Fault-tolerant Processing Technology And Appliction In Momentum Wheel Speed Data Processing

Because of the influence of system running environment,the sensitivity limitation of sensor and the error of recording process,the measurement data inevitably appears a few abnormal data(also known as Outliers)in the sampling process,which obviously deviates from actual value.A majority of classical optimal identification,filtering and prediction algorithms lose optimality or validation because of the abnormal data.Taking the momentum wheel speed data of the satellite attitude control system as example,there is some sampling data that is seriously different from most of the sample point because of the equipment temperature or environment mutation.The existence and disadvantage of abnormal data can seriously interfere with the control of satellite attitude.Therefore,the analysis of the momentum wheel speed data,which contains the outliers,is of great significance to determinate,predict and analysis the attitude of the satellite.The classical recursive maximum likelihood algorithm,Kalman filter algorithm and autoregressive model are studied respectively in this paper.When momentum wheel speed data appears outliers,it is found that model identification accuracy would become lower,filtering effect would appear obvious deviation,and the result of forecast would be not ideal.Based on the idea of fault-tolerant data processing,the tolerance of the classical algorithm is improved and a series of fault-tolerant identification,fault-tolerant filtering and fault-tolerant prediction algorithm is implemented in dealing with outliers.Computational software with the fault-tolerant algorithm is developed independently.Meanwhile,the fault-tolerant algorithm is verified by using the momentum wheel speed measurement data with outliers.The results show that the established fault-tolerant algorithm is ideal in identification and filtering effect.Furthermore,the momentum wheel speed data of the next period can be forecast accurately even if momentum wheel speed data contains abnormal data.The innovation of this paper is the fault-tolerant algorithm of recursive maximum likelihood parameter identification,Kalman filtering algorithm and autoregressive prediction.The fault-tolerant algorithm can be directly used for fault tolerance identification,fault tolerant filtering and prediction to deal with measurement data with abnormal data.In additions,this paper provides practical tools for modeling,model identification,state filtering and prediction in the case of abnormal data.This paper is founded by the National Natural Science Foundation of China(61473222,91646108).