Reverse Engineering (RE) is an activity which consists in digitizing a real part in order to create a numerical or virtual model of it. It is conducted on components that does not have any Computer Aided Design (CAD) model, or only a semantically poor 3D representation. Industrial applications for RE are: (A) manufactured parts inspection and dimensional control, (B) CAD model redesign and modifications; and (C) Product Data Management (PDM) system and Database search for 3D models and relative heterogeneous data retrieval. Aeronautical components present several challenges for RE activities, such as complex structures and shapes, large volume of data and a high need of precision in CAD models rebuilt. Moreover, development of local freeform shapes descriptors remains an area of research for applications such as complex surface labeling and retrieval in raw data, and for global mesh semantic segmentation (i.e., decomposition of the mesh into meaningful regions that can be associated with definition features). This paper presents a study on semantic segmentation techniques for complex aeronautical components. Machine Learning and Deep-learning model-based methods for geometry processing will be evaluated on a set of real aircraft engine parts.