Development of a machine learning algorithm for geometric compensation of Single Point Incremental Forming (SPIF) process

Single Point Incremental Forming (SPIF) is a highly versatile sheet metal forming process that enables the production of complex geometries without requiring dedicated tooling. This flexibility has attracted significant interest, particularly in project prototyping and medium-scale industries. In the medical sector, for instance, where it can be used to manufacture anatomically customized prostheses and in the aerospace sector, where it can be used to produce complex lightweight panels. However, the inherent elastic recovery in SPIF processes presents a significant challenge, often resulting in over- or under-formed components. Conventional linear compensation methods fail to achieve satisfactory geometric accuracy, while iterative geometric modulation compensations are resource-intensive and time-consuming, making them less practical for industrial applications. This study outlines the development of a machine learning algorithm designed to generate optimized CAD geometries, minimizing geometric deviations in SPIF-formed components. The algorithm's performance will be validated through the production of fixed and variable-angled cones and pyramids using AW6082-O aluminium sheets. These formed components will be scanned and compared to their theoretical geometries, demonstrating the effectiveness of the proposed approach.