Process Improvement Using Data

This book is free to read here, or to download as a PDF. It covers the methods engineers and scientists use with process data: visualization, univariate statistics, process monitoring, least squares models, designed experiments, and latent variable models such as PCA and PLS.

The Python examples in chapters 1 to 6 each have a Run in browser button, as under the raincloud plot in box plots. A click starts Python in your browser tab, runs the example on its data set and draws the result under the code, with nothing to install. The examples use process-improve, the book’s open-source Python package.

Table of Contents

• 1. Visualizing Process Data
  • 1.1. Data visualization in context
  • 1.2. References and readings
  • 1.3. Time-series plots
  • 1.4. Bar plots
  • 1.5. Box plots
  • 1.6. Relational graphs: scatter plots
  • 1.7. Tables as a form of data visualization
  • 1.8. Topics of aesthetics and style
  • 1.9. General summary: revealing complex data graphically
  • 1.10. Exercises
• 2. Univariate Data Analysis
  • 2.1. Univariate data analysis in context
  • 2.2. References and readings
  • 2.3. What is variability?
  • 2.4. Histograms and probability distributions
  • 2.5. Some terminology
  • 2.6. Binary (Bernoulli) distribution
  • 2.7. Uniform distribution
  • 2.8. The normal distribution and checking for normality
  • 2.9. The t-distribution
  • 2.10. Poisson distribution
  • 2.11. Confidence intervals
  • 2.12. Testing for differences and similarity
  • 2.13. Paired tests
  • 2.14. Other types of confidence intervals
  • 2.15. Statistical tables for the normal- and t-distribution
  • 2.16. Exercises
• 3. Process Monitoring
  • 3.1. Process monitoring in context
  • 3.2. References and readings
  • 3.3. What are process monitoring charts?
  • 3.4. Shewhart charts
  • 3.5. CUSUM charts
  • 3.6. EWMA charts
  • 3.7. Other types of monitoring charts
  • 3.8. Process capability
  • 3.9. The industrial practice of process monitoring
  • 3.10. Industrial case study
  • 3.11. Summary
  • 3.12. Exercises
• 4. Least Squares Modelling Review
  • 4.1. Least squares modelling in context
  • 4.2. References and readings
  • 4.3. Covariance
  • 4.4. Correlation
  • 4.5. Some definitions
  • 4.6. Least squares models with a single x-variable
  • 4.7. Least squares model analysis
  • 4.8. Investigating an existing linear model
  • 4.9. Summary of steps to build and investigate a linear model
  • 4.10. More than one variable: multiple linear regression (MLR)
  • 4.11. Outliers: discrepancy, leverage, and influence of the observations
  • 4.12. Enrichment topics
  • 4.13. Exercises
• 5. Design and Analysis of Experiments
  • 5.1. Design and analysis of experiments in context
  • 5.2. Terminology
  • 5.3. Usage examples
  • 5.4. References and readings
  • 5.5. Why learning about systems is important
  • 5.6. Experiments with a single variable at two levels
  • 5.7. Changing one single variable at a time (COST)
  • 5.8. Full factorial designs
  • 5.9. Fractional factorial designs
  • 5.10. Blocking and confounding for disturbances
  • 5.11. Response surface methods
  • 5.12. Evolutionary operation
  • 5.13. Multi-response optimization: the sweet spot and desirability
  • 5.14. Optimal designs of experiments: D-, A-, and G-optimality and the information matrix
  • 5.15. Definitive screening designs (DSD): screen factors and detect curvature in one experiment
  • 5.16. OMARS designs: orthogonal minimally aliased response surface designs
  • 5.17. A worked OMARS study: four factors on a fed-batch bioreactor
  • 5.18. Judging and comparing experimental designs
  • 5.19. An omnibus comparison across design families
  • 5.20. A worked screen-then-optimise study: six factors on a reactor with a heat limit
  • 5.21. Mixture experiments
  • 5.22. General approach for experimentation
  • 5.23. Extended topics related to designed experiments
  • 5.24. Exercises
• 6. Latent Variable Modelling
  • 6.1. In context
  • 6.2. References and readings
  • 6.3. Extracting value from data
  • 6.4. What is a latent variable?
  • 6.5. Principal Component Analysis (PCA)
  • 6.6. Principal Component Regression (PCR)
  • 6.7. Introduction to Projection to Latent Structures (PLS)
• 7. Applications of Process Improvement using Data
  • 7.1. Improved process understanding
  • 7.2. Troubleshooting process problems
  • 7.3. Multivariate process monitoring case studies
  • 7.4. Soft sensors and inferential sensors
  • 7.5. Keeping a model current: an adaptive soft sensor
  • 7.6. Multivariate image analysis
  • 7.7. Batch process monitoring and improvement
  • 7.8. Learning from batch trajectories: the DuPont polymerization reactor
  • 7.9. Diagnosing a known fault with batch PLS: the SBR reactor
  • 7.10. Combining initial conditions and trajectories: multiblock batch PLS on a batch dryer
  • 7.11. Product development and product improvement
  • 7.12. A mixed-level split-plot design with a profile response