关键词: 民机空调系统, 异常检测, 自编码器（AE）, 长短期记忆网络, 孤立森林（iForest）

Abstract: The key to quickly judging the operating state of the air conditioning system and efficiently arranging maintenance tasks lies in the abnormal performance detection technology of the civil aircraft air conditioning heat exchanger. However， traditional abnormal detection methods struggle to effectively handle high-dimensional time series data and fail to provide early fault warning for the system. In light of this， this paper proposes an unsupervised anomaly detection method based on long-short term memory (LSTM) and autoencoder (AE) to identify abnormal operating state of the civil aircraft air conditioning system. Firstly， characteristic monitoring parameters that reflect the per鄄 formance of the air conditioning heat exchanger is constructed based on the original sensor parameters of the civil aircraft air conditioning system. Subsequently， the LSTM-AE model is employed to reconstruct the data features and calculate the reconstruction errors. Finally， unsupervised anomaly monitoring is conducted using isolation for鄄 est (iForest). The constructed unsupervised anomaly detection method is compared with the traditional method， and a model is established to evaluate indicator. The verification results demonstrate that the proposed model can effec鄄 tively detect abnormal states of the performance of civil aircraft air conditioning heat exchanger

Key words: civil aircraft air conditioning system, abnormal detection, autoencoder (AE), long-short term memory (LSTM), isolation forest (iForest)