Aero-engine rotor-stator rubbing position identification based on casing velocity signal

Abstract

A rotor-stator rubbing position identification method based on casing velocity signal is proposed. Considering that velocity is an ideal parameter to reflect vibration, and its effective value is a standard to the measuring of vibration fault diagnosis in the world, firstly, the rotor experiment rig of aero-engine was used to simulate rubbing faults in different rubbing positions, and casing vibration acceleration signal was collected and changed to velocity signal through integral and polynomial least square fitting method. Secondly, low-frequency normalized energy characteristics of velocity signal and normalized mean-square value characteristics of acceleration signal were extracted; finally, normalized characteristic parameters including energy and mean-square value were input to nearest neighbor classifier and support vector machine(SVM) to identify the different rubbing positions. The results show that low-frequency energy characteristics of velocity signal can effectively identify the rotor-stator rubbing positions of aero-engine, and reach to 93 % of recognition rate based on nearest neighbor classification method and 98 % based on SVM, while mean-square value characteristics of acceleration signal recognition rate can only reach 81 % based on nearest neighbor algorithm and 85 % based on SVM.