A Feature Representation Method Based on Dual Segment and Entropy Evaluation for Aeroengine Gas Path Anomaly Detection

Traditional methods for gas path anomaly detection cannot fully extract remarkable shape features that can represent the gas path anomaly mode. Therefore, a feature representation method based on dual segment and entropy evaluation for aeroengine gas path anomaly detection is proposed in this paper. Taking the temporal and spatial correlations of the multivariate time series into consideration, the expression rule of the anomaly mode in the multivariate gas path parameter deviation time series is analyzed, on this basis, time series subsequence segment method is determined. To obtain the features that best fit the anomaly expression rule, a dual segment method based on piecewise optimal fitting is proposed. The entropy evaluation method is introduced to comprehensively evaluate and optimize the primary features while calculating the common shape features of subsequence, and then the remarkable shape feature matrix for anomaly detection is determined. Finally, the early warning for the gas path anomaly is realized by mining the potential anomaly mode of the gas path state using isolation forest model. The experimental results show that this method can improve the accuracy of aeroengine gas path anomaly detection.