Research On Sensor Data Validation For Aero-Engine Vibration Testing

The research of sensor data validation method for aero-engine vibration testing includes two parts. The first is determining the validation of the vibration signals; The second is detecting the state of the sensors and diagnosing the faults if the data exists abnormal phenomenon.This paper researches the application of the time series prediction based on AR model, and introduces the signal processing method which is named Empirical Mode Decomposition and Hilbert transform in the field of aero-engine vibration signal analisis creatively, then combining both of them, proposes a kind of sensor validation method based on time series prediction and EMD, and proves its validation with simulation. This method can produce the remanet vector between the normal and abnormal data, which can be used to judge the validation of the sensor's output signal and determine the failure grade of them. What's more, this method can also reestablish the signal, gets rid of the fault component and rescovers the valid information. At last this paper applies the method to the signal analysis of the twin-spool torbo-fan engine vibration test, focusing on the data abnormity, validates the data, and diagnoses the fault of the sensor.The analysis results prove that the time series prediction based on AR model has a superior skill in validating the vibration signal of the aero-engine, the data model is reasonable, the precision is satisfying; The signal processing method based on Empirical Mode Decomposition and Hilbert transform could zoom the output signal of sensors using multi-resolution analysis, diagnose the sensor's faults that is usually neglected by the traditional method. The sensor data validation method constructed basing on both of them has some merits that are accuracy, intelligence and visualization, and it is of good value in the research field.