45 research outputs found
Statistical methods for quality control in additive manufacturing
No full textAdditive manufacturing (AM), or three-dimensional (3D) printing, holds the promise of direct fabricating products of highly complex and individualized geometric shapes. However, shape deviation suffered by fabricated products remains one of the most concerned quality issues that hinders the wider application of AM technologies. Quality control for AM involves improving the shape accuracy for any new and untried products. The objective of the thesis is to develop statistical methods for shape deviation prediction and derive compensation strategies for shape accuracy improvement. In contrast to mass production, due to the huge variety of product shapes and low volume of production in AM processes, it is usually cost-prohibitive to collect sufficient sample data, which means only limited sample data for limited shapes are available. Achieving high and consistent shape accuracy with limited sample data in such one-of-a-kind AM processes is a challenging task, since the shape deviation of fabricated products usually depends on the setting of certain process parameters and varies from shape to shape. To address this, three research issues are investigated and corresponding methods are developed for shape accuracy control. First, a two-step Gaussian Process and Kernel Smoothing (GPKS) scheme is proposed to predict the in-plane (x-y plane) shape deviations with the information on process parameters. Based on this prediction scheme, a shape compensation strategy is derived that greatly improves the shape accuracy of products under different settings of process parameters. Second, a novel statistical transfer learning framework is proposed to predict and compensate the in-plane shape deviations for freeform products with arbitrary untried shapes based on a small number of fabricated products. In this framework, transferring information from source shapes to new target shapes is achieved through establishing shape deviation models based on error decomposition and learning of a common representation for local shape features. Third, the statistical transfer learning framework for in-plane shape accuracy control is extended to 3D shape accuracy control by learning the additional error induced due to different layer features. Experimental studies of the fused filament fabrication processes validated the effectiveness of proposed methods.</p
Erstellung eines Konzeptes für ein automatisch arbeitendes Fertigungssystem zur Schleifbearbeitung mit integrierter Qualitätskontrolle für Werkstücke und Schleifkörper
No full textZunehmend gewinnt die Automatisierung in mannigfaltiger Form in der
Produktionstechnik an Bedeutung, so dass es bei der Auslegung der einzusetzenden
Schleifverfahren und -anlagen eine Vielzahl an technischen und organisatorischen
Rahmenbedingungen zu beachten gilt.
In Rahmen der Bachelorarbeit sind, gemäß der o.g. Themenstellung, das
Qualitätssicherungskonzept zu erstellen, deren Struktur darzustellen sowie
dazugehörige maschinenbautechnische und organisatorische Besonderheiten
aufzuzeigen und deren Einbindung in den Prozessablauf zu analysiere
Statistical transfer learning: A review and some extensions to statistical process control
No full textThe rapid development of information technology, together with advances in sensory and data acquisition techniques, has led to the increasing necessity of handling datasets from multiple domains. In recent years, transfer learning has emerged as an effective framework for tackling related tasks in target domains by transferring previously-acquired knowledge from source domains. Statistical models and methodologies are widely involved in transfer learning and play a critical role, which, however, has not been emphasized in most surveys of transfer learning. In this article, we conduct a comprehensive literature review on statistical transfer learning, i.e., transfer learning techniques with a focus on statistical models and statistical methodologies, demonstrating how statistics can be used in transfer learning. In addition, we highlight opportunities for the use of statistical transfer learning to improve statistical process control and quality control. Several potential future issues in statistical transfer learning are discussed.</p
A prediction and compensation scheme for in-plane shape deviation of additive manufacturing with information on process parameters
No full textShape fidelity is a critical issue that hinders the wider application of Additive Manufacturing (AM) technologies. In many AM processes, the shape of a product is usually different from its input design and the deviation usually depends on certain process parameters. In this article, we aim to improve the shape fidelity of AM products through compensation, with the information on these parameters. To achieve this, a two-step hierarchical scheme is proposed to predict the in-plane deviation of the product shape, which relates to the process parameters and the two-dimensional input shape. Based on this prediction procedure, a shape compensation strategy is developed that greatly improves the dimensional accuracy of products. Experimental studies of fused deposition modeling processes validate the effectiveness of our proposed scheme in terms of both predicting the shape deviation and improving the shape accuracy.</p
A Statistical Transfer Learning Perspective for Modeling Shape Deviations in Additive Manufacturing
