A Procedure for Automated Gas Turbine Blade Fault Identification Based on Spectral Pattern Analysis

Abstract

A method for diagnosing the existence and the kinds of faults in blades of a gas turbine compressor is presented in the present paper. The innovative feature of this method is that it performs the diagnosis automatically, that is, it gives a direct answer to whether a fault exists and what type of fault it is, without requiring the interpretation of results by a human expert. This is achieved by derivation of the values of discriminants calculated from spectral patterns of fast response measurement data. A decision about the corresponding engine status is then derived according to the values of the discriminants. In the paper, the procedure of examining the suitability of particular parameter discriminants and the constitution of a related knowledge base is described. The derivation of decisions by a computer, and on what engine condition a particular measurement data set corresponds, are then described.