Sparse component analysis using time-frequency representations for operational modal analysis

Area of Science:

• Structural Dynamics
• Signal Processing
• Mechanical Engineering

Background:

• Sparse Component Analysis (SCA) is a key technique for Blind Source Separation (BSS).
• SCA has recently been adapted for Operational Modal Analysis (OMA), also known as output-only modal identification.
• Existing methods may face challenges when the number of active modes exceeds the number of sensors.

Purpose of the Study:

• To propose a new time-frequency (TF) domain SCA method within the OMA framework.
• To leverage the sparsity of source signals in the TF domain for improved modal identification.
• To address limitations of current OMA techniques, particularly when dealing with a high number of active modes.

Main Methods:

• Transforming sensor measurements into the TF domain to achieve a sparse representation.
• Detecting single-source-points (SSPs) to identify hyperlines corresponding to the mixing matrix columns.
• Employing K-hyperline clustering to determine hyperline direction vectors and compute the mixing matrix.
• Utilizing basis pursuit de-noising to recover modal responses for parameter computation.

Main Results:

• The proposed TF-domain SCA method successfully identifies modal parameters.
• The method demonstrates robustness even when the number of active modes surpasses the number of sensors.
• Numerical simulations and experimental data confirm the effectiveness and good performance of the approach.

Conclusions:

• The developed TF-domain SCA offers a powerful tool for output-only modal identification.
• This approach enhances modal analysis capabilities, especially in complex scenarios with numerous active modes.
• The study validates the practical applicability and accuracy of the proposed method through simulations and experiments.