Génération automatique d'assemblage CAO basée sur des méthodes d'apprentissage machine

Automation in Computer-Aided Design (CAD) is a complex task due to the intricate engineering constraints involved during the design process. This thesis focuses on the automatic generation of mechanical part assemblies. This automatic generation can be used for design assistance, expanding mechanical assembly databases, or reusing CAD models. The proposed method is divided into three parts. The first part involves the creation of a pipeline based on rules, which enables the generation of new mechanical assemblies from existing ones. The second part focuses on the automatic assembly of parts from existing assemblies, based on a machine learning model. A predictive interface model for dumb B-Rep models is developed to address the limitation concerning the origin of the parts in previous works. This final piece of work ultimately allows for the generation of mechanical assemblies from a set of arbitrary B-Rep models. The generated assemblies were compared both qualitatively and quantitatively with those produced by existing methods in the literature. This work is the first to enable the assembly of multiple B-Rep models simultaneously and offers a first approach to addressing the broader challenge of automatic assembly.