High-speed Nanoindentation Data Analysis of WC-based Cemented Carbides using Gaussian Mixture Model Clustering and Skew-normal Mixture: Beyond Gaussian Deconvolution

Abstract

The mapping of micromechanical properties of heterogeneous materials with high-speed nanoindentation faces challenges in data interpretation and extraction of reliable phase properties. Gaussian deconvolution has commonly been used to treat this data, but problems arise when microstructures are fine compared to the indentation imprint size and the mechanical properties of the different phases are not extremely different. Recently machine-learning methods, such as clustering techniques, have emerged as good approaches to assess these challenges.

Within this context, it is important to understand the micromechanical properties of cemented carbides, materials usually referred to as hardmetals, but their complex microstructure poses challenges for assessment and interpretation. The study reveals insights from high-speed nanoindentation data treatment on hardmetals with different microstructures and binder compostion. The data has been statistically analyzed by means of a clustering method: Gaussian Mixture Model (GMM) and the fitting of a mixture of skew-normal distributions. Findings underscore the asymmetry in phase properties, and the challenges GMM encounters in some samples. The skew-normal method offers enhanced precision and addresses issues related to scatter in phase intersections, providing a more accurate representation of fine microstructural features. The combined approach of GMM and skew-normal proves consistent for reliable evaluation of micromechanical properties from nanoindentation maps in cemented carbides, and demonstrate the potential of this technique to be applied to novel hardmetal compositions as well as other composites.