Research On Reconfiguration Of Redundant Inertial Measurement Unit With Multi-information

Inertial Measure Unit(IMU)is one of the key units of Guidance,Navigation and Control(GN&C)system.In order to ensure that flight vehicle can perform reliably,Redundancy and fault-tolerant design are widely discussed at home and abroad.The current researches mainly focused on the optimal configuration and Fault Detection and Isolation(FDI),but ignore the post-fault reconfiguration.For the requirements of improving the reliability and accuracy on abnormal conditions,the theory on post-fault reconfiguration is investigated in this thesis.Main contents are about the key technologies of RSIMU,including mathematical model,strategy of information application,compensation of parity check errors,fast switch on models,correlated noise and integrated navigation solution.The details of main contents are the following aspects:(1)Considering various error sources,a more precise output model is established.Take into account the correlation and the variance differences between residuals which are always ignored,Double transformations are proposed.Mahalanobis distance method is applied for standardization and unitary transform is applied for projection redundant information into data estimation space and null space for fault monitor.Similarities and differences are analyzed between SVD,PCA and proposed method.(2)Based on Bayesian Decision,the strategy for configuration is investigated.Considering the outliers,the probability model for information reliability is proposed.In that context,there is a risk to get all sensors outputs involved on account of false alarm and miss alarm.First,the fusion of two sensors with different precision is investigated by Bayesian decision stochastic model.Then,the impact is analyzed whether the fusion includes skew axe for the configuration of three orthogonal axes and one skew axe.It is shown that in the presence of fail alarm and uncertainty the fusion accuracy of all sensors involved would not improve but worse.The strategy named Triads First is proposed based on decision model.(3)The optimal configuration based on proposed strategy is investigated.Two indicators are put forward to judge the reliability of configuration,respectively the number of combination and minimum intersection angle.A structure named dual triad is optimized by geometric constraints and indicators.The precision of optimized dual-triad structure is compared with the configuration of dodecahedron in different abnormal modes.(4)A method based on the time series analysis is proposed for the residuals of Parity-check equations which are widely used in practical application.With the purpose of extracting the information in the residuals,AR(1)models are set up.Kalman filter is designed to eliminate the error of reconfiguration estimation.The method analyzes the error of reconfiguration model instead of every single sensor.(5)A new algorithm called Intersection fusion model is proposed which can reconfigure quickly with switching indicator.Three models are designed by prior rules,the indicator which can locate the failed sensor is obtained through the outputs.It is proved that a weighting combination of three outputs can reach the optimum estimation.Considering the error coefficients of gyros,a unified model is built for redundancy configuration.Unscented Kalman filter is used to monitor the error coefficients of gyros online.The simulation with configuration of 5 sensors and 6 sensors verifies that proposed approaches have a strong adaptability and high precision.(6)Error propagation models for RSINS are investigated in navigation algorithm.The optimal navigation solution for the SRIMU based INSs will be introduced.Considering the compensation on sensor outputs,the cascaded filter models are constructed.The proposed construction is verified by the configuration of 4 sensors which cannot locate the failed sensor by itself.