me 388 – applied instrumentation laboratory design of experiments (doe)
DESCRIPTION
ME 388 – Applied Instrumentation Laboratory Design of Experiments (DOE). Reference (If you can find it). R.D Moen, T.W. Nolan, L.P. Provost, Improving Quality Through Planned Experimentation , McGraw-Hill, 1991 - PowerPoint PPT PresentationTRANSCRIPT
ME 388 – Applied Instrumentation Laboratory
Design of Experiments (DOE)
Reference (If you can find it)
• R.D Moen, T.W. Nolan, L.P. Provost, Improving Quality Through Planned Experimentation, McGraw-Hill, 1991
• D.C. Montgomery, Design and Analysis of Experiments, 5th Edition, Wiley, 2001
Six Sigma (black belt)• System for constant improvement
• Use of statistical tools for process analysis, problem solving and improvement
• Six Sigma statement by GE: …The central idea behind Six Sigma is that if you can measure how many "defects" you have in a process, you can systematically figure out how to eliminate them and get as close to "zero defects" as possible…
Definition
• A powerful “statistics-based” experimental methodology that is used to efficiently determine how multiple independent variables affect dependent variables of a system or process.
ProcessIndepententVariables
DependentVariable
A properly executed DOE will…
• Provide the most information
• With the fewest amount of tests
• Compared to a sequential “string-of-pearls” type approach
DOE’s are used for
1. Screening experiments
2. To determine Interactions between independent variables
3. Optimization
Motivation
• Knowledge
• Optimization
• Improvement
Steps to use DOE
• Have some technical knowledge of process or system (assumed here)
• Have some statistics background (assumed here)
• Design experiment (and run)
• Analyze data
Terminology
• Factor = independent variable
• Level = a given value or setting for an independent variable (for example, a 2-level experiment involves testing a high and low value for each independent variable)
2 level, 2 factor design (22)
F2 - F1 +
-
+
2 level, 3 factor design (23)
F2 - F3 +
- F1 + - F1 +
-
+
2 level, 4 factor design (24)
F4 F2 - F3 +
- F1 + - F1 +
- -
+
+ -
+
Fractional Factorial Designs
• Reduce number of test
• At the expense of complete data
• Rely on reasonable judgments and technical knowledge– Assume triple, quadruple and greater
interactions are not significant
27-3 design
27-4 design
Analysis
• Develop extended test/design matrix
• “Effects” are assessed for each variable and combination of variables
• Effects plots can also be generated