基于试飞数据的故障率预测及预警监控  

Failure Rate Prediction and Early Warning Monitoring Based on Flight Test Data

作　　者：胡毅[1] 李飞敏[1] 杨胜学[1] HU Yi;LI Feimin;YANG Shengxue(Chinese Flight Test Establishment, Xi’an 710089, China)

机构地区：[1]中国飞行试验研究院,西安710089

出　　处：《兵器装备工程学报》2020年第7期224-227,共4页Journal of Ordnance Equipment Engineering

基　　金：航空科学基金项目(2012ZA20003)。

摘　　要：针对试飞阶段故障率时间序列样本量小、非线性和非平稳性导致难以有效预测和监控的问题,提出了基于经验模态分解和最小二乘支持向量机的故障率时间序列预测方法,并引入了故障率警戒值监控的方法。该方法针对分解后规律性更强的子序列分别建立预测计算模型,使预测的风险分散化,提高预测精度;同时采用短周期历史数据计算故障率警戒值,解决预测结果无评价标准的问题。利用外场试飞数据进行了验证,充分表明了预测和预警监控方法的有效性,为试飞阶段的故障趋势监控、预测性维修和定检周期优化等提供技术支撑。The problems such as small samples,non-linearity and non-stationarity of failure rate time series in flight test phase make it difficult to predict and monitor effectively.A method of prediction based on EMD and LSSVM was proposed,and the method of warning value monitoring was introduced.This method established prediction calculation model for sub-sequence after decomposition,which can disperse the risk of prediction and improve the accuracy of prediction.At the same time,it calculated the warning value of failure rate based on short-period historical data,which can solve the problem that the prediction results have no evaluation criteria.The validity of the forecasting and early warning monitoring methods was fully demonstrated by the flight test data.Continuous and in-depth study of the method can provide technical support for failure trend monitoring,predictive maintenance and optimization of the inspection cycle in the flight test phase.

关 键 词：故障率 经验模态分解 最小二乘支持向量机 预测 警戒值 

分 类 号：TJ85[兵器科学与技术—武器系统与运用工程] V217.39[航空宇航科学与技术—航空宇航推进理论与工程]