Data-driven methods for operational modal parameters identification: A comparison and application. (January 2019)

• Record Type: Journal Article
• Authors: Guan, Wei; Dong, L.L.; Zhou, J.M.; Han, Yi; Zhou, J.
• Is Part Of: Measurement. Volume 132(2019)
• Page Start: 238
• Page End: 251
• Publication Date: 2019-01
• Subjects: Operational modal identification -- Structural dynamics -- Data-driven methods -- Blind source separation -- Manifold learning; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2018.09.052
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: Comparison and application of data-driven methods for operational modal analysis. Damping ratios extraction using Hilbert Transform and Random Decrement Technique. Proposed methods are validated by a discrete system and a continuous steel plate. Dynamic parameters (modal shapes, frequencies, damping ratios) are identified. Present the results of research from the field of signal theory and engineering. Abstract: The operational modal parameters identification under ambient excitation has been widely applied in structural dynamics analysis, where the input force is difficult to measure. This paper makes comprehensive and systematic comparisons of four statistical learning algorithms (PCA, ICA, SOBI, LLE) on analyzing their performance for resolving operational modal parameters identification. It mainly focuses on the comparison and evaluation of the four data-driven methods based for operational modal analysis and explores the use of Hilbert Transform and Random Decrement Technique for the identification of modal damping ratios. The further tests, the robustness of four algorithms, are conducted for investigating the influences of external noise to the performance of the algorithms. On the basis of the modal expansion theory, the vibration response signals could be decomposed into the inner product of modal shapes matrix and modal responses vector in the modal coordinate, from which the modal shapes, modal natural frequencies and modal damping ratios can be well estimated. Hence, a one-to-one mapping between the mathematical model of four statistical learning algorithms and physical model of dynamic systems is established. To validate the effectiveness of the proposed methods, performance and comparison of the proposed methods are studied using a discrete three degree-of-freedom (DOF) system and continuous cantilever steel plate. The simulation and experimental results show that, for the same experimental data, the SOBI algorithm has better performance than other algorithms, and it is more suitable for operational modal identification. Finally, the characteristics of four algorithms and further directions are concluded and discussed in this paper.