Experimental Implementation of Neural Network Springback Control for Sheet Metal Forming

Abstract

The forming of sheet metal into a desired and functional shape is a process, which requires an understanding of materials, mechanics, and manufacturing principles. Furthermore, producing consistent sheet metal components is challenging due to the non-linear interactions of various material and process parameters. One of the major causes for the fabrication of inconsistent sheet metal parts is springback, the elastic strain recovery in the material after the tooling is removed. In this paper, springback of a steel channel forming process is controlled using an artificial neural network and a stepped binder force trajectory. Punch trajectory, which reflects variations in material properties, thickness and friction condition, was used as the key control parameter in the neural network. Consistent springback angles were obtained in experiments using this control scheme.