Additive manufacturing (AM) methods like powder bed fusion-laser beam (PBF-LB) enable complex geometry production. However, understanding and predicting the microstructural properties of AM parts remain challenging due to the inherent non-homogeneity introduced during the manufacturing process. This study demonstrates a novel approach for 3D microstructure representation and virtual testing of non-homogeneous AM materials using 2d electron backscatter diffraction (EBSD) data. By employing the representative volume element (RVE) method, we reconstruct the 3D microstructure from 2D EBSD datasets, effectively capturing the grain morphological characteristics of PBF-LB-produced Hastelloy X. Using validated RVE data, we artificially generate combinations of two grain textures to gain deeper insight into locally affected areas, particularly the stress distribution within the interfaces, as well as global material behavior, exploring non-homogeneity. Computational homogenization (CH) utilizing a crystal elasticity finite element (CEFE) method is used to virtually test and predict directional elastic properties, offering insights into relationships between microstructure evolution and property correlation. The experimentally validated results show a strong correlation, with only 0.5-3.5% correlation error for the selected grain tessellation method. This consistency and reliability of the methodology provide high confidence for additional virtual tests predicting the properties of non-homogeneous, artificially generated combined-grain structures.