Visualization of engine design, simulation and optimization

• Main Author: Wang, Yan Chao
• Other Authors: Lin Feng
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Computer science and engineering::Information systems::Information interfaces and presentation
• Online Access: https://hdl.handle.net/10356/137190
• Institution: Nanyang Technological University

The development of techniques contributes to the digitalization of engine-related procedures, generating a massive amount of engine-related data generated. An effective and efficient understanding of these datasets would provide domain experts with new insights. However, it is impractical for human beings to go through the raw data to have a clear understanding to conduct data analysis, as these datasets have a large-scale size and complex data formats. Hence such data understanding requires closely coupled human and data analysis, where information visualization comes into play. The task on how to utilize information visualization to efficiently understand engine-related datasets can be considered from two aspects. Firstly, visualization can focus on a specific visual analytic target, by integrating heterogeneous datasets and providing meaningful visualization methods. Secondly, visualization can generically provide a tidy and useful view for certain data types. Hence, our work is organized in two aspects: designing specific visualization system for certain analysis target and proposing novel visualization methods for certain data structures in the engine-related dataset, such as high-dimensional data and hierarchical data. For specific visualization system, this thesis presents EngineQV, a single page web-based integrated visualization system that integrates multiple geo-temporal engine-associated datasets, with the aim of assisting domain experts to explore the external factors that affect the aircraft engine performance. The system features a dynamic query on the datasets and incorporates several customized interactive visualizations, which provides intuitive exploration and understanding of the data from various aspects. With the system, a user may query a certain group of engines or compare multiple engine groups, identify an issue, and find its potential causes. The functionality and usability of EngineQV are evaluated by several case studies. The validity of this specific system is confirmed by expert feedback. For generic high-dimensional data visualization, we focus on easing the clumping effect in the two promising plots, RadViz and Star Coordinate. We observe that with the increase of dataset size and dimensionality, the clumping of projected data points towards the origin in RadViz causes low space utilization and degenerates the visibility of the feature characteristics. Hence, to better evaluate the hidden patterns in the centre region, we proposed a new focus+context distortion approach, termed PolarViz, to manipulate the radial distribution of data points in RadViz. PolarViz cannot be directly used in Star Coordinate, as the Star Coordinate does not have a boundary and PolarViz needs the plot boundary to calculate the relative distance. In this case, we proposed a geometric algorithm to determine an axis-aligned bounding box, minimum bounding box, and bounding polygon of the star coordinate with an evenly distributed configuration and an arbitrarily distributed configuration respectively. These boundaries also can provide an overview of the plot and enable a continuous point rendering that adapts to the screen space boundary. For generic hierarchical data visualization, we consider providing a tidy and flexible layout for implicit hierarchy visualization. To achieve this, we propose a novel space partitioning strategy. Starting from a new distance function and neighborhood relationship between sites, we divide the space by axis-aligned segments. Then a sweepline+skyline based heuristic algorithm is proposed to allocate the partitioned spaces to form an orthogonal Voronoi diagram with orthogonal rectangles. To the best of our knowledge, it is the first time to use a sweepline-based strategy for the Voronoi treemap. Moreover, we design a novel strategy to initialize the diagram status and modify the status update procedure so that the generation of our plot is more effective and efficient. We show that the proposed algorithm has an $O(n · log(n))$ complexity which is the same as the state-of-the-art Voronoi treemap. To this end, we show, via experiments on artificial and real-world datasets, the performance of our algorithm in terms of computation time, converge rate, and aspect ratio. Most of our work are implemented in JavaScript and the source code of the proposed PolarViz and the orthogonal Voronoi treemap have been uploaded into the GitHub. The thesis ends with the conclusions of our works and a discussion of future work.