Conference article

This paper describes the design of a visual analysis tool for our Inclusion Processing Framework, called IPFViewer. The tool has been designed in cooperation with a large German steel production facility in order to acquire knowledge from data collected about nonmetallic inclusions and other defects in steel samples. We have highlighted parts of the framework in previous publications in interdisciplinary journals. Here, we describe our contribution in the area of grouped or clustered data items. These data groups are visualized by techniques known from uncertainty visualization to make visible fluctuations and corresponding variations in steel samples. However, our results are also transferable to other ensemble data. To find an optimal way to design algorithms and visualization methods to process the huge data set, we discuss the project-specific requirements regarding memory usage, execution behavior and precision. By utilizing approximate, incremental analysis techniques when needed, the responsiveness of the application is ensured while high precision is guaranteed for queries with fast response times. The design allows workers at the steel production facility to analyze correlations and trends in billions of data rows very quickly and to detect outliers in routine quality control.

No keywords available