Detecting and Analysing Spatial-Temporal Aggregation of Flight Turbulence with the QAR Big Data

Flight turbulence is a non-ignorable risk during flying, and its timely detection and prediction are greatly important for avoiding safety accidents. Currently, flight turbulences are detected simply by the measured vertical acceleration value, or reported by the pilots after a flight. All these means or criteria are relatively not substantial, either persuasive or scientific enough. In particular, the turbulence data reported traditionally is far from sufficient for analyzing this kind of events in depth. On the other hand, the QAR data collected, which is characterized by completeness, high accuracy, real-time and fine-grained time-series, can provide a much better data source for flight safety studies. With it, flight turbulences could be detected in real-time via the random forest method. Results show that the detection model can effectively identify flight turbulence within an acceptable tolerance. The different degrees of attribute importance reveal the spatio-temporal aggregations of flight turbulences. This will provide pilots with warning of flight turbulence in real-time to reduce flight risk. In addition, this study also provides important references for flight routing and schedule planning.