INTRODUCTION FOR DESIGN OF EXPERIMENT (DOE)
A design of experiment (DOE) is a structured method for determining the relationship between factors affecting a process and the output of that process. Here, you will focus heavily on the scientific and statistical methods involved in planning and analyzing DOEs in order to yield practical results. Some understanding of basic statistical methods will be needed to complete this course.
COURSE OBJECTIVES
- Decide whether to run a DOE to solve a problem or optimize a system
- Set-Up a Full Factorial DOE Test Matrix, in both Randomized and Blocked forms
- Analyze and Interpret Full Factorial DOE Results using ANOVA, (when relevant) Regression, and Graphical methods
- Set-Up a Fractional (Partial) Factorial DOE, using the Confounding Principle
- Analyze and Interpret the results of a Fractional Factorial DOE
- Recognize the main principles and benefits of Robust Design DOE
- Decide when a Response Surface DOE should be run
- Select the appropriate Response Surface Design (either Plackett-Burman, Box-Behnken, Central Composite, or D-Optimal)
- Interpret Response Surface Outputs
- Utilize the MiniTabTM Software tool to analyze data
COURSE CONTENT
Icebreaker: Team Problem Solving Exercise Using Engineering Judgment
What is DOE?
- Types of Designed Experiments
- Application Examples
- Where DOE Fits in with Other Tools/Methods?
DOE Requirements: Before You Can Run an Experiment
- Writing Problem and Objective Statements
- Ensuring DOE is the Correct Tool
- Selecting Response Variable(s) and Experimental Factors
- Actual vs. Surrogate Responses
- Attention to Experiment Logistics
- Test Set-up and Data Collection Planning
- Selecting and Evaluating a Gage
Full Factorial Experiments
- Introduction to Cube Plots for 3- or 4-factor 2-level Experiments
- Experiment Set-Up
- Factor Levels, Repetitions, and “Right-Sizing” the Experiment
- Experiment Terms to Estimate (Main Effects and Interactions)
- High-Level Significance Evaluation
DOE Statistical Analysis
- ANOVA Principles for Simple Full Factorial Experiments — Statistics Basics; Significance Test Methods; Effect of Non-Random Experiments; Estimating Significance Test “Power”; Confidence Intervals; Estimating Random Error
- Analysis Plots — Normal and Half-Normal Plots; Main Effect and Interaction Plots
- Regression Analysis of Simple Full Factorial Experiments
- Using MiniTabTMfor Full Factorial DOE Experiments
Fractional (Partial) Factorial Experiments
- The Confounding Principle — How it Works; What Information We Lose with Confounding (and why we might not care!)
- Selecting and Using Generators (Identities) to Set Up Confounding Strings
- Determining Which Factor Combinations to Run
- Analyzing Fractional Factorial Experiment Data
- Using MiniTabTMfor Fractional Factorial Experiments
Robust Design Experiments (Overview)
- What is Robustness?
- Control and Noise Factors
- Classical and Taguchi Robust DOE Set-Up
- Robustness Metrics
- Analytical and Graphical Output Interpretation
Response Surface Modeling
- What Response Surface Models do BEST
- Available Response Surface DOEs (Plackett-Burman, Box-Behnken, etc.) — Ideal Situation(s) to Use Each Response Surface DOE Type; Cube Plot Set-up of Each Response Surface DOE
- Analyzing Response Surface Experiment Data
- Methods for Finding Optimum Factor Values
- Using MiniTabTMfor response Surface Experiments
Miscellaneous Notes and Wrap-up