基于参数递归更新的惯性器件寿命预测  

Inertia Device Lifetime Prediction Based on REM Method

作　　者：张会会[1] 张伟[2] 胡昌华[1] 

机构地区：[1]第二炮兵工程学院302教研室,西安710025 [2]第二炮兵工程学院402教研室,西安710025

出　　处：《系统仿真学报》2013年第5期1036-1040,共5页Journal of System Simulation

基　　金：国家自然科学基金重点课题(60736026)

摘　　要：在基于随机滤波理论的剩余寿命预测模型中,模型参数的获取是离线的,且当历史数据较少时,模型参数不能够进行修正,影响了预测精度。针对以上问题,采用递归期望最大化(Recursive Expectation Maximization,REM)算法来对模型参数进行递归更新,提出了基于参数递归更新的剩余寿命实时预测模型。应用实际的监测数据和参数递归更新的寿命预测模型,进行了某导弹陀螺仪的寿命预测实验。仿真结果表明,该预测模型能够根据实时的数据对模型参数进行快速地更新,满足预测的实时性要求。In residual life prediction model based on stochastic filtering theory, the estimation of initial parameters is offline, and it is inaccurate when the data is limited. In order to resolve these problems, a new analysis method for residual real-lifetime prediction based on recursive expectation maximization （REM） algorithm was proposed. In this model, REM algorithm was introduced to update the parameters. The forecasting model was used to predict the inertia system＇s residual life by using the real monitoring data. Experimental results indicate that the model can renew the parameters fast and has higher precision, which can fulfill the request of on-line prediction.

关 键 词：寿命预测 递归期望最大化 参数更新 实时预测 

分 类 号：TP202.1[自动化与计算机技术—检测技术与自动化装置]