Interactive Visual Analysis of Anomalies in Simulations of Energy Flow

Fabrizia Bando-Bechtold

Supervisor(s): Kresimir Matkovic

TU Wien


Abstract: At a time when the demand for alternatives to gas and oil is steadily increasing, the automotive industry faces the challenge of designing vehicle systems, which are as efficient as possible. Vehicle system simulation is used for concept analysis, subsystem design, and virtual component integration. The high-level simulation model accurately depicts the energy flow between different vehicle parts over an extended period of time. The the resulting data is presented as a time-dependent edge-weighted graph. The analysis and exploration of such data is a tedious task. We propose a novel approach for exploring and analyzing energy flow data at different levels of detail to assist engineers and guide them to the events of interest. We employ automatic anomaly detection techniques and propose intuitive navigation to time steps of interest. We also propose expandable cards or labels that depict current and overall data. Users can interactively and dynamically choose how much information is displayed on the labels by changing their level of detail. We provide four levels of detail, each giving more information than the previous level. The first level depicts the current amount of energy; the second level shows sparklines for the energy flow over the entire time interval; the third level shows a timeline of anomaly occurrences; level four shows the sparkline zoomed around the current time frame; finally, level five depicts expands the label to a large display of the previous four levels of detail of the flow data for the selected element. The new approach is implemented as an interactive web application. We are currently evaluating it with domain experts. Since the initial feedback is very positive, we expect rapid adoption of the newly proposed approach by automotive industry professionals.
Keywords: Scientific Visualization
Full text:
Year: 2023