低频振荡模态参数辨识新算法

摘要

TLS-ESPRIT算法的实质是通过对信号子空间进行奇异值分解(SVD),辨识出信号的频率和阻尼比.SVD虽然可以抑制部分噪声影响,但噪声过强时就会影响TLS-ESPRIT算法的辨识精度;而且SVD易把较弱信号滤除,丢失真实模态而引入虚假模态.针对此结合数学形态学能够在时域内对信号进行滤波,且滤波后信号的几何特征保持不变的特点,将实测信号先经数学形态学滤波再通过TLS-ESPRIT进行模态辨识,可以最大限度地降低噪声的影响,使辨识精度更加精确.仿真结果验证了所提算法的可行性、有效性和强抗噪性.%The essence of TLS-ESPRIT algorithm is to identify the signal frequency and damping ratio by singular value decomposition (SVD) of signal subspace. Although noise can be suppressed by SVD, The identification accuracy of TLS-ESPRIT algorithm is also affected when the noise is too strong. Furthermore, SVD is prone to filter out weak information, lost real mode and introduce false mode. Mathematical morphology filter can filter the signal in time domain but maintain signal geometric characteristics. In the proposed algorithm, the measured signal is first filtered by morphological and then went through the TLS-ESPRIT mode identification. The method can minimize noise impact and increase identification accuracy. Simulation results verify the feasibility, effectiveness and strong noise resistance of the proposed algorithm.