École Nationale Supérieure de
Techniques Avancées

Résumé

We present a Python program to detect and analyze crack patterns on cutting tool insert coatings with the goal of determining crack density. The program starts on an SEM photo of the cracked coating. To detect crack lines in the photo, built-in image processing functions from different Python libraries are used. The user must select appropriate parameters, some of which are based on empirical rules. The detection program produces a binary image with single-pixel-width lines, referred to as the “skeleton” of the original image. A custom program then enhances this skeleton, also allowing the user to manually add or remove lines. Crack density is determined using two methods. Next, we study statistical quantities to characterize the crack pattern. Some ideas come from the field of “quantitative stereology”, which studies the characterization of microstructures. Despite efforts, no definitive way of characterizing the crack patterns has been found but we investigate the isotropy/anisotropy of the crack pattern.