iSixSigma

Factor Level Settings

Six Sigma – iSixSigma Forums Old Forums General Factor Level Settings

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #50240

    Dan B
    Participant

    Can anybody tell me a rule of thumb for deciding on effective High and Low settings for a DOE?
    If, as an example, I traditionally run a process temperature setting at 150 degrees F, for the purposes of a DOE what would be appropriate hi/lo settings to use in conducting the experiment.
    I don’t have access to the MINITAB Response Optimizer, which I’ve heard is one way…does anybody know a practical way to figure out the settings?
     

    0
    #172850

    Stephens
    Participant

    There is no rule of thumb that I am familiar with. Instead you need to consider a couple of things. Firstly, what would be sensible temperatures for ongoing operation? There is no point carrying out experiments if the conditions are unrealistic or impractcal in the long term. Secondly, you need to consider how well you can control the temperature. If the variability in your temperature control is say +/- 5°F then you would want at least 10°F between levels, preferably more. Setting a centrepoint at 150°F in your case would make good sense and setting levels at +/- 10°F would seem fairly reasonable for an optimisation study. The results will tell you if you need to go any further.

    0
    #172855

    k.bhadrayya
    Participant

    Dear Mr. Dan.B
    Setting the factors levels is important and tricklish.Many DOE expts fail because of improper setting leves of factors. I am giving below some tips based on my practice.
    If it is an exsiting problem, from th eoperations / log sheets the extreme values can be collected and discussed with the concern engineer the scope of further incresing the limits, or theoriticall possible ones. Th elowest level is viewed as -2 level and highest level is viewed as +2 level.
    Then the lowest and highest levels can be added and divided by two to get mid point. Now again add mid point and lowest level and divide it by two to get -1 level. If you add highest level and mid point and divide it by two, you get + 1 level. Thus you have 5 levels from -2, -1, 0 +1, +2 levels for each factor.
    In DOE to start with it would be better to take -1 and + 1 levels and start experiemenation. After conformation that the relation ship between input factosr and out put is non linear, then we can use -2 , 0 , +2 levels to obtain quadratic mdel.
    Further any doubts , you amy mail to [email protected]
    k.bhadrayya
     
     
     

    0
    #172856

    Venugopal G
    Member

    Hell Mr. Dan.B,
    For optimization shainin DOE technique- variable search is an excellent tool. I know conventional DOE techniques i.e., factorial, fractional factorial, taguchi etc in GB training, but when I attended BB training application of shainin has saved me lot of time and is also very easy and has given me good results. If you are interested I can guide you. You need not to be worried about correct levels selection, bcoz in conventioanl DOE if your levels choosing was wrong you will come to know only after the end of the experiment. But in shainin DOE unless and until you prove that levels chosen or correct, you cannot move further for experimentation. So there you will save time, money etc. I am giving a brief introduction for your interest. But please note that you should observe and note down quality parameters first and production parameters next. Every data of QA parameter is important and needs to be noted down.
    1. First list all parameters for DOE
    2. Identify current levels and proposed levels. Current level is where you are currently operating the machine at that setting. Proposed level must be from the technical knowledge like if you want to increase output you have to increase the temperature etc.
    3. Now conduct the experiment 1A when all setting at current level and note down the QA parameters (also production). Then experiment 2A when all settings at proposed level and note down the QA parameters (also production). 1A & 2A are called as Run-1.
    4. Repeat step – 3 thrice. (Run – 1,2,3).
    5. With above runs we will know that whether there is a major change in QA parameters ( there is a small analysis to calculate change). Now if so we will find the optimal setting between good value of QA parameter and bad value of QA parameter, that is optimization. If you dont get major change, that means you have further scope for new levels identification and your earlier proposed values in step 2 will become current and you will find another new proposed levels. Then repeat step 3&4 and then based on step 5 analysis we will design further experiment.
    Even we can reduce number of runs with the simple technique of  this shainin DOE. Please contact me if you are interested.([email protected])
    With regards,
    G. Venugopal.
     
     
     
     

    0
    #172857

    Ron
    Member

    Remeber in a screening experiment you are trying to force decisions. Therefore when selecting your factor levels they should be at the maximum and mimimum that the prodcut going throught the process will experience.
    It is always best to have knowledgeable engineers, and Mfg engineerrs and Quality Engineers work together to define these limits.
    i.e. if a temperature specification states to run a process between 70C and 100 c those would be your two level s

    0
    #172858

    Jonathon Andell
    Participant

    A general guideline is to set the “+” and “-” as far apart as you can without causing damage, either to the output of the process or to the equipment & people operating the process. Your team should be able to figure that out.

    0
Viewing 6 posts - 1 through 6 (of 6 total)

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