Measurement Selection Method for Aero-engine Discrete Operating Conditions Gas Path Analysis

With Aero-engine fault detection and isolation (FDI), engineers can ensure engine works in a healthy and reliable condition. Gas path analysis (GPA) is widely used to conduct FDI. It is a data-driven method for the state-of-health diagnosis of aero-engine. However, the number of available parameters in GPA is usually limited due to the unavailability of additional sensors in the aero-engine. Considering this limitation, discrete operating conditions gas path analysis (DOCGPA) is introduced. It makes the GPA with fewer measurements possible without compromising diagnostic accuracy. The effectiveness of the DOCGPA is greatly influenced by the selection of measurements and operating points. However, no existing work investigates the systematic and quantitative way for measurement and operating points selection. Therefore, we propose a framework based on correlation analysis and a two-stage genetic algorithm to the help us select measurements and operating points. The correlation analysis identifies the similar faults from influence coefficients matrix and narrows down the range of meaningful measurements. A low-resolution genetic algorithm is carried out to find the combinations of measurements and operating points of relatively low predicted residual error sum of squares. Lastly, high-solution genetic algorithm combined with regularized least squares (RLS) is employed to refine the selection by searching the neighborhood of the previous results. Through the experiment, we find a set of four measurement parameters that could achieve a significant diagnostic accuracy.