Les pièces mécaniques fabriquées tel que les feuilles de métal et les pièces à paroi mince, ont souvent des différences géométriques significatives par rapport à leurs modèles CAD nominaux car ils ont une forme considérablement différente dans une condition d’état libre en raison de la gravité et/ou la tension résiduelle. Par conséquence, les fixtures de conformation coûteux sont traditionnellement utilisés pendant les opérations d’inspection géométriques à la phase de contrôle de qualité. L’objective de cette étude est de développer les méthodes d’inspection géométriques automatisées des pièces flexibles qui ne nécessiteraient pas d’utilisation des fixtures. / In manufacturing, quality control (QC) is an essential phase of a product’s lifecycle management (PLM) as it ensures customers receive parts within permissible tolerance ranges and free from defects. Given that all manufactured parts often have geometrical differences compared to their nominal computer-aided design (CAD) models, performing geometrical inspections becomes critical during the QC phase. Nowadays, actual measurements and defect identification during geometrical inspections have been semi-automated through the use of computer-aided inspection (CAI) software. Such software can simplify the inspection into a data acquisition task (contact-based probing or non-contact scanning of the part) followed by semi-automated procedures in a software environment. Despite their growing popularity and practicality, currently available CAI software assume the input acquired data are from a rigid part. This assumption is a major limitation given that not all manufactured parts are rigid, and in some sectors such as the aeronautical industries a considerable percentage of all manufactured components (35 to 40 percent) possess some nonrigid behavior. In other words, CAI software can only be used when a part maintains its shape in both free-state and state-of-use positions. Free-state shape is that which a part has without inspection fixture support and/or before assembly, whilst state-of-use shape is that which is defined in the nominal CAD. Although free-state and state-of-use positions are the same for rigid parts, some mechanical parts such as sheet metals and skins (thin-wall featured parts) often have significantly large geometric deviations in a free-state position compared to their nominal CAD models due to the effects of gravity and residual stress. Referring to such parts as flexible, the aforementioned deviations force the QC technicians to traditionally use a variety of inspection fixtures and conformation jigs in order to maintain flexible parts in their state-of-use position before using conventional CAI software. Without fixation, the free-state elastic geometric deviation of flexible parts would be mistaken by CAI software as plastic deformations and as a result identified as defects. With fixation, the aforementioned free-state deviations are removed before data acquisition, and whatever deviations remain can be inspected as potential defects. However, multiple disadvantages exist in using fixtures including: time consuming set-up process (e.g. 60+ hours for a skin panel in the aerospace industry), considerable purchase and operating expenses, limitations of standard fixture kits in some scenarios, big errors in CAI analysis if fixation has not been conducted correctly, etc. Such disadvantages have recently led researchers to:1) try to circumvent use of fixtures by digitally deforming (or better called registering) the acquired free-state pointcloud/mesh data of a flexible part until it superimposes onto the part’s corresponding nominal CAD model, thereby elastically deforming the data to obtain an optimal state-of-use shape whilst avoiding neutralization of any existing manufacturing defects2) and to try to introduce dedicated defect identification modules with higher degrees of automation (compared to conventional semi-automated CAI tools)In this thesis the same two goals are pursued. A bi-criterion registration method (and two algorithms/demos based upon it) is proposed to achieve the first goal, thereby enabling defect identification of flexible parts in conventional CAI software without the use of fixtures. This is followed by introducing an automated method for fast approximation of defect amplitudes (and an algorithm/demo based upon it) to achieve the second goal. Validation was conducted against a number of virtual (simulated) and experimental industrial case studies. Obtained satisfactory results reflect the effectiveness and utility of the proposed methods.
| Identifer | oai:union.ndltd.org:theses.fr/2018GREAI023 |
| Date | 15 March 2018 |
| Creators | Babanezhad, Kaveh |
| Contributors | Grenoble Alpes, Bigeon, Jean |
| Source Sets | Dépôt national des thèses électroniques françaises |
| Language | English |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation, Text |
Page generated in 0.0022 seconds