iSixSigma

Short Run SPC and Tool Wear

Six Sigma – iSixSigma Forums Old Forums General Short Run SPC and Tool Wear

Viewing 14 posts - 1 through 14 (of 14 total)
  • Author
    Posts
  • #67866

    Neil Polhemus
    Participant

    There are control charts specifically designed to monitor tool wear. STATGRAPHICS contains one such chart designed to handle wear which follows a linear trend.

    0
    #67868

    denton
    Participant

    The basic rule for any process that has a known, non-removeable pattern is to find the pattern, subtract it out, and control chart what is left. 
    This can be conveniently done in cases with a linear trend by performing linear regression, storing the residuals, and doing your control chart on the residuals.  In this case, the residuals are your detrended data.
    A good way to think of this is that the control chart of the residuals lets you look for extraordinary deviations from the trend, rather than from the average.
    Denton Bramwell
    Sr. Master Black Belt
    [email protected]
     

    0
    #67869

    Neil Polhemus
    Participant

    Good point. A special toolwear chart does have a couple advantages, however. (1) You can specify the slope of the line if desired, rather than estimating it from the data. (2) You can add a specification limit to the plot to tell when you are getting close to needing to change the tool.

    0
    #27604

    MSAFAI
    Participant

    Hi everybody,
    Here is a case in SPC, that I try to describe below:
    1- It is a process using a perishable tool with high Cpk (2 or more),
    2- There is tool wear, that causes the part diameter to gradually increase (indiacted by a slow trend on the X-bar chart),
    3- Change of part diameter cannot be helped by readjusting the machine/tool. The only way is to change the tool with a new one.
    4- The production is typically a “short-run” production. Completion of each order takes one to maximum 3 days.
    5- After completion of the first order the tool is NOT changed, but the production goes on using the same tool for another product having the same diameter, but maybe different length, until USL (upper tolerance) is reached.
    Any idea about what kind of Control Chart to use in such a situation?
    I thought maybe a Nominal Chart would be suitable?
    ThanksMSAFAI

    0
    #67895

    Dewayne
    Participant

    All of the information supplied here will obviously be helpful, but it seems to me that the root of the problem is not being addressed. As I understand it, (1) the product must have a Cpk of 2, so it may/will be difficult to meet the requirement as it is; (2) the capability of the tool’s accuracy is reduced after each run by the manufacture of another item, produced until a worse-case condition is reached, wearing out the tool completely.
    As more than one tool will be needed anyway, why not require production to remove the tool to assure that the tool’s capability, life and availability is not ruined?  After the tool has worn to the point of not holding the Cpk 2 requirement, then let it be used for production of the non-sensitive product. 
    It seems to me that production convenience is being traded for potential difficulty with meeting the Cpk 2 product requirement.

    0
    #67897

    MSAFAI
    Participant

    Dewayne,
    1- Sorry for not being accurate; the Cpk is actually around 3 .
    2- You understood the situation completely. They “eat away” the poor tool until the product dimension is near the tolerance. Mind you, in this case the tool is perishable (not to be repaired).
    But I hope you agree, it is difficult to convince production to throw away a die which can still produce within tolerance.
    Using the tool for less sensitive application, can be a good idea.
    To me, the best solution seems to use Modified control limits, as one of our friends pointed out.
    Thanks again for your input,
    MSAFAI

    0
    #67898

    MSAFAI
    Participant

    Dear Neil and Denton,
     
    Thank you very much for your comments. Surely very useful and enlightening.
     
    To be honest, I am only afraid that such control charts with slanting control  limits can be difficult to use and to maintain on the shop floor.
     
    How about “modified control limits”, which allow larger shifts in the process average ? They seem to be easier to maintain, or not ?
     
    Regards
    MSAFAI
     

    0
    #67905

    Tommy
    Member

    How about using an EWMA chart, the weighting would normalize the effect of the wear I believe. Minitab has such a chart, and it is also explained on the itl.nist.gov web site in their statistical handbook.

    0
    #67915

    MSAFAI
    Participant

    Tommy,
    Thanks for the hint.
    I don’t think EWMA chart could help. If I’m not mistaken this chart is used to detect small changes in the process average. But in our case this also change is not important to us, it is part of the process. But again I may be mistaken.
    I quote from the same site you had mentioned:
    ” An EWMA control chart is a data analysis analysis technique for determining if a measurement process has gone out of statistical control. Similar to the cusum control chart, it is better than the standard xbar control chart for detecting small shifts in the process mean. “
    Regards
    MSAFAI
     

    0
    #67931

    Fausto Flores
    Participant

    I would suggest to use the moving average chart. Whith this kind of chart you can calculate and “incline” set of control limits which follow the wear trend of the tool, and it lets you know when to change it more precisely.
    On the other hand, if the process shows a cpk of 2 or bigger and with a low trend of wear, then I think we could simplfy things and use the common x/r charts and make sure everybody is aware of this expected performance and set the criteria to know when the tool should be changed.

