Classification and selection of sheet forming processes with machine learning

ABSTRACT

Sheet metal forming is a critical component of modern manufacturing. The procedure for selecting a suitable manufacturing process to achieve the final geometry of a metal part is unstructured and heavily reliant on human expertise. Similarly, classification and design of new metal forming processes has yet to be automated. In this study, a machine learning approach was used for the first time to identify the manufacturing process that formed a part solely from the final geometry. Several neural network configurations were tested with different geometry representation methods. The best performing classifier employed a deep convolutional neural network and achieved an accuracy of 89%, namely when the geometry was given through a mapping of the mean and Gaussian curvatures. The high accuracy rate establishes that automated methods can perform this step between design and manufacture, thus eliminating the need for human experts in matching each product to a suitable forming method.

KEYWORDS:

• Process selection
• machine learning
• convolutional neural networks
• sheet metal forming

Disclosure statement

No potential conflict of interest was reported by the authors.