Study On Some Key Techniques Of Autonomous Health Management For Spacecraft Propulsion System

As a key technology for realizing the spacecraft autonomy,autonomous health management has important practical significance and engineering value.It can ensure reliability and safety of the spacecraft under the unknown and uncertain environment,with reducing reliance on ground-based intervention and management cost.Taking the typical spacecraft propulsion system such as the DFH-4 & DS-1 propulsion system as the object,this dissertation thoroughly investigates some key technologies involved in autonomous health management,including robustly fault detection and diagnostic methods,fault recoverability evaluation,reconfiguration under failure,etc.The hardwares and some core modules of a prototype system are also designed and implementated.The main research achievements are summarized as follows.In order to deal with the problems such as imbalanced-class problem,unequal misclassification costs in fault detection for propulsion system,by introducing cost-sensitive mechanism,a CS-SVDD-BC(Cost-Sensitive Support Vector Data Description with Boundary Calibration)method is proposed.In order to solve the deficiency of empirical values of CS-SVDD-BC parameters,CS-SVDD-BC parameters are optimized using PSO(Particle Swarm Optimization).In view of the existence of higher misjudgment risk and low training efficiency of standard SVDD,Inc CS-SVDD-BC based on incremental learning is also proposed.Verification through test data of propulsion systems shows that the PSO-based parameter optimization greatly reduces the error rate and average misclassification cost with improving the detection performance of CS-SVDD-BC.Through online adaptive learning,Inc CS-SVDD-BC enhances the separability between the normal sample space and fault space,with improving the timeliness and validity of fault detection,which has better fault detection performance than the static one.In order to deal with the time-variance and the characteristic of “bidirectional failure transmission” of spacecraft propulsion system,a tense-oriented fault diagnostic method based on the combination of Z-test and analytical redundancy relations is proposed.The time-invariant structural information mined through the analytical redundancy relation residuals can effectively avoid the problems existed in the acquisition of time-variant diagnostic criterions.The time-invariance of the system's structural characteristics overcomes the difficulty of fault location caused by bidirectional transmission.Simultaneously,in order to solve the disturbance of modeling error and environment noise,the Z-test is used to deal with the residual qualitation,which can ensure the robustness during fault diagnosis.Verification through the fault simulation data shows that the combination method is simple and easy to be implemented with achieving real-time and efficient fault location,which is suitable for on-orbit autonomous fault diagnosis of the spacecraft propulsion system.In order to deal with promoting the fault recoverability evaluation,combining with the lattice of system configurations,the fault recoverability analysis framework based on the property span of spacecraft propulsion system is approached,and proposes three kinds of fault recoverability evaluation metrics such as the coverage,redundancy degrees and the MTTF(Mean-Time To Failure)based on the component reliability estimation.Based on the above metrics,the recoverability evaluation on the booster gas path of DFH-4 and DS-1 propulsion system is carried out.The evaluation results show that the design requirements of propulsion systems are satisfied,which verify the effectiveness of the proposed fault recoverability measure metrics.In order to deal with the existed complex knowledge representation and low search efficiency at solving configuration and reconfiguration of the spacecraft system,based on the translation-based insight,a reconfiguration problem description framework is proposed by translating the reconfiguration problem into an optimal satisfiability(Op SAT)problem.Meanwhile,a new optimal search strategy--CWBA* is designed and achieved for the Op SAT problem.Verification with single fault,multiple faults of the DS-1 propulsion system shows that CWBA* is increased by nearly an order of magnitude than CDA* in search efficiency by adopting a tightly coupled approach.So the CWBA* is more suitable for the spacecraft system reconfiguration because of its fast response to faults.According to the engineering requirements of autonomous health management for spacecraft propulsion system,the functional requirements of an autonomous health management system are firstly analyzed.Then the hardware and software parts of the prototype system is designed,and the prototype hardware system is achieved under laboratory.Additionally,two core modules of the prototype system are developed and validated.For the fault detection and alarm module,verification results from the DFH-4 propulsion system simulations show that fault detection based on CS-SVDD-BC has well detection performance,and the prototype system has good stability and reliability.Based on the insight of modularization design,the configuration and reconfiguration module with extensible capacity is realized.Verification results from multiple faults of the DS-1 propulsion system show that the module can provide the optimal reconfiguration schemes meeting the requirements for fault control along with good solving performance and accuracy.Lastly,in order to solve the problems existed in the ground-based health management systems because the traditional telemetry data is extremely finite,the real-time recording system--APSDR-I is developed and achieved.The on-fire tests show that APSDR-I possesses good ability in the simultaneous recording of multiple parameters,and also the recording data is accurate.