Research On Typical Units Fault Diagnosis Technology Of Launch Vehicle Control System

The launch vehicle control system is the core delivery vehicle for launching satellites,space laboratories and space exploration.Its high reliability work is a prerequisite for the peaceful and effective development and utilization of space resource.Although the control system which is one of the key subsystems of launch vehicle,is usually designed with high reliability,failures are often caused by human and environment factors.Despite the failure online reconstruction technology with the accumulation and development of technology,most of launch failures in history are serious or catastrophic failures that cannot be reconstructed.If the failure is found in the ground test phase or the potential for success or failure is met,the high reliability of the flight phase can be guaranteed.The thesis takes the typical units of the typical control system of a certain type of launch vehicle as the research object.Firstly,from the perspective of test optimization design,in order to achieve higher test diagnosis efficiency with less test points or test cost,the test selecting optimization technology is studied,and then different fault diagnosis methods for the characteristics of different types of units are studied,which lays the foundation for system-level fault diagnosis through the fusion of different diagnosis methods.The main work of the thesis includes:Test optimization selection technology based on multi-signal model is studied.Under the constraints of equipment testing requirements and test indicators,optimize the test resource and obtain the best test indicators with the minimum test cost.Aiming at this problem,combined with the advantages of particle swarm optimization and genetic algorithm,an improved binary particle swarm genetic algorithm is proposed.The simulation analysis shows that the algorithm can converge to the optimal solution with a probability of 97.2%,and optimize tests combined with the multi-signal model of the integrated controller,the lowest test cost is obtained without changing the testability indicators.A deeply research on fault location technology based on multi-signal model.Aiming at the problem of fault location under credible test condition,two methods of information entropy based,fault diagnosis inference engine of TEAMS-RT are achieved and propose a method of test consistency and fault prior probability based.Simulation results show that fault diagnosis inference engine has higher accuracy and effectiveness,and is suitable for multi-fault location.Aiming at the problem of fault location under the condition of missed detection and false alarm,the location method based on the minimum posterior probability and the location method based on test consistency and fault prior probability are proposed.Simulation results show the latter is less effective but has higher location accuracy.Taking a typical electronic unit,integrated controller,as an example,after the test optimization,single fault and double fault location tests based on the test consistency and the fault prior probability are operated,it achieves a single failure 100%,double failure 86.16% isolation rate,and implements a hierarchical diagnosis strategy to improve diagnosis efficiency.Data-driven fault diagnosis technology is studied.Aiming at the dilemma of difficulty to establish an accurate model for electromechanical units,a data-driven diagnostic algorithm combining wavelet packet,information entropy and neural network theory is achieved.The algorithm firstly decomposes and reconstructs the wavelet packets,and then extracts the entropy of wavelet packets in order to constitute a fault feature vector,and finally classifies the fault models with neural network.The research work of the thesis combined with the typical electronic unit integrated controller and the typical electromechanical unit laser inertial measurement combination is to prove that the proposed methods are effective and feasible.