Microstructural Characterization of QC-Forming Al-Mn-Based Alloy Using Machine Learning Software

Abstract The main objective was to investigate and evaluate the influence of TiC and TiB 2 inoculants on the formation of not only the icosahedral quasicrystalline phase (IQC) but also the β -phase in our Al-Mn-Si-Cu-Mg alloy. First, the presence of both phases was confirmed using electron backscatter diffraction (EBSD), followed by microstructural segmentation and quantification using the open-source machine learning software ilastik and Fiji. The ilastik software was selected because it allowed us to use different parameters to distinguish between the IQC and β -AlMnSi phases, which otherwise have similar color/Z contrast and are difficult to distinguish in a timely manner using other methods. The analyses were performed on a total of 3662 images taken during optical light microscopy. The results show that TiC inoculants better promote the ability to form IQC compared to TiB 2 . The use of TiC resulted in an increase of 40% compared to only 14% when TiB 2 was used. Exceeding the TiC threshold of 0.0224 wt.% resulted in a 571% increase in the amount of β -phase compared to our non-inoculated alloy. Microhardness measurements were carried out on the IQC phase using the Vickers method, and an average value of 680 HV0.01 was obtained.