Interactive Visual Analysis in Engineering: A Survey Zolt´ an Konyha * VRVis Research Center Kreˇ simir Matkovi´ c † VRVis Research Center Helwig Hauser ‡ University of Bergen Abstract Interactive visual analysis has become a very popular research field. There is a significant body of literature on making sense of mas- sive data sets, on visualization and interaction techniques as well as on analysis concepts. However, surveying how those results can be applied to actual engineering problems, including both product and manufacturing design as well as evaluation of simulation and measurement data, has not been discussed sufficiently to date. In this paper we provide a selection of demonstration cases that doc- ument the potential benefits of using interactive visual analysis in a wide range of engineering domains, including the investigation of flow and particle dynamics, automotive engine design tasks and change management in the product design process. We attempt to identify some of the proven technological details such as the link- ing of space-time and attribute views through an application-wide coherent selection mechanism. This paper might be an interesting survey for readers with a relation to the engineering sector, both reflecting on available technological building blocks for interactive visual data analysis as well as exemplifying the potential benefits on behalf of the application side. CR Categories: I.6.6 [Simulation and Modeling]: Simulation Output Analysis; I.6.9 [Simulation and Modeling]: Visualization— Applications Keywords: interactive visual analysis, engineering, simulation, coordinated multiple views 1 Introduction Making sense of the massive amounts of data from engineering de- sign, simulation or measurement processes is by no means an easy task [Hauser 2006]. Traditional analysis procedures are based on computing various statistical properties of the data. Interactive vi- sual analysis [Thomas and Cook 2005] is a relatively new alterna- tive that has already gained a lot of interest. It allows the gradual exploration of data in a guided human-computer dialogue. Inter- active visual analysis takes full advantage of the advanced human visual and cognitive system to find unknown or unanticipated de- tails that could otherwise go unnoticed. The interactive nature of the analysis can often reveal more information than complex, but static visualizations. In fact, it calls for simple and effective visual- ization techniques that can be rapidly adjusted to provide the exact piece of information that the analyst needs at a given stage of the exploration. * e-mail: Konyha@VRVis.at † e-mail:Matkovic@VRVis.at ‡ e-mail:Helwig.Hauser@UiB.no Data in engineering application domains usually stems from mea- surements, or, more typically, from simulation. These data sets are large, complex, multidimensional and multivariate. The data is of- ten time-dependent and dependencies in the data are intricate. Dur- ing analysis, experts working with these data sets want to under- stand the behavior of the simulated or measured system, discover relations, create and support hypotheses. They often look for fea- tures and phenomena that they cannot exactly describe before the analysis. Interactive tools can assist them in the process of making sense of their data. This constitutes a domain where one can expect interactive visual analysis to be of great added value. There is often no single visual representation that can encode all of the important information contained in the data. Interactive visual analysis systems can show different aspects of the same data in sev- eral distinct views [Baldonado et al. 2000]. Each individual view can be textual (e.g. a table) or graphical (histogram, scatter plot, etc). The views can differ in the data they depict or in the visual representation of the data. The data displayed in one view can be a subset or an aggregate of the data depicted in another view, or it can be completely different information, e.g. a map that provides geo- graphical context. Using multiple views in a visualization system can be advantageous when the data attributes are diverse, when dif- ferent levels of abstraction need to be represented or when the users of the system exhibit different levels of expertise. Different views can highlight different properties and correlations in the data. They can help the user understand intricate, often surprising and unex- pected relations in the data. The visualization system can follow a “divide-and-conquer” approach and present partitions of the en- tire data set in individual views. This avoids the cognitive overload that users could face when they need to consider the entire data set at one time. Efficient interaction with the visualization is crucial in the analysis procedure. There are several concepts of coordinat- ing multiple views, including linked navigation, focusing, brushing and linking [Buja et al. 1991; Becker and Cleveland 1987]. There are numerous well known visualization systems based on these principles, including XmdvTool [Ward 1994], GGobi [Swayne et al. 2003], Snap-Together Visualization [North and Shneiderman 2000], WEAVE [Gresh et al. 2000] and Improvise [Weaver 2004]. Please refer to the paper by Matkovic et al. for a list of academic and commercial visualization tools [Matkovi´ c et al. 2008a]. After more than a decade of related research, there is a significant body of literature. Roberts provides an overview of many research publications related to coordinated and multiple views or one of the many associated boundary sciences [Roberts 2007]. He discusses developments in data preparation, concepts of creating and linking views, exploration techniques, window and sessions management, usability and perceptual issues and various display mediums. The book by Thomas and Cook [Thomas and Cook 2005] covers many aspects of visual analytics. We do not attempt to present another overview of the very broad topic, but rather focus on a very limited subset: interactive visual analysis in solving engineering problems. That means we do not discuss any of the very interesting devel- opments in medical visualization [Gresh et al. 2000; Oeltze et al. 2007], gene expression analysis [Weber et al. 2007; Saraiya et al. 2004] or visualization in software engineering [Bohner et al. 2007; Graˇ canin et al. 2005], either. The remainder of the paper is organized as follows: Section 2 deals with analysis in product and manufacturing design. In Section 3