基于全航段QAR数据和卷积神经网络的航空发动机状态辨识

为了充分挖掘全航段飞行数据中蕴含的丰富信息以提高发动机状态辨识的准确率，提出一种基于全航段快速存取记录器（QAR）数据和卷积神经网络的发动机状态辨识方法。该方法将每次飞行循环的全航段QAR数据变换为一个红绿蓝（RGB）多通道样本实现全航段数据图像化处理，根据发动机维修记录中的水洗时间，将发动机划分为不同的衰退状态，采用卷积神经网络对不同衰退状态进行分类和辨识。该方法经某航空公司飞机QAR数据验证，结果表明：基于全航段QAR数据的衰退状态辨识算法的精确度相比于仅使用巡航段数据的精确度提升超过13%，辨识准确率达到98%。

Aero-engine status identification based on full-segment QAR data and convolutional neural network

To improve the identification accuracy of engine states, an identification method based on full-segment quick access recorder (QAR) data and convolutional neural network was proposed by extracting and mining rich information from the real flight data. The full-segment QAR data of each flight cycle was transformed into a multichannel imaged sample in the red, green and blue (RGB) channels to realize the imaging process of the full-segment QAR data. Engine state was divided into different decay states according to the washing time in the engine maintenance record. The convolutional neural network was used to identify the different decay states of the engine. The proposed method was verified using the QAR data of a commercial plane. The results showed that the proposed method based on the full-segment QAR data improved the identification accuracy up to 98% and was 13% higher than that based on the data of the cruise segment only.