Search

This paper combines adaptive-network-based inference system (ANFIS) and elasto-plastic finite element to predict the ductile fracture initiation and the preform shape of the upsetting process. From the hybrid-learning algorithm in ANFIS, it can efficiently construct rule database and optimal distribution of membership function to solve the punch stroke which causes the ductile fracture, and the preform shape which results a desired cylindrical workpiece after forming in the upsetting process. As a verification of this system, the punch stroke for ductile fracture initiation and the free boundary radius of the billet after forming are compared between ANFIS and FEM simulated results. In the ductile fracture prediction, it is proved that ANFIS can efficiently predict the ductile fracture initiation successfully for arbitrary friction coefficient and aspect ratio. In the preform shape prediction, the simulated cylindrical radius shows good coincidence with the desired radius after forming. From this forward and inverse investigation, the ANFIS is proved to supply a useful optimal soft computing approach in the forming category. [Copyright &y& Elsevier]