Session: 01-08-01: Vibration and Acoustic Measurements, Signal Processing, and Test Facilities

Paper Number: 68378

Start Time: Wednesday, 01:10 PM

68378 - A Comparative Study of Adaptive Mode Decomposition Methods for Modal Response Extraction 

Adaptive mode decomposition (AMD) methods have received significant interest in recent years as an effective means for analyzing signals of multi-components and high complexity. The methods are data-driven and posterior. They do not rely on a predefined basis as for conventional decomposition methods. They adapt to the transient, local characteristics of the signal, and extract the constituent oscillation modes of mono-component nature, making them appealing for analysis of nonlinear and nonstationary signals. On the other hand, modal information of a structure is vital to the analysis and understanding of its dynamic behavior. Many modal extraction techniques have been developed over the years either in the time or frequency domain. In many cases, the free response of a structure in the time domain can be easily obtained, which typically consists of contributions from multiple modes. By decomposing this global response into individual responses associated with each mode, the mode shape, natural frequency and damping can be estimated. This paper investigates the feasibility of integrating AMD methods and modal response extraction, and performs a comparative study of few representative AMD methods including the empirical mode decomposition (EMD), variational mode decomposition (VMD), empirical wavelet transform (EWT), adaptive local iterative filtering (ALIF), nonlinear mode decomposition (NMD), and local mean decomposition (LMD) methods. The fusion of AMD and modal analysis adds adaptivity and flexibility into data processing and helps automate the modal analysis process. The comparative study will provides insights on the advantages and disadvantages of the AMD methods as to the application of modal analysis. This is important because AMD methods can be considered as filtering algorithms that can be loosely grouped into two categories based on their capabilities for precise or rough filtering. As desired for modal analysis, the precise filtering aims to accurately extract mono-components from the raw signal with no or very little loss of the modal information; while the rough filtering targets to highlight particular characteristics of the signal, and some information loss can be tolerable. In this comparative study, the six representative AMD methods are first applied to the free response of a simulated three-degree-of-freedom (3-DOF) system, to extract the modal responses associated with the three modes. The decomposition results are compared to the analytical solution of the responses. Then, noise at various levels is added to the simulated signal to study the effect of noise on the decomposition performance. After that, the methods are applied to a measured free-response signal of a polymethyl methacrylate (PMMA) beam to assess their capability of analyzing real signals. Finally, the findings are summarized and conclusions are drawn.

Presenting Author: Yabin Liao Embry–Riddle Aeronautical University

Authors:

Yabin Liao Embry–Riddle Aeronautical University
Mark Sensmeier Embry–Riddle Aeronautical University