Control Charts on Bulk materials

Six Sigma – iSixSigma Forums Old Forums General Control Charts on Bulk materials

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
  • #35961


    My company makes paste-type materials in large batches, 1 at a time.  The material is made in bulk, and then packaged in smaller containers. Only 1 sample from each batch is measured.  Does it make sense to control chart the measurements?  Since each measurement is a from a new batch, there is not a continuous process that is being monitored.   Also, what type of chart is recommended, if at all.  I would think in individuals and moving range, since the sample size is always 1.



    I think it is definitely a good idea to use some form of control charts and the fact that you are dealing with a batch process does not make any difference.  Control charting will help you monitor your processing conditions and possible drift in your incoming raw materials that you may not capture by simply looking at COA’s.
    As far as what type of chart to use it depends on the type of data you are looking at (variable or attribute).  IMR chart would work for variable type data, but have you thought of taking more than one sample per batch to check for within batch variation (if that is important in your product).



    Yes, it is variable (usually viscosity).  We’re looking to add measurments per batch, but you are correct it could help.  I have doubts about measurement repeatability, and we’ll look at that before getting into SPC.


    Ken Feldman

    Are the process steps used to make each batch different batch to batch?  If not, then you are dealing with a process.  I agree it might be a good idea to check whether the measurement system of measuring viscosity is giving you valid information.



    One of the processes that  worked on in the past used a shear mixer to mix very large volumes of ink-jet ink … a mixture of MEK and various pigments.
    One of the problems you’ve correctly identified is that the mix has to be adjusted each time you make up a batch, due to variability in the raw materials, and does not reflect the natural variation in the process and therefore does not conform to one of the assumptions of Shewhart Charts.
    This procedure is futher complicated for us by the fact that our ink had three main responses: viscosity, conductivity, and print quality, which are suffer covariance.
    My recommendation would be to use a multivariate control chart for the unadjusted process if you have more than one response, and plot each unadjusted response on separate Shewhart Charts. This will give you a good idea of the ‘natural variation’ in the process, but it is unlikely that you’ll be able to work with the suppliers to improve assays – at least in my experience.
    You will have to treat this data with circumspect otherwise some auditor might tell your report that your process is not capable. I don’t see any point of plotting the adjusted process as the only goal is to adjust each response into specification, and to maintain the covariance of the responses. (In our case the print quality suffered if the viscosity was at the high end of the spec. and the conductivity at the low end of the spec. In other words, high viscosity requires a higher conductivity and vice versa – even when all responses are within specification. This is why we needed multivariate control.)
    Although we used a Heuristic approach to adjust, I would now consider using a Simplex method and try to adust ‘in a controlled way’, so as to improve the standard time of the process. This is how a well-known manufacturer of chocolate achieves similar results.)
    There may also be a way of using PCR to achieve the same result … perhaps Robert Butler, or some other eminent statistician, could give you a better insight; as this subject is the limit of my own capability!



    As I was reading the posts on this topic, it brought up a question. How would I handle control charting an extrusion process where a days work would have data coming from 4 separate alloy types? There is no order to what alloy type is run first…the orders are scheduled more or less based on extrusion speed. for instance, a fast runner may be the first order followed by a slow runner to enable the folks at the stretch and saw stations to clear the table in time.
    Also, there are no hard spec limits in this process. We are striving for a goal of 82% recovery, and industry average is 76%. I have month-to-month history of recovery rates for this process. Would I use a one-sided spec of 82% minimum?Also, since sample sizes are 1, I will be using an I-mR chart.
    Any advice would be greatly appreciated!

Viewing 6 posts - 1 through 6 (of 6 total)

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