The Research on Missing Data Imputation Method of Aero-Engine’s ACARS Based on GAN-Attention

The ACARS data is an important data source for Real-time flight status monitoring. However, ACARS data is affected by unstable factors during transmission and storage, resulting in complex missing situations. We focus on a data-driven class of methods based on deep learning applied to the reconstruction of ACARS data. The main work and innovation points are as follows: We categorized four missing patterns and generated corresponding datasets for training and testing models. The GAN architecture is adopted to deal with the problem that the ACARS data are dimensionally wide and difficult to be extracted. Encoder and decoder are used to form the generator of GAN, allowing the model to compress existing information into deep features and then generate a complete data set. And the attention mechanism is introduced to enhance the learning ability of the model for long-distance information. Finally, the effectiveness of the proposed method is verified by comparing it with different reconstruction models under four missing modes and ablation experiments.