Within recycled aluminum, impurities of mainly Fe cause iron-rich intermetallic phases with undesired effects on the casting process and the mechanical properties. Therefore, melt conditioning of AlSi9Cu3(Fe) is conducted generating large intermetallic particles, which are separated under a subsequent filtration process. The analysis of conditioned and filtrated castings compared to a reference state includes the following main goals: (i) investigation of the morphology of iron-rich intermetallic particles and (ii) behavior of intermetallic phases under fatigue loading. X-ray microtomography (µCT) before and after ultrasonic fatigue testing reveals the initial microstructure and the fatigue crack. Segmentation of materials phases as well as the fatigue crack by machine learning algorithms of Trainable Weka Segmentation allows superposition of the initial microstructure with the fatigue crack. Together with Digital Volume correlation, this reveals the fatigue crack path through the matrix and the intermetallic phases as well as their surrounding local strains.