Study Of Fault Detection And Diagnosis Methods At Start-Stage Of Liquid-Propellant Rocket Engines

In order to solve the difficulties caused by the large fluctuation of parameters in fault detection and diagnosis at start-stage of liquid-propellant rocket engine (LRE), the cloud model theory based fault detection and diagnosis methods for LRE were chiefly studied in this thesis.First of all, the cloud-model-theory was deeply studied. And then, methods for LRE fault detection and diagnosis that respectively based on cloud model association rules, cloud model neural network and cloud classifier were discussed in detail, the contents covered basic theory, fault diagnosis strategies and example analysis.Three methods were mentioned in this degree paper. By combining cloud model with traditional association rule technology, the definition and mining algorithms of cloud model association rules were given, and then, the LRE fault detection strategies based on cloud model association rules were discussed seriously as well as a detection example. A fuzzy system based on cloud model theory and neural network was considered next. The system inherited the advantage of traditional fuzzy system but without the weakness. The usage of the cloud-model and neural work based fuzzy system in LRE fault diagnosis was given next, as well as a simulation. In the last part of this dissertation, a new classifier based on cloud model was developed; classification rule extraction algorithm and the application in LRE fault diagnosis were carried out.The results reveal that the cloud model theory based fault detection and diagnosis methods for LRE start-stage have good performance in reliability and stability as well as solving fuzzy and stochastic problems. Moreover, the methods given in this thesis can predict faults early than the traditional methods such as RS (Red Line System), EA (Envelop Algorithm) and neural networks algorithm in some conditions.