Research About Rub Feature Analysis And Identification Technique Of Aero-Engine Based On Casing Response

To pursue the high thrust weight ratio and low oil consumption rate of modern aero-engine, the rotating speed is becoming higher and higher and the various kinds of clearance are becoming smaller and smaller, so the rotor-stator rubbing occurs more frequently at work. Accurate and timely diagnosis of rubbing fault and determining rubbing positions, which can not only avoid serious accidents caused by rubbing fault but also discover the rubbing causes and improve the design to increase working security of aero-engine, are of great significant engineering application value. In view of this, the main research contents in this paper are as follows:(1) An aero-engine rotor test rig with casing is used to simulate rubbing faults whose rubbing positions are single-point and local partial at different rotating speed. The turbine casing vibration acceleration signals and casing strain signals are measured, and a large number of experiment data of different rubbing positions are obtained, in which the single-point rubbing positions include the upper, lower, left and right on turbine casing and local partial rubbing positions include left and right. The experiment results obtained from thin-walled and thick-walled casing are compared, which provide a large number of experiment date for the research in this paper.(2) The identification method for aero-engine rotor-stator rubbing positions based on casing acceleration signals is proposed. Considering that the vibration states are generally monitored by acceleration sensors in modern aero-engine, the projects proposed in this paper are as follows:1) mean square features of casing acceleration signals are extracted directly; 2) the normalized energy characteristics of acceleration signals are extracted by means of wavelet packet analysis; 3) the wavelet partial maximum modulu energy features of acceleration signals are extracted by means of wavelet maximum modulu energy. Then the normalized energy features extracted from the three methods are input into support vector machine(SVM) in order to identify the rotor-stator rubbing positions. The average recognition rate of the three methods applied in training samples and test samples all reach 99% and in unknown sample, the average recognition rates reach above 65%, 85% and 98%.(3) Considering the thin-walled structure adopted in modern aero-engine and the deformation caused by rubbing force, and the strain measuring technic is of high sensitivity, mature, simple and reliable, the identification method for aero-engine rotor-stator rubbing positions based on casing strain signals is proposed. The strain mean features of different rubbing positions and different measuring points are analyzed. And the analysis indicates that the strain value change rules of each measuring point in the same rubbing positions are the same. Finally the strain mean value of each measuring point is extracted as the characteristics parameters, which are used to identify the rubbing positions by SVM. The recognition rate of the method can reach 100% for training samples, test samples and unknown samples.(4) The identification method for aero-engine rotor-stator rubbing positions based on cepstrum analysis is proposed. By using the “generalization” ability of the cepstrum for side-frequency, the vibration acceleration signals of casing measuring points are directly analyzed by cepstrum, the transfer characteristics which reflect the rubbing positions are isolated. Furthermore, the features for rubbing position identification are extracted. By using SVM to identify the rubbing positions, the recognition rate can reach 100% for training samples, test samples and unknown samples.(5) By using the software of Weka, data mining for sample gallery composed of features extracted from rubbing position identification based on casing strain signals and acceleration signals is processed. The diagnosis rules which are easy to understand and explain are extracted and excavated, and it provides ways for engineering application of rubbing position indetification method.(6) The features of casing measuring points acceleration signals based on different rotating speed, different casing thickness and different rubbing positions are analyzed. It can be found that the impact frequency and its frequency doubling of the blades through casing are caused by single-point rubbing, and the size of the impact is modulated by unbalanced force. Therefore, at the blade passing frequency and its frequency doubling, the modulation sidebands whose side frequency is rotation frequency occur. In cepstrum, the quefrency components of rotating frequency and its doubling frequency is observed obviously. Finally, the actual aero-engine test data is used to verify the rubbing feature.(7) Feature analysis for the acceleration signals on different rubbing conditions is carried out by using the Hilbert spectrum envelope. Acceleration signals on normal operation, single-point rubbing condition and partial rubbing condition on aero-engine test rig are compared by means of the Hilbert transform. The effect of casing thickness on the acceleration features at low frequency is studied by changing the thickness of casing.