查看文章

Automated clustering of input nanoindentation data using a K-means algorithm was performed on dual phase (DP) and high strength low alloy (HSLA) steels with the aim of accurate microstructural classification of the phases present. Electron microscopy and diffraction techniques were used to correlate the clustering result with the microstructure, and for the DP steel an accurate separation of polygonal ferrite, martensite and anomalous indents (i.e. located on grain boundaries or adjacent to martensite islands) was achieved. In direct contrast, K-means did not reveal an accurate separation of the HSLA constituent phases. After comparing the hardness level of selected HSLA grains and the minimum distinguishable hardness level of the DP steel, it is found that K-means input variables should differ from each other by at least 10% in order to achieve reasonable separation of clustering results. Further targeted …