The digital transition in the manufacturing industry is characterised by a three or even four-decade liability. Some CAO models or digital mock-ups accumulated du ring this period are frozen, i.e. 3D models without a construction tree, which are characterised by missing geometries, due to software changes or versions of 3D formats that have not been updated Reverse engineering activities of CAO models, aiming at obtaining semantically rich 3D models, i.e. parametric and modifiable, made up of construction operations, carrying attributes and metadata, with geometric ru les and constraints, etc., thanks to the use of engineering tools such as CATIA for example, or by approaches based on point clouds coming from a scan for example. But, this is still not satisfactory, because at the end of the reverse engineering activities, we often obtain a solid with a weak semantic representation or an absent construction tree. This leads us to propose in the framework of this thesis work, a methodology for managing information linked to CAO models in order to integrate expert information that we call semantic into these CAO models. The frozen CAO models handled are usually in low-level formats such as STL, IGES or STEP AP203. They are used as input data for our methodology and they can be associated with product definition data, such as a product drawing or documents. The processing of CAO models requires a solution that is able to_manage the digital models and the information they couId possibly integrate. And also the incompleteness of some CAO models that is linked to the 3D format or to the limit of the technology used to obtain the CAO model (e.g. software li mit, 3D format for geometric representation only and that does not support a representation of the construction tree or that cannot graphically represent geometric dimensions and tolerances, etc.). Finally, the relevance of integrated information into CAO model, of a non-geometric nature, during the semantic overlay phase should make it possible, in certain cases, to produce parameterised CAO models, specific to the activity of the application domain. The state of the art, concerning the information representation contained in CAO model and the management of this information, makes it possible to identify techniques and approaches that help the semantic enrichment of CAO models at various levels of granularity. This thesis proposes a methodology named Vaquero For CAO Semantic Enrichment (VFCSE), which is made of three step access, identification and annotation. The aim of this methodology is to integrate missing and standardised information of a non-geometric nature, such as product specifications, tolerances, geometric dimensions, etc., into frozen CAO models. This information will be derived from user needs working on the CAO model and will corne from a semantically rich standard in order to be useful for many operations related to the product life cycle. The enrichment, thanks to this semantically rich standard, will allow for a perpetuation of the information and an efficient reuse of CAO model information. ln order to do this, a CAO model is retrieved from a PDM (Product Data Management) thanks to a user request. lt is visualised in a CAO viewer supporting STL, IGES and STEP AP203 formats. Then, follows a step of identifying components of CAO model. These components can be parts or assemblies. The identified components are annotated based on the STEP AP242 format, which represents the semantically rich standard. These annotations are stored in a standardised ontology, which serves as a minimal basis for carrying all the semantics to be integrated into the CAO mode in order to make the CAO model durable and reusable. The scientific contribution of this work is mainly based on the possibility of reverse engineering by using ontologies to annotate 3D models, according to user needs who has the CAO model at his disposal.