Dissertation > Aviation, aerospace > Aerospace ( Astronauts ) > Propulsion system ( engine, propeller ) > Engine control systems and devices > Engine test

The Research of the Fault Diagnoses Algorithm for the Liquid Rocket Engine Testing Bed Based on PCA-SVM

Author GuoXiaoFeng
Tutor WangQi
School Harbin Institute of Technology
Course Instrument Science and Technology
Keywords fault diagnoses PCA SVM
CLC V433.9
Type Master's thesis
Year 2008
Downloads 101
Quotes 0
Download Dissertation

With the development of the spaceflight in China, the task of the rocket engine ground testing is very heavy, so how to ensure the security and reliability of the rocket engine testing is great important. And the reliability of rocket ground testing bed is the precondition of the ground test success. In this paper, based on the testing bed of Beijing institute of aerospace testing technology, the fault diagnosis algorithm of the liquid propellant rocket engine ground testing bed is researched.The testing bed is mainly composed of liquid hydrogen subsystem and liquid oxygen subsystem which is the power source of engine. Each subsystem is divided into super-charging subsystem, providing subsystem, discharge subsystem and blowoff subsystem. In this paper, principal component analysis (PCA) is used to extract the fault feature and reduce the dimension of the data, and support vector machine (SVM) is used to design the classifier, which is implemented by bintree based on clustering.The paper is composed of four parts: fault mode analysis, feature extraction by PCA, classifier design by SVM, experiments and summing-up. First, fault mode analysis is the foundation of whole fault diagnosis system. Two methods are used to acquire it: They are fault statistics, failure mode and effects analysis (FMEA). Then, the detailed fault mode results of system are obtained. Second, the theory of PCA and SVM, implementation and application in the paper are researched in detail. And the bintree based on clustering is proposed to design the fault classifier. At last, experiments are done to verify the algorithms and the whole paper is summarized up. Meanwhile, the comparison result with BP neural networks shows that the proposed algorithm has better diagnosis results, also, the probability of enhancing fault diagnosis rate is analyzed.Based on the characteristic of the system, the paper combines PCA and SVM, gains excellent fault diagnosis results, and has instructional meaning both in theory and practice.

Related Dissertations
More Dissertations