Design of Experiments

Course Description

Statistical experimental design, also known as Design of Experiments or DoE, is a core component of QbD (Quality by Design), Robust Design, Lean 6 Sigma, and 6 Sigma quality initiatives.  This course introduces the important elements of DoE including both the design and analysis of experiments. Key components of DoE such as understanding factors of interest and improving process performance will be discussed, and how DoE can be used to minimize the experimental effort to achieve these goals. We will discuss the main phases: Experiment definition; Scoping factors to consider and response variable(s); Screening to reduce to the most important factors; Optimizing these factors, and Response surface analysis. Designed experiments’ results are usually analyzed using ANOVA, which will be explained in detail. The course will cover:

• Definition of the experimental goal including resources available and preliminary feasibility studies
• Choice of experimental design, number of factors and interactions to model, and which factor levels to use
• Describe ANOVA and its use to analyze results
• Different design types (Full factorial, fractional factorial, Box-Behnken,…), replication, blocking
• Modeling of the response surface to find the optimal factor settings

The course will end with a short section on developing designs that work well with multivariate calibration methods, including cross-validatable designs and selection of candidate reference samples based on measured values (reduceNNsamples).  Examples will be used throughout the course to illustrate the various aspect of the DOE modeling process. Examples will be used throughout the course to illustrate the various aspect of the DOE modeling process. The course includes hands-on computer time for participants to work example problems using PLS_Toolbox with MATLAB or Solo.

Prerequisites

None, but some exposure to statistics is very helpful, including hypothesis testing, ANOVA, and regression analysis. Linear Algebra for Chemometricians, MATLAB for Chemometricians or equivalent experience highly recommended.

Course Outline

• Why design experiments
• Experimental definition and scope
• Common experimental designs
• Proper execution of experiments
• Identification of important factors
• ANOVA, viewing factors’ and interactions’ statistical significance
• Relating sources of variation to factors and interactions
• Response Surfaces and optimization designs
• Design for multivariate calibration