Research On The Key Technology Of Fdi For Redundant Strapdown Inertial Navigation System

Redundant strapdown inertial navigation system(SINS)can improve reliability by a single sensor by redundant configuration technology.Based on the spatial orientation relationship of redundant sensors,fault detection and isolation(FDI)for redundant SINS can detect and isolate a fault,and complete system reconstruction,which can guarantee system work normally under bad working condition,and has been become the most important development direction for SINS in high reliability applications.However,redundant SINS still face some problems,such as the real-time performance of soft fault,feasibility of single fault isolation,uncertainty of double faults isolation and so on.Thus,researching fault detection and isolation for redundant SINS is vital in improving performance of relevant theory and promoting the application of redundant SINS.This dissertation will take four-gyro SINS and six-gyro SINS as the research object,and focuses on the key issues of SINS on the basis of optimizing redundant configuration,such as real-time fault detection approach and feasible single/ double faults isolation approach.(1)Design of redundant configuration based on optimal performance indicatorTo solve the problem that existing redundant configuration design criteria is lack of fault detection and isolation performance criteria,this dissertation views four gyros and six gyros as research objects,summaries existing redundant configuration design criteria,and analyzes the relationship between configuration and fault detection and isolation performance systematically,which improves redundant configuration design criteria.Among the existing redundant configuration,optimal one is chosen according to improved optimal redundant configuration design criteria.(2)High real-time performance approach for a soft faultGeneralized likelihood test(GLT)fault detection approach based on Kalman filter(KF)cannot detect a soft fault in time because of the feature that soft faults change slowly and KF suppresses measurement residual.In this dissertation,an improved GLT fault detection approach based on interactive multiple model(IMM)and average parity vector(APV)is proposed to improve soft fault detection.In this approach,on the one hand,IMM is applied to instead of KF to compensate parity vector,which make fault not be estimated as a part of the state,and soft fault detection performance is improved.On the other hand,APV is applied to improve GLT fault isolation approach to improve real-time performance for a soft fault.Compared with GLT fault detection based on KF,the improved GLT fault detection based on IMM and APV improve soft fault detection real-time performance.(3)Single fault isolation approach for four-gyro SINSGLT fault isolation approach isolate a fault through the direction of parity vector.However,the parity vector is not a vector in four-gyro SINS.As a result,GLT cannot isolate a fault in four-gyro SINS.In this dissertation,an improved residual chi-square test fault isolation approach based on an average reduced dimension vector is proposed.With angular velocity baseline provided by star sensor,this dissertation improves residual chi-square test approach by a reduced dimension vector,and designs residual chi-square test fault isolation function for each gyro based on a reduced dimension vector,which realize single fault isolation in four-gyro SINS;in addition,the thought of APV approach is applied to propose an improve residual chisquare test fault isolation based on a average reduced dimension vector,which can be used to isolate a soft fault in time.The improved residual chi-square test fault isolation approach based on an average reduced dimension vector can isolate a fault in dynamic environment,and can improve the isolation performance for a soft fault.(4)Double faults isolation approach for six-gyro SINSExisting double faults isolation approach based on maximum likelihood estimation(MLE)cannot isolate all of the double faults combination.In six-gyro SINS,when different double faults generate collinear parity vector,and it is hard to judge the collinear parity vector generated from which candidate.In this dissertation,the application range of double faults isolation approach based on MLE is first studied,and then improved double faults isolation approach based on MLE and reduced-order parity vector(RPV)is proposed.In this approach,double faults isolation approach based on MLE can recognize candidate faulty gyros,and improved RPV faults isolation approach is proposed by building reduce-order redundant SINS to recognize detailed double faults for six-gyro SINS.The improved double faults isolation approach based on MLE and RPV can isolate double faults for six-gyro SINS when double faults isolation approach based on MLE cannot isolate double faults.