A hybrid algorithm for detecting changes in diagnostic signals received from technical devices

In this paper, a hybrid two-level algorithm of the original changes in diagnostic signals received from multiple technical devices is presented. Research are aimed at identification of the changes, deviations or patters (events), through concurrent diagnostic signals processing, which occur in one selected signal (proposed algorithm is adjusted to omit concurrent and time-lagged changes). In the first stage, detection is based on non-stationarity detection with the short-term prediction comparison. In the second stage, dedicated distance-like measure is employed. Detection results obtained for sample random signals including simulated large deviations are presented.