A Data-driven Integrated Safety Risk Warning Model based on Deep Learning for Civil Aircraft

With the extensive application of sensor technology, airlines accumulate a lot of flight data during fleet operations. Quick Access Recorder (QAR) is an important basis for aircraft state estimation and safety assessment, it records the operating status of various aircraft systems. However, the current application of QAR is mainly reflected in the investigation and verification of accidents, but it fails to apply such continuous real-time data to safety monitoring, risk prediction, and early warning. Therefore, this paper proposes a data-driven integrated safety risk warning model based on deep learning. Combined with the characteristics of QAR data and aircraft system failures, the failure mode and impact analysis (FMEA), cause chain analysis (CCA), long and short-term memory (LSTM), and other methods are used to describe the development of risks caused by system failures and predict the trend of flight parameters of each system, improve the ability to predict the risk and risk severity of the aviation system. Taking the Aircraft Refrigeration System (ARS) as the case object, the safety risk warning model is validated to calculate the risk level may face and find out the hidden hazard sources of civil aircraft operation in advance, to improve the safety level of civil aircraft operation