Recognising Visual Patterns to Communicate Gas Turbine Time-Series Data

We have observed that visual patterns play an important part when domain experts interpret time series data. Such patterns change their appearance when displayed at different time scales and a systematic method is proposed to handle this problem. First, a rapid change detector combined with a dynamic limit checker (DRCD) is employed to detect primitive patterns at a basic time scale. Patterns obtained at that time scale are then transformed into patterns viewed at a higher time scale and the DRCD algorithm is reused at this time scale to identify new visual patterns that did not appear at lower time scales. An evaluation of the preliminary results is promising.