The filtering requirements of engine measurement signals are analyzed.A center weighted median filter and fast algorithm for sensor data validation are presented

and the filtering function and algorithmic efficiency are evaluated with numerical experiments.The simulated test signals are composed of steep edges and slopes that representing abrupt faults

linear deterioration and/or regime transitional response

superimposed with simulated random noise and high-amplitude impulsive noise.Results show that median filter is good at removing impulsive noise and outliers.In addition

random noise is suppressed to great extents while preserving the final details.Practical application to gas turbine experimental sensor data validation confirms the above conclusions.