奇异谱分析在故障时间序列分析中的应用
发布时间：2018-11-08 15:57
【摘要】：由于日益增长的飞行安全和飞机维护质量需求,飞机使用可靠性已经成为一个重要的研究领域。从某航空公司波音737飞机使用过程中现场所记录的18年的故障数据出发,应用奇异谱分析(SSA)方法,对故障时间序列进行了建模和预测,进一步以预测结果的均方根误差(RMSE)最小为优化目标对SSA模型参数进行了优选。在此基础上,提出了一种更为广泛的模型组合方法和实现算法,这种方法采用不同的时间序列模型来构造SSA分解出的趋势、周期和残差等成分。通过与三次指数平滑(Holt-Winters)、自回归移动平均(ARIMA)2种时间序列模型的实验结果对比,SSA及其参数优选和模型组合方法在故障时间序列分析中具有更好的拟合和预测精度。
[Abstract]:Aircraft reliability has become an important research field due to the increasing requirements of flight safety and aircraft maintenance quality. Based on the 18 year fault data recorded during the operation of a Boeing 737 aircraft, the fault time series was modeled and predicted by using singular spectrum analysis (SSA) method. Furthermore, the minimum root mean square error (RMSE) of the prediction results is taken as the optimization objective to optimize the parameters of the SSA model. On this basis, a more extensive model combination method and implementation algorithm are proposed. Different time series models are used to construct the trend, period and residual components of SSA decomposition. By comparing the experimental results with the two time series models of cubic exponential smoothing (Holt-Winters) and autoregressive moving average (ARIMA), SSA and its parameter optimization and model combination methods have better fitting and prediction accuracy in fault time series analysis.
【作者单位】： 北京航空航天大学计算机学院;中航工业西安航空计算技术研究所;
【基金】：中国民用航空专项研究项目(MJ-S-2013-10)~~
【分类号】：V267
本文编号：2318958
[Abstract]:Aircraft reliability has become an important research field due to the increasing requirements of flight safety and aircraft maintenance quality. Based on the 18 year fault data recorded during the operation of a Boeing 737 aircraft, the fault time series was modeled and predicted by using singular spectrum analysis (SSA) method. Furthermore, the minimum root mean square error (RMSE) of the prediction results is taken as the optimization objective to optimize the parameters of the SSA model. On this basis, a more extensive model combination method and implementation algorithm are proposed. Different time series models are used to construct the trend, period and residual components of SSA decomposition. By comparing the experimental results with the two time series models of cubic exponential smoothing (Holt-Winters) and autoregressive moving average (ARIMA), SSA and its parameter optimization and model combination methods have better fitting and prediction accuracy in fault time series analysis.
【作者单位】： 北京航空航天大学计算机学院;中航工业西安航空计算技术研究所;
【基金】：中国民用航空专项研究项目(MJ-S-2013-10)~~
【分类号】：V267
【相似文献】
相关期刊论文 前2条
1 左召军,钟新辉;航材消耗的时间序列分析[J];长沙航空职业技术学院学报;2004年03期
2 ;[J];;年期
，本文编号：2318958
本文链接：https://www.wllwen.com/kejilunwen/hangkongsky/2318958.html
