Vision-based deformation and wrinkle detection for semi-finished fiber products on curved surfaces

The paper focuses on a vision-based approach for optimizing automated deformation and draping processes of dry semi-finished fiber products at the production of large-area composite components for the aerospace industry. The vision-based approach developed at University of British Columbia, is to be utilized with the existing draping process, carried out on a form-variable end-effector, developed at the Center for Lightweight Production Technologies (ZLP) in Augsburg. During the deformation of the semi-finished product, tensions develop in the material leading to shearing and relative movements of the fiber material on the gripping surface. In turn, the resulting displacement and deformation of the cut piece negatively influences the production quality. The method proposed in the paper is designed to help in visually detecting and automatically evaluating the drape and deformation of the cut piece on a laboratory scale setup. For this purpose, RGB-D camera data is used to detect the deformed gripper surface and determine the position, the boundary geometry and any wrinkles that may have occurred in the cut piece. The accuracy of the proposed method is verified by experiments on a known target geometry.