Research on Real-time Early Warning Mechanism of Aviation Safety Based on Finite State Machine Underlying in QAR Stream Data.

Flight safety plays a crucial role in the civil aviation industry. However, contemporary research predominantly focuses on accident detection, and traditional early warning methods rely on crew members, rendering them less reliable. This study addresses these limitations by employing data mining techniques on Quick Access Recorder (QAR) stream outputs. The overarching strategy involves predicting future states based on current states. To achieve this, the paper constructs a state queue founded on the principles of Finite State Machines (FSM) to ascertain the current aircraft state from the QAR stream. Subsequently, features are extracted from the queue and utilized to predict imminent over-limit events. The efficacy of the warning mechanism is then validated using the Random Forest model. The real-time early warning mechanism proposed in this study offers an effective safety assurance for airliners without imposing undue computational power demands.