In this project from the Quality Data Analysis course we had to develop a statistical process monitoring algorithm to detect defects in 3D-printed objects, using control charts, the method we learned at lectures. We firstly performed a deep analysis of the images, applying several transformations and operations to precisely isolate each 3D-printed object, and secondly we calculated relevant geometrical variables (such as perimeter, area, and curvature) as well as more specific ones devoted to structural characteristics of the printed objects (e.g. number of voids, area of the voids, etc).
In the end, our algorithm achieved an accuracy of 95%, correctly classifying, as intact or damaged, 38 objects out of the 40 available in the dataset.
Github repository: https://github.com/abylai11/qda-project.
Link diretto al pdf qui. Direct link to pdf here.
