iSixSigma

StuW

Activity

  • I sympathize with your post. 
    Finding “enlightened” management for LSS (or TQM) is more than half the problem.   I was in a situation where management changes over a long period of time led to a swing from a team based proactive approach to problem solving, to a top-down mandated action approach.   While the new managers did not immediately undo…[Read more]

  • StuW replied to the topic Adjust IMR UCL/LCL? in the forum General 12 years, 1 month ago

    Just a few more things to consider based upon the added information provided in your last note:
    Have you checked for “fixed effects”, in other words, differences in respnse means between the brines?   These could be small shifts, but nonetheless, important in how your I-MR chart looks.
    Have you done a variance components analysis and checked f…[Read more]

  • StuW replied to the topic Adjust IMR UCL/LCL? in the forum General 12 years, 1 month ago

    I would check to see if your data shows positive auto correlation.  That would account for a series of measurements that appeared to run closer to the center line than expected.   Use the Autocorrelation and Partial Autocorrelation options under the Time Series and Stat menus in Minitab to check it.  If the data does show auto correlation than you…[Read more]

  • I’m not sure the question you are asking based upon how it is posed, but the usual sample comparison is for DOE sampling versus “one trial at a time” iterative sampling.   In that case, the DOE approach will often provide enhanced learning and require fewer samples because of the ability to estimate both main effects and interactions.   This allo…[Read more]

  • StuW replied to the topic Auto correlated data in the forum General 12 years, 1 month ago

    For highly autocorrelated data, either positive or negatively correlated, there have been multiple papers published in statistical journals over the last 25 years on how to make adjustments to standard control chart limits.  If you are in that position, I’d suggest researching those papers.
    In some cases, the autocorrelation can be significantly…[Read more]

  • I think you’ve answered your own question.  If the sample size changes daily, and units are either good or bad, then as presented the p-chart is the way to go.  If you’re counting multiple defects as a possibility for each unit, than a U-chart sounds appropriate.
    One possibility if they really want to show integers on the charts would be to r[Read more]

  • It is possible I don’t fully comprehend your specific situation entirely, however you’re still making some incorrect statements such as this one: 
    “First, you pose the regression model of the degradation in the machining behavior, and then use the model as a way to remove the trend in the data.  That approach ignores two major common causes to…[Read more]

  • I’m not sure where to start on this other than to wonder why you are addressing a 3-year old topic, but there are so many issues with the note that I feel compelled to add a response.
    On the use of SPC for “zig-zag” type machining data, there are ways to address this statistically, and not just using a guardband approach as you have…[Read more]

  • Our organization was all about taking noise out of processes.  Because the Semi world is so capital intensive, the way most older fabs survive is by taking the variability lower on a tool and finding out how to make it perform even better than the manufacturer proposed in their specs.   There are many 20-30 year old tools still in use where the…[Read more]

  • Forrest, I agree with your assessment in this case, and it will often net additional information by doing so.
    On your article, I’d make a couple of points.   I agree transforming to the most appropriate response is the right behavior although it can be difficult to convey to engineers and production workers, particularly if they are not well e…[Read more]

  • I used “generally” in the earlier post as most of my experience is in the semiconductor industry, and I am giving my insights into the question posed.   From what I have seen, it is ususally much easier to adjust a tool to a target then to figure out ways to reduce variation, which is typically much more challenging to do.  For example, many di…[Read more]

  • Generally, in my experience, moving the mean is much easier than reducing the variation.  If the desired process target is known, than it is often a first order to adjustment to a process input, such as an input temperature or time, or another similar adjustment in order to reset the process to a desired setpoint.
    Reducing variation requires ev…[Read more]

  • There are many instances where customer specs may not exist, for example early in a process where the results are not easily translated into a customer impact.  For example, in electronic circuit production there are dozens are early steps used to set the base silicon properties which are very difficult to relate to final customer related o…[Read more]

  • This was not my industry, but I will say it is certainly possible to conduct DOE’s where the final assessment of the results takes weeks, or months to get.   In the electronics industry, the impact to the final device yields may take months to see after the design takes place.
    In your case, I am more concerned as to whether you will have the d…[Read more]

  • A couple of relevant points to consider.   Prediction intervals will always be wider than confidence intervals, that is the nature of what you are asking.   In the confidence interval case you’re asking for the variation in the average response at a given value of (x1,x2), while in the prediction interval case, you’re asking for a prediction for…[Read more]

  • StuW replied to the topic Lean Confusion…Still in the forum General 12 years, 1 month ago

    I think you are confusing what the multiple lanes provide in a supermarket.  These lanes are not inventory, the inventory is the customers waiting to be served, or processed through.   The additional lanes serve as ways to handle increased volume of customers.  The more lanes, the more options for the WIP to be handled.  
    Think of it as a singl…[Read more]

  • A caveat to begin.  I worked in the electronics industry for many years, not the health industry.   I don’t know if regulations exist, or not, however, your company should have guidelines to address dealing with sub-contractors or suppliers, and I suggest you start with those.   In my industry, we would not deal with suppliers who would not ent…[Read more]

  • StuW replied to the topic Scrap Codes in the forum General 12 years, 1 month ago

    The benefit to making your own codes and operational defitinitions is that it will fit your business.   Sure, there are generic codes such as “Broken”, “Too Large”, or “Too Small”, but you want to set these up to reflect the actual operation under study.
    Here are the guidelines.   Every category needs to have a definition that describes for eve…[Read more]

  • StuW replied to the topic Lean Confusion…Still in the forum General 12 years, 1 month ago

    Don’t confuse process velocity with Little’s Law, which is the lead time, as the latter is the denominator in the computation of Process Velocity.   In other words:
    Velocity = # Steps / Lead time = # Steps / (WIP/Completion rate)
    Reducing the WIP, decreases the overall denominator which is lead time, and that increases process velocity.  As an…[Read more]

  • Consider a one-sided CpL estimate, with the lower specification being the minimum value of acceptable on-time delivery.  For example, 85% might be the value of interest.   You’ll need a rational sampling scheme for collection, in other words weekly data, monthly data, etc. where the number of transactions is reasonably similar (hundreds, or th…[Read more]

  • Load More