TTK31 Design of Experiments (DoE), metamodeling and Quality by Design (QbD)

Instructors: Øivind Riis and Frank Westad

9th semester module – 3.75 study points

Understand the theory and applications of

• Design of Experiments (DoE) in research and development (R&D)
• Quality by Design (QbD) and Process Analytical Technology (PAT)

Be able to use DoE in experimental work and in simulation systems.

None

Design of experiments (DoE) is an important tool to establish knowledge on how to optimize processes of any kind. DoE techniques are necessary for gaining knowledge and understanding the causality in a system by performing as few experiments as possible. One of the strengths of DoE is the ability to detect interactions between parameters in complex systems.

DoE is also an essential technique for tuning parameters in simulation systems and semi-physical/first-principle models (metamodeling).

Industrial applications of DoE focus on keeping the processing cost of raw materials as low as possible while at the same time consistently produce products with predefined characteristics/quality at minimum cost by optimizing settings of process parameters. Thus, DoE is one of the tools as basis for Model Predictive Control (MPC) and Multivariate Statistical Process Control (MSPC).

Related to DoE are the concepts Quality by Design (QbD) and Process Analytical Technology (PAT), which are generic approaches for improving manufacturing efficiency and quality. QbD stresses designing quality into manufacturing rather than testing the quality of finished products. PAT enables real-time, quality-based adjustments to your process.

QbD – PAT – DoE may also be utilized in Precision Production of Health services (PPH) e.g. in hospitals.

The course will cover the following topics:

• DoE as an important tool for R&D
• How to apply DoE in simulation systems
• QbD – PAT for optimizing the production of goods/materials in industry and the production of health services in healthcare.

It will focus mainly on theory and practice of DoE.

1. DoE: Introduction and motivation
2. ANalysis Of VAriance (ANOVA)
3. Factorial designs
4. Fractional factorial designs
5. Response surface designs
6. Optimal designs
7. Metamodeling
8. Combining DoE with multivariate analysis/machine learning
9. QbD – PAT
10. Practical examples of DoE related to cybernetics

Actual use cases demonstrated by the use of software

1. Assignment use case 1 (paper helicopters)
2. Assignment use case 2 (metamodeling)

1. Multivariate Data Analysis; Esbensen, Swarbrick, Westad
2. Handbook for experimenters; StatEase
3. QdB & PAT for dummies