Research on the application of random forest algorithm in predicting tail strike events
Ensuring flight safety is paramount in civil aviation operations.A tail strike incident involving an aircraft can result in structural damage to the airframe,posing significant risks that could potentially lead to accidents.To facilitate the prediction of tail strike events,a non-exceedance flight database was introduced.By integrating flight state analysis,the random forest algorithm was employed to develop a predictive model for tail strike events.The effectiveness of the model was validated across various dataset sizes.Initially,pertinent preprocessing was carried out on the obtained QAR data for the B737-800 model to safeguard the integrity of the dataset.Furthermore,thirteen distinctive parameters influencing landing tail strike events were derived through an analysis of the landing operation process and the mechanism of tail strike formation.Subsequently,the importance scores of various features were calculated using out-of-bag data.The optimal feature combination parameters were extracted through calculations of importance scores.Finally,the authors utilized the random forest algorithm to establish a predictive model for strike events during the landing phase,which was then compared with the support vector regression and long-short memory network models.Aircraft tail strike prediction methods employing random forests can effectively harness a substantial amount of potentially valuable data to provide proactive alerts regarding tail-strike risk.The results demonstrate that the coefficient of determination of the test set for the proposed model exceeds 0.85 when predicting 7 seconds before landing.Furthermore,the model maintains its accuracy even when the scale of the data is adjusted.For instance,the coefficient of determination for the test set of the proposed model exceeds 0.9.When compared to support vector regression and long short-term memory models,this approach showcases superior fitting and higher predictive accuracy.It affords pilots ample time to react and implement necessary measures to prevent tail strike events from occurring.
NETL
NSTL
万方数据
