摘要
针对传统l_(1)正则化方法在叶片振动参数识别中幅值估计偏差较大的问题,提出一种基于非凸稀疏正则化的叶端定时数据处理方法。结合叶片振动机理与压缩感知理论,建立叶端定时欠采样信号压缩感知模型;通过引入参数可分离的非凸反正切正则项提升信号稀疏性,减少幅值估计偏差,同时利用加速广义阈值迭代算法提高计算效率。高速旋转试验结果表明,该方法与应变片结果对比频率辨识误差小于0.5%;与经典的1正则化相比,在叶片共振幅值的辨识精度方面具有显著优势,能够改善幅值低估问题,为叶片运行安全评估与健康监测提供了有效技术支持。
Aiming at the problem of large deviation of blade amplitude estimation in blade vibration parameter identification by the traditional ℓ_(1) regularization method,a blade tip timing data processing method based on non-convex sparse regularization was proposed.Combining the blade vibration mechanism and compression sensing theory,a compression sensing model of blade tip timing undersampled signal was established to improve the signal sparsity and reduce the amplitude estimation bias by introducing the parameter separable non-convex arctangent regular term,and the accelerated generalized thresholding iteration algorithm was used to improve the computational efficiency.Results of the high-speed rotation test show that the frequency identification error of this method is less than 0.5%compared with the strain gauge results,and compared with the classicalℓ1 regularization,it has a significant advantage in the identification accuracy of blade resonance amplitude,which can improve the amplitude underestimation,and provide effective technical support for the blade operation safety assessment and health monitoring.
作者
乔百杰
张松林
周凯
刘美茹
梁俊
陈雪峰
QIAO Baijie;ZHANG Songlin;ZHOU Kai;LIU Meiru;LIANG Jun;CHEN Xuefeng(Institute of Aero Engine,School of Mechanical Engineering,Xi’an Jiaotong University,Xi’an 710049,China;AECC Sichuan Gas Turbine Establishment,Chengdu 610500,China)
出处
《燃气涡轮试验与研究》
2025年第3期1-8,共8页
Gas Turbine Experiment and Research
基金
国家自然科学基金(52475130,52305127)
国家财政稳定支撑项目(GJCZ-0000-2024-0015)。
关键词
转子叶片
叶端定时
非接触式测量
参数辨识
欠采样
非凸正则化
航空发动机
rotor blades
blade tip timing
non-contact measurement
parameter identification
undersampling
non-convex regularization
aero-engine
