Neural network application to the reconfigurable multipoint forming process

Abstract

Multipoint forming of thin sheets plates is based on the discrete die-punch reconfigurable tooling concept. The paper is concerned with the application of the neural network method in studying the springback phenomenon in multipoint forming. The method of neural network is first presented. An algorithm based on FEM and neural network modeling is then presented. Using the FEM simulation, the springback values for a simply curved geometry are obtained. On this basis, a neural network is trained, using as input parameters the rubber thickness, the rubber elastic modulus and the pins stroke and as output the springback in width and height defined. The conclusions obtained from the neural network modeling certify the validity of the developed method.