Visual Analysis of Defects


In everyday life, we use many objects on which we rely and expect them to work correctly. We use phones to communicate with friends, bicycles to commute, payment cards to buy groceries. However, due to defects, these objects may fail at some time, leading to adverse outcomes. Modern industry continually improves the quality of outputs (e.g., products and services) and ensures that they meet their specifications. A common quality management strategy is the defect analysis used to identify and control outputs that do not conform to their specifications. Traditional defect analysis methods are often manual and, therefore, time-consuming procedures. To build more efficient solutions, defect analysis increasingly employs visual analytics techniques. These techniques automatize and enhance the up-to-now manual analysis steps and support new visual approaches for defect representations that resolve existing defects without introducing new ones. In this dissertation, visual analytics techniques applied to defect analysis are referred to as visual analysis of defects. Being a rapidly developing area, the domain of visual analysis of defects is still missing a formalized basis.

This dissertation presents and discusses a workflow for the visual analysis of defects based on the plan-do-check-act cycle of continual improvement. The workflow consists of four steps: defect prevention, control of defective outputs, performance evaluation, and improvement. During the defect prevention step, domain experts plan the design and development processes to ensure that intended results can be achieved while forecasting risks and opportunities. During the control of defective outputs step, domain experts implement the processes and control defects arising throughout these processes. During the performance evaluation step, domain experts ensure that defective outputs are identified by measuring the object’s characteristics. During the improvement step, domain experts explore possible actions that improve the object quality.

This dissertation presents four solutions that advance the visual analysis of defects at the four distinct steps of the workflow. The first solution corresponds to the defect prevention step and provides a preview of dental treatment. It helps dental technicians to identify the most suitable treatment option and avoid cases when patients are unsatisfied with the results due to poor denture aesthetics. The second solution corresponds to the control of defective outputs step and supports dental technicians in designing aesthetic and functional dentures. The approach provides immediate visual feedback on a change in the denture design, which helps to evaluate how the change affects aesthetics. The third solution corresponds to the performance evaluation step and supports material engineers in investigating the damage mechanism in composite materials. First, the system captures and measures various defects such as matrix fracture, fiber/matrix debonding, fiber pull-out, and fiber fracture. Later, users analyze these defects using several interactive visualization techniques. The fourth solution corresponds to the improvement step and visualizes 4D dynamical systems describing various phenomena. The solution enables the 4D representation of dynamical systems and allows the 4D representation to seamlessly transition into, familiar to the user, lower-dimensional plots.


The four steps of visual analysis of defects and the solutions proposed within the scope of this dissertation to address them.

Paper A: Aleksandr Amirkhanov, Artem Amirkhanov, Matthias Bernhard, Zsolt Toth, Sabine Stiller, Andreas Geier, M. Eduard Gröller, Gabriel Mistelbauer, “WithTeeth: denture preview in augmented reality,” in Vision, Modeling and Visualization, The Eurographics Association, 2018, DOI: 10.2312/vmv.20181250.

This paper supports the defect prevention (i.e., the plan phase of the PDCA cycle) in the context of an augmented reality approach for previewing different dental treatment options. The scientific contributions of this paper include:

  • an augmented reality approach for previewing different dental treatment options;
  • a technique that achieves compelling visualizations without depending on complex reconstruction algorithms, expensive hardware, or tedious calibration procedures;
  • a real-time system that permits user’s free head movements and allows switching different teeth replacement options interactively.

Paper B: Aleksandr Amirkhanov, Matthias Bernhard, Alexey Karimov, Sabine Stiller, Andreas Geier, M. Eduard Gröller, Gabriel Mistelbauer, “Visual analytics in dental aesthetics,” Computer Graphics Forum, vol. 39, no. 7, pp. 635–646, 2020. DOI: 10.1111/cgf.14174.

This paper supports the control of defective outputs (i.e., the do phase of the PDCA cycle) in the context of an approach helping dental technicians to adjust a denture that fits a patient’s appearance. The scientific contributions of this paper include:

  • an approach that assists dental technicians in identifying the most aesthetically fitting denture and subsequently adjusting it;
  • a technique that predicts the appearance of a patient with in-line designed dentures;
  • a novel approach for quantifying several aspects of dental aesthetics;
  • a technique for visualizing dental aesthetic aspects interactively during denture design;
  • an approach that is integrated into the state-of-the-art digital workflow of dental technicians.

Paper C: Aleksandr Amirkhanov, Artem Amirkhanov, Dietmar Salaberger, Johann Kastner, M. Eduard Gröller, Christoph Heinzl, “Visual analysis of defects in glass fiber reinforced polymers for 4DCT interrupted in situ tests,” Computer Graphics Forum, vol. 35, no. 3, pp. 201–210, 2016. DOI: 10.1111/cgf.12896.

This paper supports the performance evaluation (i.e., the check phase of the PDCA cycle) in the context of an application for material characterization and conformity verification. The scientific contributions of this paper include:

  • a novel technique for defect extraction and classification in computed tomography images of glass fiber reinforced polypropylene;
  • a set of techniques for visual analysis of material defects.

Paper D: Aleksandr Amirkhanov, Ilona Kosiuk, Peter Szmolyan, Artem Amirkhanov, Gabriel Mistelbauer, M. Eduard Gröller, Renata G. Raidou, “Manylands: A journey across 4D phase space of trajectories,” Computer Graphics Forum, vol. 38, no. 7, pp. 191–202, 2019. DOI: 10.1111/cgf.13828.

This paper supports the improvement (i.e., the act phase of the PDCA cycle) in the context of an approach to analyze and study complex 4D mathematical models describing biological processes. The scientific contributions of this paper include:

  • design and implementation of a technique that supports mathematical domain scientists in understanding and analyzing the behavior of biologically meaningful 4D dynamical system trajectories;
  • a novel technique to discover new knowledge within mathematical systems;
  • a novel technique to illustrate findings of mathematical specialists for externalization and education;
  • an interactive and integrated workflow for exploring and analyzing 4D dynamical systems trajectories.