VIMA: Modeling and visualization of high dimensional machine sensor data leveraging multiple sources of domain knowledge

Joscha Eirich, Dominik Jäckle, Tobias Schreck, Jakob Bonart, Oliver Posegga, Kai Fischbach

Research output: Chapter in Book/Report/Conference proceedingConference paperpeer-review


The highly integrated design of the electrified powertrain creates new challenges in the holistic testing of high-quality standards. Particularly test technicians face the challenge, that lots of machine-sensor data is recorded during these tests that needs to be analyzed. We present VIMA, a VA system that processes high dimensional machine-sensor data to support test technicians with these analyses. VIMA makes use of the concept of interactive labeling to train machine learning models and the process model of knowledge creation in visual analytics to create new knowledge through the interaction with the system. Its usefulness is demonstrated in a qualitative user study with four test technicians. Results indicate that through VIMA, previously undetected abnormal parts, could be identified. Additionally, a model trained with labels generated through VIMA, was deployed on a test station, that outperforms the current testing procedure, in detecting increased backlashes and improved the test benches output by 15%.
Original languageEnglish
Title of host publication2020 IEEE Visualization in Data Science (VDS)
PublisherIEEE Press
Number of pages10
ISBN (Electronic)978-1-7281-9284-0
Publication statusPublished - Oct 2020
EventIEEE VIS 2020 - Virtuell, United States
Duration: 25 Oct 202030 Oct 2020


ConferenceIEEE VIS 2020
Abbreviated titleVIS 2020
Country/TerritoryUnited States
Internet address


  • anomaly detection
  • interactive labeling
  • knowledge creation
  • machine learning
  • Visual analytics

ASJC Scopus subject areas

  • Computer Science Applications
  • Media Technology

Fields of Expertise

  • Information, Communication & Computing


Dive into the research topics of 'VIMA: Modeling and visualization of high dimensional machine sensor data leveraging multiple sources of domain knowledge'. Together they form a unique fingerprint.

Cite this