Fault Detection and Identification in an IM270 Gas Turbine Using Measurements for Engine Control

Abstract

A unique fault detection and identification algorithm using measurements for engine control use is presented. The algorithm detects an engine fault and identifies the associated component, using a gas path analysis technique with a detailed nonlinear engine model. The algorithm is intended to detect step-like changes in component performance rather than gradual change of all components. Since simultaneous multiple faults are unlikely, a single component fault is assumed, which reduces the number of unknown parameters to less than two. By setting the number of adjustable parameters to that of the available measurements, the parameters are computed using an engine model. After computing all of the six possible combinations of adjustable parameters, the average magnitude of the parameter deviation vectors is used to detect an engine fault. Component performance deviation (efficiency and flow rate) is represented by a magnitude and a phase. The phase is selected to minimize the error of matrices consisting of normalized adjustable parameter deviation vectors. Then the magnitude is computed by the average magnitude ratio of the vectors. Since the algorithm is simple, it is easily applied to newly developed engines. A fault detection and identification program was specifically developed for IM270 engine, a single shaft gas turbine with 2MW output capacity. By utilizing operational data obtained at a remote monitoring center, the algorithm was able to quantitatively identify the compressor and the turbine performance deviation. Although the algorithm correctly identifies the turbine as the faulty component, there remains some ambiguity. Analysis of linear dependency of the measurement deviation vectors shows that identification capability varies with phase. There are several phases where identification is impossible in the current IM270 sensor system.