考虑误差不确定性的航空安全预测新方法

A novel method for aviation safety prediction considering error uncertainty

航空安全精确预测对预防事故意义重大。目前航空安全预测主要是确定性预测,忽略了各类不确定性对预测影响。在确定性预测基础上,考虑误差不确定性开展航空安全预测,通过非参数方法获得航空安全预测误差不确定性描述,基于最高密度域求解一定可靠程度航空安全预测值最可能落入区间,量化不确定性引起航空安全预测结果的变动,从而确定该区间包含航空安全预测值的可靠程度,更好地认识被预测量在未来变化中可能存在的不确定性和面临的风险。以某航空公司1994—2015年航空安全数据为例,采用所提方法对航空安全开展预测,结果表明,所提方法能提供航空安全预测值及其更精确的不确定性变化范围,更有利于从不确定性角度对航空安全进行分析,解释航空安全预测结果的可能性水平,能为航空安全预警和管理提供理论依据。

Accurate aviation safety prediction is of great significance for preventing accidents.At present,aviation safety prediction is mostly deterministic prediction,which ignores the influence of various uncertainties on the prediction results.Based on the traditional deterministic prediction of aviation safety point,this paper presents the prediction of aviation safety interval considering the uncertainty of error.First,through the description method of nonparametric uncertainties,aviation safety prediction error probability density function is derived.Then,the highest density domain method is applied for the most likely future value interval under a certain reliability of aviation safety,and quantitative uncertainty factors cause changes in the aviation safety prediction results.This method aims at determining that the area contains aviation safety forecast reliability,and better understanding the uncertainty and risk of those being predicted in the future change.Taking aviation safety data of civil aviation of an airline from 1994 to 2015 as an example,we predict the aviation safety using aviation safety interval prediction.The results show that the proposed method can provide aviation safety prediction curve and more accurate variation range of uncertainty,which is more conducive to modeling uncertainty of aviation safety and explaining the possibility level of aviation safety prediction results,which can provide theoretical basis for aviation safety early warning and management.