COMET Project Highlight “Visual Signal Search for Engine Anomaly Identification”

Knowledge Visualization

A signal search service enables experts to find, explore and annotate signal patterns in large scale signal repositories.

The service is built upon novel technology developed at KC (KV, KD and Software Services) to store, retrieve and browse data in large signal repositories. Users can visually spot candidate signal patterns, which the service uses to retrieve related occurrences in the whole repository. Sophisticated algorithms were developed to i) index large signal repositories ii) retrieve signal candidates from them iii) compare and rank based on similarity, considering noisy and time-amplitude-scaled patterns. A web-based visual interfaces was introduced for search and exploration of large scale signal data. It uses server-side hierarchical aggregation, with a down-sampling algorithm optimized to preserve visual features. Experts can smoothly explore a large signal space, brushing pattern candidates to search, while the client dynamically fetches data fragments depending on time-interval and zoom level to reduce network traffic. The tool has been developed in a three year project with AVL, and transferred to signal spaces in other different fields of operation. This major effort on Big Data Analytics is a source of research questions addressing visual exploration of large scale data to be investigated as part of our next funding period.