Vision based in-process inspection for countersink in automated drilling and riveting

Abstract

Countersinks have been widely used for flush rivets of aircraft panels. The quality of countersinking significantly impacts the performance of riveted joints. Thus, a reliable countersink inspection method is typically required in automated aircraft assembly. In current practice, the in-process inspection of countersinks, especially the concurrent inspection of their normal deviation and depth, has not been reported. This paper presents an in-process countersink inspection approach based on machine vision in automated drilling and riveting systems. In this study, the inspection system is firstly developed. With a telecentric lens, the system generates an orthographic view of the subject being observed to avoid the scaling effect. Moreover, the system is mechanically coupled with the drill unit on the shuttle; consequently, the countersink image is obtained accurately and rapidly. Thereafter, an improved edge-following method is proposed to extract the countersink contour. As the contour is the combination of a circle and an ellipse in a countersink image, the RANSAC algorithm is employed to fit them together. Finally, by orthographic projection, the countersink imaging process model is established; furthermore, the normal deviation and depth of countersink are derived. The results of a series of experiments demonstrate that the proposed method performs accurately and robustly.