基于ARIMA模型的航空装备事故时序预测

为提高航空装备事故预防的针对性、有效性和主动性,预防和减少事故的发生,降低事故造成的损失,提出一种时序的差分自回归滑动平均(ARIMA)模型。其建模过程先在时间序列基础上辨识一个试用模型,然后加以诊断,并作出必要调整,反复进行辨识、估计、诊断,直至获得较为满意的ARIMA预测模型。在实例验证中,所构建的用来预测美国空军飞行事故万时率的ARIMA模型,能够将预测的平均相对误差控制在7%以内,预测结果总体反映航空装备的实际安全状况。

Time Series Prediction of Aviation Equipment Accident Based on ARIMA Model

To improve the pertinence,effectiveness and proactiveness of aviation equipment accident prevention,prevent and reduce accident and its loss,a prediction model of time series based on ARIMA was built.Its modeling process was that: first a trial model was identified on the basis of time series,and then the diagnosis was conducted with the necessary adjustments.This process,that includes identification,estimation and diagnosis,was repeated until a satisfactory model was obtained.In the actual example,the model built for predicting flight accident 10 000-hour-rates of USAF can obtain good prediction results with average relative errors being not larger than 7%,which reflect the actual safety condition of aviation equipment on the whole.