gives the least error as your benchmark.
    bitoy

    0
    #93401

    Mike Carnell
    Participant

    Thomas,
    I would only disagree with one part of your post. I would not start with the re-calibration. I would leave them alone ntil I had time to figure out which one(s) were the problem.
    Good luck.

    0
    #93420

    Thomas C. Trible
    Member

    Mike:
    The equipment recalibration process always begins by determining the “as found” condition.  The “as found” condition is determined by measuring a standard reference with equipment “A” and “B”.  Do the measured values for “A” and “B” fall within the manufacturer’s tolerance?  If they do, then we can conclude that A and B are operating in a calibrated state.  If A or B do not, then adjustment of the equipment is required, or if adjustment is not possible, then revision of the calibration factor is required – before further measurements are made.  If A or B are determined to be out-of-calibration, then we can begin to make sense of the differences between values measured, or found by A and B.
    Once we have determined that A and B are operating in a calibrated state, we can begin to draw conclusions about total measurement uncertanty – but not before.  Also, we can conduct further measurements with A and B to determine if the total measurement process is operating in a state of statistical control.
    When data genererated from measurement equipment are suspect, always begin the troubleshooting process by conducting the recalibration process as outlined above.  Gathering further data from out-of-calibration equipment, or equipment whose calibration status is unknown, or suspect, provides meaningless data.  No valid comparison conclusions may be drawn from that data.
    I refer you to Calibration: Philosophy in Practice. 2nd Edition, 1994, published by Fluke.  This excellent book may be purchased directly from Fluke.
    Good luck. 
     

    0
    #93422

    Mike Carnell
    Participant

    Thomas,
    Thanks for the reference.
    Good luck.

    0
    #93427

    walden
    Participant

    Jackey,
    There’s been a lot of posts offering good suggestions for performing a GR&R, but none have asked you any basic questions to actually define the problem you’re experiencing.

    What type of feature(s) are you measuring? (Length, diameter, etc.)
    What is the nominal size of the measurement?
    How much difference are you observing between the equipment sources?
    Are your observed measurements smaller or larger than you expect?
    Does your equipment use “edge finding” capabilities, or do the operators manually determine the edges?
    Does the equipment use profile or surface illumination, or both?
    Performing a GR&R is obviously recommended, but you’ll need to identify the sources of any errors to resolve the issue. Answers to the above items should help define the problem your seeing so possible suggestions for a solution can be offered. Of course, your answers may also lead to more questions being asked of you.
    Chris

    0
    #93459

    Jackey
    Participant

    Hello all,
    Thank you very much for your great comments and suggestion to my question.
    Hi Chris,
    I am also agreed with you that the answers to yuur questions are very important to solve this problem.
    My answers are following:1. We are measuring the coplanarity of solder pins of my fished part2. The norminal size is <0.13. The deviation between the results of different ME is very large, and the most interesting thing I observed the results are not on the same trend but interlaced with each other. By the way we didn't conduct a Gage R&R befor we did such test. I have to think we should do Gage R&R at first4. I am puzzled with the results now and I don't know what the sense of this questions? Would you please give me more comments on this questions?5. Online optical and Smartscope is using "edge finding" techinical but profile project has to be determined by manual;6. I think they are on both.Do you have more suggestion when you know my answers to your questions? Waiting for them :-)
    Jackey

    0
    #93482

    walden
    Participant

    Jackey,
    I have to apologize that I have no experience with coplanarity measurement. As such, I can’t offer any further real assistance with your problem.
    You may have already investigated this, but the only other items I can suggest are: 

    With any vision system, proper illumination is critical to obtaining accurate values and reducing variability. Especially when using automated edge finding and/or focusing. Hopefully your vision systems have an enhanced light source. Often times the basic light source that comes with vision systems is not sufficient for more sophisticated measurements, parts that lack good contrast, etc. An accessory to the SmartScope is the Smartring. This consists of 6 or 8 concentric LED ring lights. Each ring also has 8 sectors, which allow you to divide the “pie of light” into 8 slices. You can then turn on 1 or more slices to select the direction the light illuminates the object to be measured. Whether or not you have a Smartring, possibly try different levels/conditions of illumination to see if it helps.
    As far as establishing your planes, try increasing the number of points used in the measurement. As with radii measurement, an increase in the number of points used will increase accuracy and reduce variability.
    Good luck. I wish I could be of more help to you.
    Chris

    0
    #93505

    DaveG
    Participant

    Jackey,
    I can’t improve upon the good advice you’ve been given, except before I investigate an “event” like yours I always ask “Is this a problem?”  Will my customer feel the effect of these nonconformances?  Are the specification limits developed from objective correlations to CTQs?  If not, you may be chasing a non-issue.  However, if it is a true problem, you are increasing your chances of properly controlling your process.

    0
Viewing 19 posts - 1 through 19 (of 19 total)

The forum ‘General’ is closed to new topics and replies.