Automatic fixtureless inspection of non-rigid parts based on filtering registration points. Issue 1 (October 2016)

• Record Type: Journal Article
• Authors: Sattarpanah Karganroudi, Sasan; Cuillière, Jean-Christophe; Francois, Vincent; Tahan, Souheil-Antoine
• Is Part Of: International journal of advanced manufacturing technology. Volume 87:Issue 1/4(2016)
• Page Start: 687
• Page End: 712
• Publication Date: 2016-10
• Subjects: Geometric inspection -- Non-rigid parts -- GNIF -- Fixtureless inspection -- Principal curvatures -- von Mises stress; Manufacturing processes -- Periodicals; Production engineering -- Periodicals; 670.427
• DOI: 10.1007/s00170-016-8496-5
• Languages: English
• ISSNs: 0268-3768

Abstract

Abstract Computer-aided inspection (CAI) of non-rigid parts significantly contributes to improving performance of products, reducing assembly time and decreasing production costs. CAI methods use scanners to measure point clouds on parts and compare them with the nominal computer-aided design (CAD) model. Due to the compliance of non-rigid parts and for inspection in supplier and client facilities, two sets of sophisticated and expensive dedicated fixtures are usually required to compensate for the deformation of these parts during inspection. CAI methods for fixtureless inspection of non-rigid parts aim at scanning these parts in a free-state for which one of the main challenges is to distinguish between possible geometric deviation (defects) and flexible deformation associated with free-state. In this work, thegeneralized inspection fixture (GNIF ) method is applied to generate a prior set of corresponding sample points between CAD and scanned models. These points are used to deform the CAD model to the scanned model via finite element non-rigid registration. Then, defects are identified by comparing the deformed CAD model with the scanned model. The fact that some sample points can be located close to defects results in an inaccurate estimation of these defects. In this paper, a method is introduced to automatically filter out sample points that are close to defects. This method is based on curvature and von Mises stress. Once filtered, the remaining sample points are used in a new registration, which allows identifying and quantifying defects more accurately. The proposed method is validated on aerospace parts.