    0
    #67962

    Schuette
    Participant

    I think I have to agree with Tommy.  The process your are depicting is somewhat similar to a process I dealt with a few years ago and when I saw your posting I immediately thought of the EWMA chart.  We had a centerless grind operation where normal tool wear cycles required periodic redressing of the grinding wheels and caused you to continually “bounce-off” the upper control limits of a conventional xbar control chart.  By not acknowledging the ‘normal’ small shifts in the process mean in this process and thereby adjusting control limits, you had an xbar chart with a saw tooth pattern.  You were aslo continually violating the standard Western Electric rules.
    We had numerous consultants, many at the doctorate level of education in statistics look at this process, and virtually every one ultimately came down to the EWMA as the best application.  And, yes, it was tough at the operator level and required special training.
    No offense, but I twinged a little at the comment that this small shift in process mean is not important to you, it is part of the process.  Aren’t your parts a product of the process and your goal to control the process not just chart it?
    Aren’t these discussions great!?!

    0
    #67998

    MSAFAI
    Participant

    Thank you Jim,
    I’ll check again, maybe I got it wrong.
    Regards
    MSAFAI

    0
    #124447

    ameneh
    Participant

    i m searching papers about “control chart for tool wear” .may you help me? if you have any things about it , please send me. i seriousely need them.
    my email address is”ameneh _shga @ yahoo.com”
    thanks,thanks,thanks

    0
    #124468

    MSAFAI
    Participant

    MSAFAI,Well, I’ve read all of the postings listed including yours, and would like to give it a try. As I understand here are some of the conditions you cited:1. The process has a high capability, above 2 and closer to 3.2. The tool is used until it can no longer produce acceptable parts.3. It appears the production folks really do not want to cycle the tools out simply because it is out of the tight control limits, and I agree. (It’s not very econmical)4. You would prefer not using a complicated control chart approach like adjusting the control limits to tool wear slope, EWMA. (may not be necessary due to the high process capability)So, what should you do that’s simple, easy to maintain, makes sense from a process control and economics perspective, and can be implemented quickly?How about taking your own suggestion, and adjust the existing control limits from +/-3SD’s to a larger number of SD’s from the center? While using +/-3SD’s provides the most economic limits for most processes, I’ve used +/-4SD’s for processes that are the most well-behaved, i.e., Cpk > 2.0, or potential 6 Sigma Quality Level. This provides a little incentive to those folks doing process control extremely well. For processes that have poor capabilities, I typically reduce the limits to about +/-2.5SD’s to insure the process team investigates the causes of high variability more frequently. 2.0, or potential 6 Sigma Quality Level. This provides a little incentive to those folks doing process control extremely well. For processes that have poor capabilities, I typically reduce the limits to about +/-2.5SD’s to insure the process team investigates the causes of high variability more frequently. 2.0, or potential 6 Sigma Quality Level. This provides a little incentive to those folks doing process control extremely well. For processes that have poor capabilities, I typically reduce the limits to about +/-2.5SD’s to insure the process team investigates the causes of high variability more frequently. Control limits set at +/-3SDs level of control provide a limit width in terms of the spec range equal to approximately 1/3(USL-LSL), i.e., Cp=3. Control limits set at +/-4SDs level of control provide a limit width in terms of the spec range equal to approximately 2/9(USL-LSL) yielding an approximate an worst-case capability(Cpk) of 1.67, or about 5-Sigma Quality Level of control. Extending the control limits out to +/-5SDs provides a worst-case capability(Cpk) of 1.33, or about a 4-Sigma Quality Level of control. My suggestion to establish a balance between economics and quality to extend the control limits out to between 4 and 4.5SDs, and establish a tool replacement when two or more values exceed these limits.Approximate defect levels given the following capabilities are:Cp Cpk SQL Approx Defects3.0 3.0 >>6 ~0
    3.0 2.5 >6 ~0
    3.0 2.0 ~6.3 0.78 dpmo
    3.0 1.5 6.0 3.4 dpmo
    3.0 1.0 ~4.0 ~9,000 dpmo (~1%)>6 ~0
    3.0 2.5 >6 ~0
    3.0 2.0 ~6.3 0.78 dpmo
    3.0 1.5 6.0 3.4 dpmo
    3.0 1.0 ~4.0 ~9,000 dpmo (~1%)>6 ~0
    3.0 2.5 >6 ~0
    3.0 2.0 ~6.3 0.78 dpmo
    3.0 1.5 6.0 3.4 dpmo
    3.0 1.0 ~4.0 ~9,000 dpmo (~1%)6 ~0
    3.0 2.5 >6 ~0
    3.0 2.0 ~6.3 0.78 dpmo
    3.0 1.5 6.0 3.4 dpmo
    3.0 1.0 ~4.0 ~9,000 dpmo (~1%)6 ~0
    3.0 2.5 >6 ~0
    3.0 2.0 ~6.3 0.78 dpmo
    3.0 1.5 6.0 3.4 dpmo
    3.0 1.0 ~4.0 ~9,000 dpmo (~1%)6 ~0
    3.0 2.0 ~6.3 0.78 dpmo
    3.0 1.5 6.0 3.4 dpmo
    3.0 1.0 ~4.0 ~9,000 dpmo (~1%)6 ~0
    3.0 2.0 ~6.3 0.78 dpmo
    3.0 1.5 6.0 3.4 dpmo
    3.0 1.0 ~4.0 ~9,000 dpmo (~1%)Ok, there’s my shot.Ken
    Ken

    0
Viewing 14 posts - 1 through 14 (of 14 total)

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