Manufacturing processes generate enormous amounts of data, yet traditional statistical methods often focus on univariate relationships already understood, e.g., resin addition percent, weight variation, line speed, moisture variation, etc. Machine learning (ML) provides a complementary approach by allowing the data to reveal previously unknown sources of variation, nonlinear relationships, and interactions.
This hands-on workshop introduces participants to the use of machine learning to identify the variables and operating conditions most strongly associated with variation in product quality and process performance that may have been unknown. Boosted tree algorithms will be emphasized which identify interactions of the process has a hierarchy of correlations.
The workshop emphasizes practical industrial application rather than machine-learning theory. Participants will learn how to develop and evaluate ML models, interpret variable importance and interaction effects, visualize nonlinear relationships, and translate analytical findings into actionable process improvements. Introduction to practical algorithms such as: Decision trees, Random forests, and Gradient-boosted trees will be taught in the Minitab and JMP software platforms with additional demonstrations in running Python code from the platforms.
The workshop concludes with an integrated approach for using ML, statistical methods, and DOE to move from unknown sources of variation → validated root causes → process improvement → measurable ROI.