Enhanced Subspace Dynamic Mode Decomposition for Operational Modal Analysis of Aerospace Structures

Hao Zheng, Rui Zhu, Yanbin LiAerospace 2025 · DOI

TL;DR

This paper introduces an "Enhanced Subspace Dynamic Mode Decomposition" (ES-DMD) method to improve the accuracy of operational modal analysis (OMA) for aerospace structures, especially in the presence of noisy data. By combining subspace orthogonal projection for denoising with DMD for dimensionality reduction, the method achieves more precise identification of modal parameters like frequency and damping ratio.

What Problem Does This Paper Address?

Accurately identifying the vibration characteristics (modal parameters: natural frequency, mode shape, and damping ratio) of aerospace structures is crucial for their design and safety. However, this process is challenging due to the inherent noise in vibration measurements, which can significantly degrade the performance of existing modal analysis techniques. Traditional methods like Dynamic Mode Decomposition (DMD), while powerful for extracting dynamic features, struggle with noisy data, leading to inaccurate modal parameter estimations. This paper aims to bridge this gap by developing a more robust OMA method that can effectively handle noisy operational data.

Methods

The core of the proposed method, ES-DMD, lies in a three-step process:

  1. Denoising with Subspace Orthogonal Projection: The raw vibration signals are first organized into a Hankel matrix. Subspace orthogonal projection is then applied to this Hankelized matrix. This step is designed to filter out noise while preserving the essential dynamic information of the system.
  2. Optimal System Matrix Realization via SVD: A compact Singular Value Decomposition (SVD) is performed on the projection matrix obtained in the previous step. This allows for the extraction of an optimal, lower-dimensional representation of the system's dynamics.
  3. Dimensionality Reduction and Modal Feature Extraction with DMD: The standard Dynamic Mode Decomposition (DMD) is then applied to this reduced-order system matrix. DMD effectively identifies the dominant modes of the system, allowing for the extraction of modal frequencies and damping ratios.

This approach leverages the strengths of subspace methods for noise reduction and DMD for modal identification, creating a synergistic effect.

Datasets

The paper evaluates the proposed ES-DMD method using two primary sources of data:

  • Numerical Simulation: A classical wing model (HA145G) was used for numerical simulations. This model was modified with direct damping ratios and subjected to external forcing inputs to generate time-domain response data. The simulation data was intentionally corrupted with uncorrelated discrete white noise signals at various levels to test the method's robustness.
  • Experimental Verification: A metal beam vibration experiment was conducted. The experimental setup involved a vibration loading system, a high-precision measurement system (laser displacement meter), and a data acquisition system. The beam structure was subjected to stochastic excitation, and its displacement response was recorded. This real-world data provided a practical test case for the ES-DMD method.

Key Insights and Ideas

  • Synergy of Projection and DMD: The paper demonstrates that combining subspace orthogonal projection for denoising with DMD for modal analysis is more effective than using either method alone, especially in noisy environments.
  • Hankel Matrix for Temporal Structure: The use of a Hankel matrix in the initial denoising step is crucial for capturing the temporal dependencies and structure of the vibration signals.
  • SVD for Optimal Representation: Employing SVD on the projected matrix helps in obtaining a clean and optimal representation of the underlying system dynamics, which is vital for accurate DMD application.
  • Improved Damping Ratio Estimation: A significant finding is the enhanced performance of ES-DMD in accurately identifying damping ratios, an area where standard DMD often struggles, particularly with noisy data.

Key Results

The ES-DMD method was compared against two established OMA techniques: Stochastic Subspace Identification (SSI-DATA) and standard Dynamic Mode Decomposition (DMD).

  • Numerical Simulations:
    • ES-DMD showed marginally improved identification accuracy for modal frequencies compared to SSI-DATA and DMD.
  • It demonstrated significantly enhanced performance in identifying damping ratios, with much smaller errors than DMD, and was marginally superior to SSI-DATA.
  • The method maintained relatively low errors for modal frequencies even under increasing levels of measurement noise.
  • While damping ratio errors increased with noise for ES-DMD and SSI-DATA, ES-DMD consistently showed superior accuracy.
  • Experimental Verification:
    • The identified modal frequencies and mode shapes from the experimental data showed good consistency with theoretical values for all three methods.
  • The damping ratios identified by ES-DMD and SSI-DATA were in good agreement, while DMD showed limited accuracy.
  • The peak information in the power spectral density (PSD) curve of the experimental response was in excellent consistency with the stable modes extracted by ES-DMD.

Limitations and Future Work

  • Redundant Modes: The authors acknowledge that, like other OMA methods, ES-DMD can extract redundant false modes as the model order increases. Further research is needed to address this.
  • Nonlinear Systems: DMD is fundamentally designed for nonlinear systems. While applied here to linear mechanical systems, further investigation is recommended to explore the integration of DMD for nonlinear modal parameter identification.
  • Noisy Experimental Data: Despite improvements, the paper notes that accurately identifying modal parameters from noisy experimental datasets remains a challenge. Future work could focus on further enhancing robustness against complex noise scenarios.
  • Mode Shape Accuracy: The paper mentions that the MAC criterion for mode shape was ignored due to only having one displacement measurement in the experimental setup. Future studies with multi-point measurements could provide a more comprehensive evaluation of mode shape identification.

Summary updated

Abstract

To address the issue of low accuracy in the dynamic modal decomposition (DMD) method used for operational modal analysis (OMA) under noise conditions of aerospace structures, an enhanced identification approach is proposed in this paper, which integrates subspace orthogonal projection with DMD to better determine the modal properties of linear mechanical systems with noisy observations. Subspace orthogonal projection applied to the Hankelized matrix is utilized for denoising observation signals. Compact singular value decomposition (SVD) is employed on the projection matrix in order to acquire the optimal realization of system matrix. Subsequently, DMD is introduced to reduce the dimensionality of the system matrix and extract the dominant modal features. The effectiveness and practicality of the proposed method are confirmed through numerical and experimental examples. The proposed method demonstrates marginally improved identification accuracy in modal frequency and enhanced performance in damping ratios when compared to representative OMA methods under different white noise conditions.
BibTeX
@inproceedings{zheng2025enhanced,
  author = {Hao Zheng and Rui Zhu and Yanbin Li},
  title = {{Enhanced Subspace Dynamic Mode Decomposition for Operational Modal Analysis of Aerospace Structures}},
  booktitle = {Aerospace},
  year = {2025},
  doi = {10.3390/aerospace12090776},
}
Figure