摘要
基于数据驱动的寿命预测是目前故障预测与健康管理(PHM)技术寿命预测环节的主流方法。通过对燃油泵出口压力数据进行分析,建立了基于随机效应维纳(Wiener)过程状态退化模型,采用贝叶斯(Bayes)方法对模型进行在线更新,应用最大期望(EM)算法实现模型的超参数估计,得到燃油泵剩余寿命在线预测信息。结果显示,其预测均值路径与实际退化情况基本吻合,寿命预测区间控制在40.62~20.57h,在接近寿命阈值阶段其预测不确定率保持在10%左右。本文所采用的分析方法与仿真得出的结果对于基于PHM技术下的维修保障活动具有一定的指导意义。
Data-driven life prediction method is the mainstream of PHM technology life prediction links. Through the analysis of fuel pump outlet pressure data, a process of degradation of Wiener process state based on random effect was established. Application of EM algorithm to realize super-parameter estimation model, the Bayes method was used to update the model online, getting the fuel pump residual life online forecast information. The result show that the predicted mean path is consistent with the actual degradation, and the life prediction interval is controlled from 40.62 to 20.57 hours, the uncertainty rate is about 10% near the threshold of life. The analysis method and simulation results obtained in this paper have some guiding significance for maintenance and protection activities based on PHM technology.
出处
《航空科学技术》
2017年第11期47-53,共7页
Aeronautical Science & Technology
