A New Anomaly Detection Method Based on Multi-dimensional Condition Monitoring Data for Aircraft Engine

Abstract

Identifying the degradation state of aircraft engine accurately and carrying out suitable maintenance is significant for the safe and reliable operation of the aircraft system. In this paper, an anomaly detection system is proposed for multi-dimensional time series for aircraft engine. To represent the degradation degree of the device, the degradation index is proposed utilizing the Dynamic Time Warping algorithm (DTW) which possesses good similarity measure quality and can capture the changing characteristics of the indicator well. On the basis of the degradation index, three decision-making strategies including the Cumulative Sum algorithm (CUSUM), the Youden Index, and the Impact Factor are applied to determine the threshold values. For the threshold, two detection time points are obtained to discriminate different degradation stages in the device, in which the former refers to the initiation of degradation and the latter time point indicates that the device suffers from the anomaly. The experimental results demonstrate that all three decision strategies are able to detect system performance effectively before the device fails, which facilitates the incipient fault warning of aircraft engine in order to keep the equipment safe and reliable.