RéSeaux Neuro Neuro-Flous Multicouches Pour La DéTection De Collisions Outils-Pieces En Tournage Simulé

In Computer Aided Design and Manufacturing (CAD/CAM), the technicians have often been confronted with situations of collisions on a manufacturing site, between mobile elements and static elements of the machine. The objective of this study is to develop a neuro-fuzzy technology for the recognition of machining states in CAD/CAM, in particular, the collision states between different profiles, and the treatment after a collision detection. This work is divided in two parts; the first is to design a multiple layers neural-network, which after training evaluates the quality of a machining state at a point of the profile, and therefore the probability of existence of a collision between two profiles. The second part is a design of a fuzzy system that intervenes after the passage of the first tool. If there was a bad quality of machining, the system decides to pass one or several other tools, in order to find the necessary tools for this phase and the corresponding zones, while preventing collisions.