Docstoc

Experimental Design

Document Sample
Experimental Design Powered By Docstoc
					Experimental Design

     Fr. Clinic II
                       Planning
• Begins with carefully considering what the
  objectives (or goals)are
  – How do our filters work?
  – Which filter is best?
     •   Performance
     •   Cost
     •   Ease of Use
     •   Durability…
                 Variables
• Process variables include both inputs and
  outputs - i.e. factors and responses
• The selection of these variables is best done
  as a team effort
                       Terms
• Factor
  – Something that can be varied, e.g.,temperature,
    pressure, material …
• Response
  – the results
• Level
  – Each variable can be “set” to or measured at different
    levels
• Run
  – How many times each unique combinations of factors
    and levels will be tested
        What are our factors?
• How do filters work?
  – Could be the membrane/porous filter, the
    activated carbon, the viral guard…
• Comparative Assessment
  – We have one factor, the portable water filters
    themselves!
Water are our possible responses?
•   Removal of contaminants
•   Produced water quality
•   Cost
•   Ease of use
•   …
        Experimental Design
• Choosing an experimental design depends
  on the objectives of the experiment and the
  number of factors to be investigated.
   Experimental Design Objectives
• Comparative objective
   – 1 or several factors under investigation
   – primary goal to make conclusion about 1 important factor
      • in the presence of, and/or in spite of the other factors
• Screening objective
   – select or screen out few important main effects from many
     lesser important ones
• Response Surface (method) objective
   – estimate response to multiple factors
      • find improved or optimal process settings, troubleshoot process
        problems and weak points, or make a product or process more robust
        against external and non-controllable influences.
         Experimental Designs
• Completely Randomized Designs
   – Comparative objective, 1 factor
• Randomized Block Designs
   – Comparative objective, multiple factors
• Full or fractional factorial
   – Screening objective, multiple factors
• Many more!…
 Completely Randomized Designs
• One factor, with multiple levels of interest
  – For example – effect of temperature on a chemical
    reaction
     • three levels, two runs each gives 90 unique orders to
       conduct experiment (e.g., T1, T1, T2, T2, T3, T3;…)
• In a completely randomized design, you would
  randomly select the order of runs
    Randomized Block Designs
• One factor or variable is of primary interest.
  However, there are also several other
  nuisance factor variables
  – Nuisance variables are those which may affect
    the measured result, but are not considered of
    primary interest.
     • For example: specific operator who prepared the
       treatment, the time of day the experiment was run,
       and the room temperature.
     • All experiments have nuisance factors.
Randomized Block Designs (cont.)
• Blocking
   – Run every level of the primary factor with the nuisance
     factor(s) held the same
   – Minimizes total # of runs
• Random
   – Run with nuisance factors selected randomly.
   – More runs may be required
   – Likely to get more variability (error) in results
• May be able to block some nuisance factors, but
  not all
Randomized Block Design Example

 • Engineers at semiconductor manufacturing facility
   want to test effect of 4 different wafer implant
   material dosages using 3 runs for each level
 • The nuisance factor is "furnace run", since it is
   known that each furnace run differs from the last
   and impacts many process parameters
    – Block: run all 4x3=12 wafers in the same furnace run
    – Completely Random: randomly select order and and
      include each run each on a different furnace run
                 Full Factorial
• Run all combinations of factors and levels.
  – For example: A basic experimental design is
    one with all input factors set at two levels each
     • These levels are called ‘high’ and ‘low’ or ‘+1’
       and‘-1’ respectively.
     • A design with all possible high/low combinations of
       all the input factors is called a “full factorial design
       in two levels”
         Fractional Factorial
• A factorial experiment in which only an
  adequately chosen fraction of the treatment
  combinations required for the complete
  factorial experiment is selected to be run
           What should we do?
• For our competitive assessment, we have
  one factor – portable water filters
• The levels are the different filters
• What are our nuisance factors?
  –   Properties of pond water?
  –   Day of experiment?
  –   Operators of experiment?
  –   ???
      Competitive Assessment
          Experiment
• We will run a randomized block design
• However, we may only be able to do one
  run per level
• We’ll try to block as many of the nuisance
  factors as possible
• There are some that we won’t be able to
  block or randomize

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:6
posted:7/15/2012
language:English
pages:17