Heavy landing prediction method based on deep learning

Abstract

The invention discloses a heavy landing prediction method based on deep learning, which is used for solving the problem of predicting whether a heavy landing risk occurs in an aircraft or not according to parameters of the flight stage of the aircraft. The method mainly comprises a construction method of an autoregressive landing model based on deep learning and a method for heavy landing detection by adopting the autoregressive landing model, wherein the method comprises the steps of: (1) an autoregressive landing model which is a model trained based on a deep learning model by using a largeamount of sample data of a normal landing flight and used for capturing an airplane landing state conversion mode, and (2) a calculation method: flight process data, such as data acquired from a QuickAccess Recorder (QAR), is taken as input, and an aircraft landing multivariate time sequence is converted into a symbolic representation sequence by utilizing time sequence symbolic representation and a landing state dictionary. The symbolic representation sequence of the flight to be predicted is subjected to fitting by using a pre-trained autoregressive landing model, and the prediction resultof the heavy landing risk occurrence probability is given.

Description

technical field [0001] The invention relates to the field of civil aviation, is a landing quality prediction method based on flight data, and is a method for intelligently detecting landing risks under the condition of big data. Background technique [0002] In the field of civil aviation, the landing phase is the phase with the most accidents. Studies have shown that 47% of accidents occur during the landing phase. Safety is the lifeline of the civil aviation industry, and ensuring the safety of civil aviation flight operations is the prerequisite for all specific civil aviation activities. The landing phase is a high-risk phase of flight operation. The accident rate accounts for 20% of the total flight phase, and the accident death rate accounts for 23% of the total flight phase. The incidence and death rate of accidents and unsafe events are significantly higher than other flight phases. It can be seen that the risk of aircraft landing is the biggest source of risk in t...

AI Technical Summary

Problems solved by technology

First of all, the aircraft landing process is very complicated. Using the method of expert extraction feature engineering, the prediction effect of heavy landing depends to a large extent on the feature structure, and the complex landing process is very difficult for feature engineering.

Secondly, the number of heavy landing samples is very small, so when training the model, there will be an imbalance of positive and negative samples, or there are few training samples, making the model less effective

Images