A sequential model-based approach for gas turbine performance diagnostics

Highlights: • A novel sequential approach is proposed for gas turbine performance diagnostics. • The new method can successfully detect faults using small number of measurements. • The engine model is developed in Visual Studio C# and validated against GasTurb. • Computational efficiency is improved by reducing the iterative matrix dimensions. • The proposed method is superior to the conventional GPA method. The gradual degradation of gas turbine components is an inevitable result of engine operation, impacting engine availability, reliability, and operating cost. Gas path analysis plays an essential role in engine fault diagnosis. Accurate and fast diagnosis of multiple simultaneously degraded components has always posed a challenge, especially when the number of available measurements is limited. This paper proposes a novel performance diagnostic method that partitions the engine diagnosis into a series of steps to remove the "smearing effect" and reduce the matrix dimensions in the iterative diagnostic algorithm. An engine performance model of a triple-shaft gas turbine has been developed and validated against commercial software, in order to assess the accuracy and computational performance of the proposed method. The advantage of the proposed method lies in its capability to detect the severity of engine component degradation, such as compressor fouling and turbine erosion, with greater accuracy and computational efficiency than other model-based methods that use the same number of measurements. The newly developed method provides an accurate diagnosis with a reduced set of measurements. The method can deal effectively with the presence of random noise in the measurements and carries a significantly lower computation burden in comparison to existing methods. The proposed method could be used as a tool for supporting condition monitoring systems for improved gas turbine reliability and energy efficiency.