Stereovision measurement of layer geometry in wire and arc additive manufacturing with various stereo matching algorithms

Abstract

Robust measurement of layer geometry can help better understand the complex deposition process and provide feedback control to increase process stability of wire and arc additive manufacturing (WAAM). In this study, a virtual binocular vision sensing system is designed to measure the layer width and torch height from top layer simultaneously. Considering that stereo matching is the most crucial step for 3-D reconstruction in stereovision sensing, various matching algorithms, i.e. adaptive Census transform (ACT), global-based crosswise iteration (GCI), and semi global matching (SGM), are employed to calculate disparity images of image pairs. The matching algorithms are tested based on the standard datasets, indicating that the highest matching accuracy comes from the GCI matching algorithm. Then, a standard cylinder is taken as an example to verify the effectiveness of the sensing system and algorithms. The results show that the cylinder width errors with the ACT, GCI and SGM matching algorithms are 3.17%, 3.02%, and 3.08%, respectively, while the cylinder height errors are 5.83%, 4.15%, and 4.83%, respectively. Finally, layer geometries in WAAM with various process parameters are determined. The width and height errors of layer geometry with the GCI matching algorithm are less than 3.2%. This study will lay a solid foundation for subsequent feedback control for layer geometry in WAAM.