摘要
首先利用基于线性漂移的布朗运动(Brownian Motion,BM)表征陀螺仪退化过程,然后构建状态空间模型表征实际退化状态与带噪声的监测量之间的随机关系。利用期望最大化(Expectation Maximization,EM)算法和卡尔曼滤波(Kalman filter)估计与更新未知参数和退化状态。并且考虑状态估计的不确定性,将状态估计的分布函数引入剩余寿命的预测过程,实现了剩余寿命的实时预测与更新。利用该方法对陀螺仪的剩余寿命实时预测问题进行了分析,比较结果表明,该方法能够较好地解决剩余寿命估计过程中直接监测量含有噪声的问题。
Linear drift-driven Brownian Motion (BM) is used to characterize the actual degradation process of the gyro and the state space model is established to model the stochastic relationship between the actual degradation state and the direct measurement. With the Expectation Maximization (EM) algorithm and Kalman filter, the updating parameters and states are jointly estimated. Furthermore, considering the uncertainty of the state estimation, we introduce the distribution of the estimated state into the derivation of the residual life to realize the real-time residual life prediction for the gyro. The results show that the method can precisely estimate the residual life of the gyro with the monitoring noises.
出处
《长春工业大学学报》
CAS
2013年第2期155-159,共5页
Journal of Changchun University of Technology
基金
国家自然科学基金资助项目(NSFC:61004069)
关键词
剩余寿命
预测
状态空间
期望最大化
卡尔曼滤波
residual life
prediction
state space
expectation maximization
kalman filter.