Abstract:

Two algorithms for aero engine fault diagnosis based on principal component analysis are presented. The principal component estimate algorithm(PCEA) and the principal component dimension reduction algorithm (PDRA). These algorithms can resolve efficiently the two difficult problems in aero engine fault diagnosis, e.i., reduce the number of fault factors in the fault equation, and improve its abnomal condition. The concept of optimum principal component number and its determination for the PCEA are presented in the paper, which resolves the difficult poblem of selecting the number of principal components. The application of the given algorithms is demonstrated for JT9D engine diagnosis. Investigation shows that the 26 fault factors of JT9D engine can be reduced to 9, or the 10 fault factors of 5 engine modules can be reduced to 5 combined variables, and satisfactory results can be obtained by use of these algorithms in the case that serious multi collinearity exists in the fault equation and no constrained conditions are available. As a middle link of Primary Factor Model and Random Search Model presented by the authors, these algorithms can be used in the case that the fault pattern number is greater than the measurement number.

Key words: aer o engine, fault diagnosis, principal component analysis, opt imization