Visual Analysis of Complex Simulation Data using Multiple Heterogenous Views
Helmut Doleisch · reposiTUm (TU Wien) · 2004
Computational Fluid Dynamics (CFD) simulation has become very popular and is used in a wide variety of applications. Applications range from the automotive industry to aerodynamics to environmental and weather simulation, and many more. CFD simulation is popular for several reasons, including that many phenomena can be studied more easily through simulation. Measuring approaches might influence and change flow behavior. Computational simulation speeds up the design and development process of many products. Typically, CFD simulation results in very large data sets. Results are also usually time-dependent and multi-variate, including many attributes for each simulated point in space and time, e.g., flow vectors, pressure, temperature, mass fraction values of chemical substances, etc. Analyzing such data sets is not an easy task for the engineers, who have to investigate and evaluate the results. Visualization can be used to support the exploration and analysis of these data sets. Most current visualization methods for data from 3D flow simulation focus either on displaying geometric objects (e.g., streamlines, isosurfaces, etc.), or on feature-based methods employing special feature extraction and tracking techniques. However, these approaches usually do not allow the user to easily and interactivly investigate the multi-dimensional interrelations between different data attributes. The feature extraction process is usually done in a (semi-)automatic way, not allowing for interactive changes of the feature specification. The central theme of this thesis is to provide a flexible framework for interactive visual analysis of large, multi-dimensional, and time-dependent data sets resulting from flow simulation. In other words, the focus of this work is to develop a framework, which combines multiple, rather well-known concepts from scientific and information visualization, to build a new feature-based visualization framework which is based on user-driven visual analysis. This framework is called SimVis. The major strength of the newly presented visualization approach lies in a balanced combination of several different innovations. These by themselves are not all completely new and some may (to a certain extend) also be found as isolated solutions in other approaches (or in other combinations). Nevertheless, in the combinations proposed here, each component builds an integral part of the framework, which combines different individual solutions to attain maximal flexibility, while still providing solid and stable analysis tools. The innovations that contribute to this interactive feature specification framework include (1) the combination of views and methods from scientific visualization and information visualization, (2) a sophisticated interaction scheme allowing for fast and flexible information drill-down by means of advanced brushing mechanisms, (3) a fuzzy notion of feature specification and composite specifications, (4) enabling focus+context visualization (especially in the spatial domain of 3D rendering), (5) providing proper access to the special data dimension of time, and (6) coping with interactive visualization of relatively large data sets on standard PCs. Also, with the help of integrating attribute derivation (a mechanism for interactive calculation of derived data attributes) and advanced brushing mechanisms, the specification of time-dependent features, i.e., features inherently depending on the special data dimension of time, is realized. Finally two case studies are presented that demonstrate that the framework presented here is indeed generally applicable (e.g., to the automotive industry, aerodynamics, molding, climate simulations, etc.), and how it compares to other solutions and how it adds additional information and value to current methods.