A New Point Anomaly Detection Method About Aero Engine Based on Deep Learning

Abstract

Aero engine is one of the most important parts of the aircraft, so timely and efficient discovery of aero engine anomalies is essential for aircraft flight safety and airline economy. Aero engine anomaly detection method based on data driven can effectively improve the accuracy and economy of aero engine fault diagnosis. Based on data difference abnormalities can be divided into two categories: point anomaly and collective anomaly. This paper presents a method of aero engine point anomaly detection based on deep learning, and the method is composed of two main parts: the feature extraction and the anomaly detection. Feature extraction is primarily based on the construction of stacked denoising autoencoder to extract the features of the original input data, and anomaly detection is to build a single gaussian model to detect anomalies. The experimental results are analyzed to verify that the method proposed in this paper has obvious effect on point anomaly detection in the aero engine.