Jet engine gas-path measurement filtering using center weighted idempotent median filters

Key indicators of jet engine health are deviations in gas-path sensor measurements from a "good" baseline engine. These measurement deviations or deltas are used to detect and estimate engine deterioration and faults. Typical measurements are exhaust gas temperature, low rotor speed, high rotor speed; and fuel flow. The measurement deltas are displayed to powerplant engineers through computer visualization tools such as trend plots, which are then used for diagnostic and prognostic decisions. The measurement deltas are also used in pattern recognition and state estimation algorithms to detect and isolate faults. Both the fault detection and the visualization process are hindered by the presence of noise in the data. Traditional linear filters used by the gas turbine industry for smoothing gas-path measurement deltas tend to smooth out the trend shifts in the signal that can signify a fault or repair event. The linear filters also perform poorly when high-amplitude impulsive noise and outliers are present in the signal. However, nonlinear filters can be designed to suppress noise while preserving the final detail in the gas turbine measurements. Results with simulated signals show noise reduction of about 60% can be obtained with a nonlinear center weighted idempotent median filter.