Investigation of industrial die-cast Al-alloys using X-ray micro-computed tomography and machine learning approach for CT segmentation

The die-casting process creates the opportunity to design and manufacture complex parts in different industrial fields. However, the presence of micro-porosity and cracks can significantly impact the functionality and lifetime of the components. The impregnation technique improves the defect and upgrades the alloy's usability. Hence, identifying the impregnation technique's usefulness in sealing porosity is vital for enhancing the quality of the die casting products. It is hard to detect low atomic number impregnation resin located in the casting defects due to low X-ray attenuation. In this study, microfocus X-ray computed tomography (XCT) with advanced direct conversion detectors could effectively be employed to visualize casting defects in 3D. Also, to recognize the impregnated resin in Al-alloy both qualitatively and quantitatively. Dual-energy XCT recognized the resin material as P601 super sealant quantitatively. Casting defects could be identified in 2D CT images, and it is not easy to detect the impregnated resin with simple intensity-based image processing algorithms. Hence, an approach to improve resin detection through machine learning was studied. With a random forest classifier, trainable weka segmentation was used with three pre-defined classes. It precisely segmented the casting pore, Al-alloy, and resin material after well-trained the known data set.