Research On Aircraft Ice Accretion Prediction Method And Route Distribution Characteristics

With the rapid development of the aviation industry,unsafe operation incidents have increased,and the meteorological conditions(especially aircraft icing)is an important factor affecting the safety of aircraft.While the aircraft in the phases of taking-off or landing and fly in the severe icing area,may occur accumulation of icing,or even cause flight incidents.Based on the analysis of aircraft report of 2013-2016,this paper summed up aircraft icing route distribution in our country.Then this article uses MATLAB to decode and extract the reanalysis data of NCEP(including temperature and relative humidity,etc),screen out meteorological factor data related to aircraft icing formation,and establish China’s aircraft icing Prediction model based on the multiple linear regression,BP neural network method,and compare the forecast consequent with icing index forecasting equation recommended by the International Civil Aviation Organization.The BP neural network model has achieved satisfactory results.The main conclusions are as follows:1.Using the 2013-2016 China aircraft icing flight report to analysis the aircraft icing route distribution characteristics of China to summarize the aircraft icing prone areas distribute in Zhejiang,Xinjiang,Sichuan,Chongqing,Yunnan and Henan,etc;The aircraft icing height distributed under 400hPa(inclusive)and accounts for 98% showing the aircraft icing is most likely to happen in the middle and the low altitude in China;aircraft icing frequency is highest in winter from November to February of the following year.2.The statistical analysis of aircraft icing meteorological factors shows 83.3% of aircraft icing occurring between-20 to 0 degrees;The uniform distribution of the relative humidity indicates that there may be water vapour in the air and aircraft icing accumulation may occur.;By statistically analyzing the vertical velocity and plotting the water vapor flux divergence map and relative vorticity profile in Sichuan Province,it is concluded that the weaker rising motion creates more favorable conditions for the formation of aircraft icing.3.Using multiple linear regression and BP neural network algorithm to build icing forecasting model,the results show that the CSI score of multiple linear regression model is 48.8%,the CSI score of BP neural network regression model is 63.4%,the CSI score of International Civil Aviation Organization recommended aircraft icing formula is 17.1%.BP neural network model predicts the best.