iSixSigma

Questions About Sampling Procedures in Agriculture Production

Six Sigma – iSixSigma Forums General Forums Tools & Templates Questions About Sampling Procedures in Agriculture Production

  • This topic has 7 replies, 5 voices, and was last updated 9 months ago by UL.
Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
    Posts
  • #244405

    Alberto Brito
    Participant

    Hello everyone.

    I work for a company that produces Sugar Cane. When Sugar Cane is harvested some percentage of the total amount is left in the field (it would be very expensive assure that all the Sugar Cane in the field arrives to the factory).

    So, my team and I need to create a Sampling Plan to measure how much Sugar Cane is left in the field; to do so, after the harvest, you need to select an area of land (measure in hectares), collect all the Sugar Cane left in the field, weigh what was collected (measure in tons) and finally you calculate how many tons per hectare of Sugar Cane is left (for example if you select an area equal to 0.25 Ha. and you collect in that land 0.5 ton of Sugar Cane left, the measure would be 0.50/0.25= 2 Tons/Ha). Then you need to measure the percentage of Sugar Cane left, taking in to account the total amount of Sugar Cane in that area (lot) (for example: if that lot produce 100 tons/ha and 2 tons/ha was left, 2% of the total amount of sugar cane was left in the field; if that lot have better conditions and produce 125 tons/ha, 1.6% of the total amount of sugar cane was left in the field).

    The main question is: How much area of land do I need to select to collect the sugar cane and measure it? Some considerations are:

    • All the lots have different areas (may vary from 5 ha. to more than 100 ha.)
    • All the lots have different productivities (may vary from 60 tons/ha. to 150 tons/ha.)

    Here are two alternatives the we have discussed (using ANSI/ASQ Z1.9):

    1. Take always all the area of the lot as 100% (independently of the total area of the lot), if you use a AQL of 6% and assume that you lot size always  will be 100 (independently of the total area of the lot), you will always have a set Inspection Level (General II – F). The Variabilty is Unknow and we have a Single Specification Limit so the sample size wil be 4 (4% of the area of the lot)
    2. Define the Inspection Level every time that we take the measure, taking as units the total area of the lot. This is somekind complicated.

    Is one of these two alternatives correct? If not: Can we use ANSI / ASQ Z1.9 in another way to establish the sample size? If not: What other alternatives do we have?

    Thank you very much in advance for your responses.

    1
    #244445

    Strayer
    Participant

    I was going to let y0ur complicated question go for those contributors who are much more knowledgeable about statistics and sample size.  But on second thought, maybe you’re overcomplicating the matter.  Why not measure the waste on a hectare, or even just an acre?  Do RCA to identify the root cause of waste.  Improve.  Remeasure.  If that reduced waste, standardize the improvement.  Repeat, using a different hectare or acre the next time.  Granted with an annual production cycle it may take years to significantly reduce waste.  But that will also be the case if you employ more complicated statistics.  Consider that the difference between waste vs. production on one hectare/acre vs. another probably has more to do with harvesting process than any other variable, and harvesting process is probably the same.

    2
    #244459

    Alberto Brito
    Participant

    Thank you very much for your response @Straydog, I totally agree with your point of view of how we can improve the harvesting process. Nevertheless the Managers want to this measure become a KPI in the near future, is going to be measure frecuently, so we need a Sampling Procedure to determine the areas.

    2
    #244460

    Chris Seider
    Participant

    Yes, the power of statistics is to use RANDOM sampling to get population mean estimates and see if you can shift those estimates after process changes.  Yes, it could be the same for variation.

    1
    #244496

    Strayer
    Participant

    I might suggest that you do some root cause analysis to discover the KPI.  From what you say it appears that the harvesting process is the source of waste and the root cause lies somewhere there.  Assuming that the harvesting process is the same in all hectares/acres, the variance in yield/waste per plot is probably noise.  You’re being led down the garden path if your KPI is noise.  You’ll put a of effort into collecting and analyzing meaningless data.  If you’re looking at the harvesting process you may not need to randomly sample various plots.  You may only need to look at one test plot.  I’d advise to do a fishbone (Ishikawa diagram), maybe some additional RCA tools, and do some research on harvest waste for this and similar crops.  Convince yourself that harvesting process is the source of waste but don’t assume it is.   Find the KPI.  If you can do that, you should be able to sell it to management and avoid a lot of wasted effort.

    2
    #244558

    MikePorter
    Participant

    You may want to consider using a Variogram (geostatistics tool) to help decide what a representative sampling size should be.  Geologists use this to help determine the characteristics of an area of interest.  The US DOD has also used this tool for sampling an area hit by a chemical/biological weapon attack.

    I have used it in determining sampling frequency in Chemical Production processes.

     

    1
    #244563

    Alberto Brito
    Participant

    Thank you All for your kind responses.

    The Variogram looks very interesting @MikePorter, this is a new tool for me. Is there a website or book where I can learn more about it and how to apply it?

    Thank you very much in advance.

    0
    #244642

    UL
    Participant

    If you can try cluster sampling where you have clusters of equal areas ie homogenous and internally heterogenous and then you randomly select clusters.  advantage is clusters will be of equal area and thus it will be easy to arrive at conclusion.

    0
Viewing 8 posts - 1 through 8 (of 8 total)

You must be logged in to reply to this topic.