Data-driven metal spinning using neural network for obtaining desired dimensions of formed cup

Abstract

A data-driven metal spinning process was developed to systematically generate tool paths to obtain desired dimensions such as height and thickness of formed parts for various sizes of blank disks and tools without preparing big data. An artificial neural network was modeled using tool-path parameters, the sizes of the blank disks and tools, and the height and thickness of the formed parts. The tool-path parameters to obtain the desired dimensions were iteratively calculated by the steepest descent method using a Jacobian submatrix. This method was applied to a multi-pass conventional spinning process, and thickness of spun cups was successfully controlled.