Hard coatings are primarily designed to offer excellent wear and corrosion resistance. However, these properties are often adjusted at the expense of the damage tolerance of the applied materials. By introducing the concept of high entropy nitrides (HEN) new property combinations are expected. In this contribution, the damage tolerance of different HEN coatings as well as industrially established coatings will be evaluated by tensile testing of coated compliant substrates.

Evaluating the damaging behavior of hard nitride coatings by loading a compliant substrate and analyzing crack formation is a suitable way to quickly screen different coating materials for high damage tolerance. However, manually evaluating the crack density at different applied strains by using the line counting method is very time-consuming and prone to human error. In this contribution, we present an automated approach to analyze not only the crack density but also further microstructural features (such as the number of macroparticles and local spallations) of multiple nitride coatings deposited on compliant steel substrates. This is done by semantic segmentation utilizing a DeeplabV3+ convolutional neural network. Six different nitride coatings with varying hardness, residual stress and fracture toughness were analyzed. Augmentation methods to deal with high image resolutions and underlying class imbalances are presented and the limits of the methods are shown. A local crack density is introduced to further evaluate differences in the damaging behavior. Based on the findings a more meaningful comparison of the different coatings could be derived.