In most output-only-based modal identification methods, the excitation and the measurement error are assumed to be white Gaussian noise. One of the obstacles appears once the excitation has periodic nature and measurements are contaminated by colored noise. In this paper, a time-domain filtering technique is introduced for harmonic excitation/noise elimination in operational modal analysis. The procedure is based on the singular value decomposition of the vibrational response. If excitation, noise, and modal harmonics have no intersection, they will have independent counterparts within the decomposed row space. Therefore, the decomposed components are partly selected and explored in a few iterations and appropriated to the transient and steady-state response and noise contaminations. Then, the components associated with the steady-state response and noise are removed, and the transient response is used for identifying the observability matrix and modal parameters. The performance of the proposed procedure is investigated through several numerical and experimental case studies of vibrating structures, and compared with other methods. The effectiveness of the technique for eliminating the excitation and noise harmonics from the modal parameters, and reconstructing the vibration signals is verified.

在大多数仅基于输出的模态识别方法中，假设激励和测量误差为高斯白噪声。一旦激发具有周期性并且测量被有色噪声污染，就会出现障碍之一。在本文中，在操作模态分析中引入了用于谐波激励/噪声消除的时域滤波技术。该过程基于振动响应的奇异值分解。如果激励、噪声和模态谐波没有交集，它们将在分解的行空间内具有独立的对应项。因此，分解的组件在几次迭代中被部分选择和探索，并适用于瞬态和稳态响应以及噪声污染。然后，与稳态响应和噪声相关的分量被去除，瞬态响应用于识别可观察性矩阵和模态参数。通过振动结构的几个数值和实验案例研究，并与其他方法进行比较，研究了所提出程序的性能。验证了该技术从模态参数中去除激励和噪声谐波并重建振动信号的有效性。