基于叶尖间隙时频特征融合的转子碰摩诊断方法  

Rotor rubbing diagnosis method based on fusion of time-frequency features of blade tip clearance

作　　者：谷昭鹏 王维民[1,2] 刘炳成 米珂嘉 GU Zhaopeng;WAGN Weimin;LIU Bingcheng;MI Kejia(State Key Lab of High-end Compressor and System Technology,Beijing University of Chemical Technology,Beijing 100029,China;Beijing Municipal Key Lab of Health Monitoring and Self-recovery for High-end Mechanical Equipment,Beijing University of Chemical Technology,Beijing 100029,China)

机构地区：[1]北京化工大学高端压缩机及系统技术全国重点实验室,北京100029 [2]北京化工大学高端机械装备健康监控与自愈化北京市重点实验室,北京100029

出　　处：《振动与冲击》2024年第23期94-101,共8页Journal of Vibration and Shock

基　　金：国家自然科学基金重大研究计划重点项目(92160203,92360306)。

摘　　要：以航空发动机与船舰燃气轮机为代表的高端透平机械在高效率,大机动飞行等工况要求下,内部结构趋于紧凑化和复杂化,使得内部转静子发生碰摩的概率也随之增加,由于其内部少测点、难监测,往往难以及时准确地发现碰摩故障。提出了一种叶尖间隙时频特征融合的转子碰摩故障诊断方法,首先通过叶尖振动-间隙复合传感器获取叶尖间隙(blade tip clearance,BTC)序列,然后通过机器学习算法评估BTC序列的异常性;若有异常警报,则对异常信号进行傅里叶分解并汇总各分量的时频能量谱;最后通过时频能量谱中的碰摩带判断故障的发生。试验结果表明,在对转子施加短时、持续碰摩两种工况后,机器学习模型能够实时诊断,且报警后所提方法能完整分辨出两种碰摩的发生,以及持续碰摩缓慢施加、快速结束的过程。通过与转子振动信号进行对比,表明该方法可通过机匣测点获取叶尖间隙并感知转子碰摩故障,实现了“一传多感”,为透平机械的状态监测和健康管理提供了新的方法。High end turbomachinery represented by aero-engine and ship gas turbine tends to have a more compact and complex internal structure under high efficiency and large maneuverability flight conditions,and cause an increasing probability of internal rotor-stator rubbing.Due to the lack of inner measured points and difficulty in monitoring,it is often difficult to detect rubbing faults in a timely and accurate manner.Here,a rotor rubbing fault diagnosis method was proposed based on time-frequency features fusion of blade tip clearance.Firstly,blade tip clearance(BTC)sequence was obtained using blade tip vibration-clearance composite sensor,and then the abnormality of BTC sequence was evaluated with machine learning algorithm.If abnormal alarm appearing,Fourier decomposition was performed for abnormal signal and time-frequency energy spectra of various components were summarized.Finally,the occurrence of faults was determined with rubbing bands in time-frequency energy spectra.The experimental results showed that after exerting short-term and continuous rubbing conditions to rotor,the machine learning model can conduct fault diagnosis in real time,and the proposed method can fully distinguish the occurrence of the two rubbing conditions after alarm and the process of continuous rubbing slowly exerting and fast ending.Compared with the rotor vibration signal,it was shown that the proposed method can obtain BTC through engine case measured point,and sense rotor rubbing faults to realize“one sensor multiple messages”and provide a new method for condition monitoring and health management of turbomachinery.

关 键 词：叶尖感知 碰摩故障 机器学习 傅里叶分解 叶尖间隙(BTC) 

分 类 号：TH113.1[机械工程—机械设计及理论]