Study On Key Techniques Of Fault Detection And Diagnosis For New Generation Large-scale Liquid-propellant Rocket Engines

Some key technologies involved in fault diagnosis for newly large-scale Liquid-propellant Rocket Engines(LRE) are thoroughly investigated, analyzed and developed in the dissertation. The research works related to modularization fault simulation, fault detection for startup and steady process, fault detection and isolation for gas/liquid pipeline with abrupt or incipient fault, fault isolation in components level based on fuzzy model and testoff analysis method, the integrated design, implementation and validation of the state judgement system for rocket engine adjust test on ground. The research results can not only provide solid theoretical and application foundation for developing health-monitoring system but also provide significant reference for improving reliability and safety on future reusable liquid-propellant rocket engines.Due to the lack of fault sample data, it is difficult to obtain the feature of fault modes. So modularized simulation models of faults are founded and a fault simulation software system is developed. Comparison results show that the simulation results are in good agreement with real test data. It can provide important fault sample data in the research of fault detection and diagnosis methods for the engine.Two real-time fault detection method based on the fuzzy clustering for LRE startup process is proposed to deal with nonlinear process and intensively fluctuant parameters. The adaptive threshold algorithm(ATA) and the adaptive correlation algorithm(ACA) are developed and improved for the steady process of LRE. The results validated by test data of LRE show that all these algorithms can detect faults effectively in real-time.The adaptive threshold algorithm based on the characteristic of pressure difference proportion analysis was designed and realized in order to locate the block and leak out fault in pipeline, throttle circle or nozzle. The algorithm is verified by a great lot of test data from three-type engines. Results show that the algorithm could location the fault parts or components such as pipeline, valve, throttle circle and nozzle quickly and exactly when single abrupt fault occurs at them, and there is no false alarm for the normal components. However, the adaptive threshold algorithm is insensitivity for incipient faults. Therefore, a statistic threshold detection algorithm is also designed and realized for the slope of fault characteristic parameters. This algorithm is also validated by firing-test data. Results show that the statistic threshold algorithm could detect the incipient and abrupt fault clearly to locate the fault, and no false alarm for normal test data.A fault isolation method based on fuzzy model for component level is proposed to confirm the fault location in LRE by depending on the analysis of fault isolability and the monitored parameters for components. The fault isolation algorithm based on fuzzy model is validated by simulation data and firing-test data. The results show that the algorithm could isolate effectively the fault that its parameters change obviously. The further detection for special parameters is necessary to distinguish the faulty components whose performance parameters change unobviously.In order to evaluate the difference between the specifically engine performance and the design object, a test-off state evaluating method based on error analysis is investigated. This method is validated by data of a certain engine and the practical quality is obtained. The results can provide proof for the engine post evaluation.An applicable fault detection and diagnosis system for large-scale LRE on ground test is built and implemented based on integration by the developed theory and method. The fault detection and diagnosis system is validated by the ground firing-test and history firing-test data. The results show that the real-time fault detection and alarm subsystem can monitor the stage of engine in laboratory and ground test, and without false alarm and no missing alarm. The post-test analysis subsystem can diagnose and analyse the engine comprehensively based on integration manifold algorithm. The post-test analysis sunsystem is significant for ensuring the reliability and security to achieve the target in launch vehicle. It’s a powerfull support for implementing deliver directly and undisassembling after adjustment of ground firing-test in flight product.