摘要
为了掌握运输航空飞行事故征候的发展趋势,提高运输航空安全管理水平,构建了组合预测模型对运输航空飞行事故征候万时率进行预测研究。基于2003-2015年运输航空飞行事故征候万时率的原始数据,采用灰色GM(1, 1)模型、长短期记忆神经网络模型(LSTM)、自回归滑动平均模型(ARMA)作为单一模型进行预测,建立了基于诱导有序几何加权平均算子(IOWGA)的飞行事故征候万时率组合预测模型,并将单一预测模型与组合预测模型进行了对比。结果表明,组合预测模型对飞行事故征候万时率的拟合程度较好,具有较高的预测精度,运算相对简便。最后利用基于IOWGA的组合预测模型对2016-2018年运输航空飞行事故征候万时率进行了预测。
In order to improve the level of air transport safety management and grasp the trend of air transport flight accident symptoms,this paper forecasts the aviation incidents per 10000 flight hours by a combined prediction model.Based on the original data of the aviation incidents per 10000 flight hours from 2003 to 2015,the gray GM(1,1)model,long short-term memory(LSTM),and auto-regressive moving average model(ARMA)are used as a single model for prediction.Then,a combined model is established and the weight coefficients of the combined forecasting model are determined by the induced ordered weighted geometric averaging(IOWGA).The prediction accuracy of the four methods is analyzed and compared.The results of calculation and investigation indicate that the gray GM(1,1)model shows a downward trend from 2003 to 2004,and the rest of the years show a monotonical increase.The fitting curve is relatively stable and could not reflect the fluctuations well.Therefore,in the case that the original data fluctuate greatly from 2005 to 2009,the grey model has a lower prediction accuracy and a larger error.LSTM model and ARMA model can reflect the reality of the original data between fluctuations,especially under the condition of the same trend.The prediction results are very good while when faced with the actual value of a turning point in its predictive value changes lags behind the actual value.It shows that these two models have a great dependence on the original data of the previous period when making predictions.Since the IOWGA combined prediction model is a dynamic fusion of three single prediction models,the prediction value of the combined prediction model is closer to the actual value at each time point,has higher prediction accuracy,and is relatively simple to operate.Finally,the combined prediction model based on IOWGA was used to predict the ten thousand hour rate of flight accident symptoms from 2016 to 2018.
作者
文军
陈建勤
罗心雨
徐志航
WEN Jun;CHEN Jian-qin;LUO Xin-yu;XU Zhi-hang(Airport Engineering and Transportation Management College,Civil Aviation Flight University of China,Guanghan 618307,Sichuan,China)
出处
《安全与环境学报》
CAS
CSCD
北大核心
2022年第1期256-262,共7页
Journal of Safety and Environment
基金
国家自然科学基金项目(61179074)
中国民用航空飞行学院研究生科研创新计划项目(X2020-16)。