No full textQuality control of additive manufacturing applications is required to improve the shape fidelity of the products, which relies on increasing the predictive performance of statistical deviation models for any new shape. Building a single comprehensive model for a wide range of shapes is a very challenging problem, since the error generating mechanism of additive manufacturing applications is usually of high complexity, the amount of training data is usually limited, and the connection among different shapes is unknown. In this study, a novel shape deviation modeling scheme is proposed. In this scheme, the dimensional error of the product is modeled in a parameter-based transfer learning approach. In particular, the shape deviation is decomposed into two components: The shape-independent error and the shape-specific error. The shape-independent error is described by a statistical model that incorporates the engineering knowledge. Guidelines to investigate modeling of the shape-specific error are also given.</p
A hybrid transfer learning framework for in-plane freeform shape accuracy control in additive manufacturing
No full textShape accuracy control is one of the quality issues of greatest concern in Additive Manufacturing (AM). An efficient approach to improving the shape accuracy of a fabricated product is to compensate the fabrication errors of AM systems by modifying the input shape defined by a digital design model. In contrast with mass production, AM processes typically fabricate customized products with extremely low volume and huge shape varieties, which makes shape accuracy control in AM a challenging problem. In this article, we propose a hybrid transfer learning framework to predict and compensate the in-plane shape deviations of new and untried freeform products based on a small number of previously fabricated products. Within this framework, the shape deviation is decomposed into a shape-independent error and a shape-specific error. A parameter-based transfer learning approach is used to facilitate a sharing of parameters for modeling the shape-independent error, whereas a feature-based transfer learning approach is taken to promote the learning of a common representation of local shape features for modeling the shape-specific error. Experimental studies of a fused filament fabrication process demonstrate the effectiveness of our proposed framework in predicting the shape deviation and improving the shape accuracy of new products with freeform shapes.</p
Photo-SLAM: Real-time Simultaneous Localization and Photorealistic Mapping for Monocular, Stereo, and RGB-D Cameras
No full textThe integration of neural rendering and the SLAM system recently showed promising results in joint localization and photorealistic view reconstruction. However, existing methods, fully relying on implicit representations, are so resource-hungry that they cannot run on portable devices, which deviates from the original intention of SLAM. In this paper, we present Photo-SLAM, a novel SLAM framework with a hyper primitives map. Specifically, we simultaneously exploit explicit geometric features for localization and learn implicit photometric features to represent the texture information of the observed environment. In addition to actively densifying hyper primitives based on geometric features, we further introduce a Gaussian-Pyramid-based training method to progressively learn multi-level features, enhancing photorealistic mapping performance. The extensive experiments with monocular, stereo, and RGB-D datasets prove that our proposed system Photo-SLAM sig-nificantly outperforms current state-of-the-art SLAM systems for online photorealistic mapping, e.g., PSNR is 30% higher and rendering speed is hundreds of times faster in the Replica dataset. Moreover, the Photo-SLAM can run at real-time speed using an embedded platform such as Jet-son AGX Orin, showing the potential of robotics applications. Project Page and code: https://huajianup.github.io/research/Photo-SLAM/
A hybrid transfer learning framework for in-plane freeform shape accuracy control in additive manufacturing
No full textShape accuracy control is one of the quality issues of greatest concern in Additive Manufacturing (AM). An efficient approach to improving the shape accuracy of a fabricated product is to compensate the fabrication errors of AM systems by modifying the input shape defined by a digital design model. In contrast with mass production, AM processes typically fabricate customized products with extremely low volume and huge shape varieties, which makes shape accuracy control in AM a challenging problem. In this article, we propose a hybrid transfer learning framework to predict and compensate the in-plane shape deviations of new and untried freeform products based on a small number of previously fabricated products. Within this framework, the shape deviation is decomposed into a shape-independent error and a shape-specific error. A parameter-based transfer learning approach is used to facilitate a sharing of parameters for modeling the shape-independent error, whereas a feature-based transfer learning approach is taken to promote the learning of a common representation of local shape features for modeling the shape-specific error. Experimental studies of a fused filament fabrication process demonstrate the effectiveness of our proposed framework in predicting the shape deviation and improving the shape accuracy of new products with freeform shapes.</p
