Denoising jet engine gas path measurements using nonlinear filters

Traditionally, linear filters have been used to smooth time series of gas path measurements before performing fault detection and isolation. However, linear filters can smooth out sharp trend shifts in the signal and are also not good at removing outliers. Since most fault detection and isolation algorithms are optimized for Gaussian noise, they can show performance degradation when outliers are present. In this study, numerical results with simulated data for engine deterioration and abrupt fault show that the nonlinear rational filter with median preprocessor are useful for gas turbine health monitoring applications resulting in noise reduction of 73%-96% while preserving signal features and removing outliers.