Research On Prediction Of Civil Aircraft Hard Landing And Reporting Customization

Reducing the occurrence of aircraft heavy landings is one of the important issues in the field of aviation safety.A heavy landing event will cause structural damage to the aircraft fuselage.In severe cases,it will directly lead to landing failure,resulting in serious casualties and economic losses.With the further application of aircraft real-time monitoring technology,predicting possible heavy landings in advance will become a reality,like the anti-runway system already used on many Airbus models.For this purpose,based on the current aircraft operating state parameters,pilot control,and meteorological environment,this paper not only designs a model that can predict heavy landing in advance,but analyzes whether there is a risk of heavy landing and provides technical support to warn pilots to respond accordingly.The main work and contributions of this paper are as follows:Firstly,the formation mechanism of heavy landing was studied and analyzed in this paper.From the perspective of aeronautical engineering,the QAR parameters that affect the occurrence of heavy landing during the approach and landing of the aircraft were screened.They mainly include the engine parameters that affect the flight performance of the aircraft,aircraft state parameters which reflect the pilot’s operation and meteorological environmental parameters from external influences.Based on this,the decoding principle of QAR data is analyzed in this paper.In accordance with the current recording defects of QAR parameters and the requirements of model input,interpolation and data compression methods are used to achieve the consistency of the sampling frequency of heavy landing QAR parameters.At the same time,based on feature engineering,grey correlation analysis are applied to eliminate high-relevant row parameters and reduce information redundancy and model training difficulty.Secondly,using the heavy landing QAR parameter sequence,combined with the multi-level classification and evaluation indicators of heavy landing in the AMM manual,a sequence-to-sequence(Seq2Seq)prediction method based on long-term and short-term memory networks(LSTM)is proposed.The landing data of multiple flights was selected as the training set,the flight height is used as the training set interval,and the input and output steps are adjusted to obtain the optimal model.In addition compared with the traditional LSTM sequence prediction model,it is concluded that the prediction model proposed in this paper realizes multi-step prediction of multiple evaluation indicators for heavy landing.It can predict the occurrence of heavy landing 8 seconds in advance,and provide sufficient event margin for the pilot to make a decision.MAE of the model is 0.012,which has higher accuracy and practicability than the traditional LSTM sequence prediction model.Thirdly,the heavy landing customized message designed in this study has more rigorous trigger conditions,and the message parameter records are more abundant.At the same time,based on the serial data of the heavy landing impact parameters recorded in the message,the gray correlation analysis was used to obtain the ranking of heavy landing impact parameters that have high correlation with the heavy landing evaluation indicators.This study not only reflects the causes of heavy landings,but provides maintenance personnel with a basis for judging the causes of heavy landings and assists in troubleshooting.