High-resolution modal analysis

Usual modal analysis techniques are based on the Fourier transform. Due to the Delta T . Delta f limitation, they perform poorly when the modal overlap mu exceeds 30%. A technique based on a high-resolution analysis algorithm and an order-detection method is presented here, with the aim of filling the gap between the low- and the high-frequency domains (30%<mu<100%). A pseudo-impulse force is applied at points of interests of a structure and the response is measured at a given point. For each pair of measurements, the impulse response of the structure is retrieved by deconvolving the pseudo-impulse force and filtering the response with the result. Following conditioning treatments, the reconstructed impulse response is analysed in different frequency-bands. In each frequency-band, the number of modes is evaluated, the frequencies and damping factors are estimated, and the complex amplitudes are finally extracted. As examples of application, the separation of the twin modes of a square plate and the partial modal analyses of aluminium plates up to a modal overlap of 70% are presented. Results measured with this new method and those calculated with an improved Rayleigh method match closely.

💡 Research Summary

The paper addresses a fundamental limitation of conventional modal analysis techniques that rely on the Fourier transform. Because of the time‑frequency uncertainty principle (ΔT·Δf ≥ 1/4π), these methods lose the ability to separate individual modes once the modal overlap factor μ exceeds roughly 30 %. In many engineering applications—particularly in the mid‑ to high‑frequency range where many closely spaced modes coexist—this restriction hampers accurate identification of natural frequencies, damping ratios, and mode shapes.

To overcome this, the authors propose a complete experimental‑computational workflow that couples a high‑resolution spectral estimation algorithm with an automatic order‑detection (model‑order selection) scheme. The workflow proceeds as follows:

1. Excitation and Measurement – A pseudo‑impulse force, i.e., a short‑duration broadband excitation that approximates an ideal Dirac impulse, is applied at a point of interest on the structure. Simultaneously, the structural response is recorded at one or more measurement points using accelerometers or laser vibrometers.

2. Impulse‑Response Reconstruction – The measured force and response signals are deconvolved in the digital domain to obtain the pure impulse response function (IRF) of the structure. Prior to deconvolution, the data are conditioned through windowing, high‑pass filtering, and amplitude normalization to suppress noise and baseline drift.

3. Band‑wise Segmentation – The reconstructed IRF is divided into a series of narrow frequency bands (typically a few hertz wide). This segmentation reduces the effective modal density within each band, making subsequent order detection more reliable.

4. Order Detection – For each band, the number of true modes is estimated automatically using information‑theoretic criteria such as Akaike’s Information Criterion (AIC) or the Minimum Description Length (MDL). This step prevents over‑parameterization and eliminates spurious peaks caused by measurement noise.

5. High‑Resolution Parameter Estimation – With the model order known, a high‑resolution algorithm (e.g., ESPRIT, MUSIC, or Matrix Pencil) is applied to the band‑limited IRF. These sub‑space methods treat the signal as a sum of damped complex exponentials and solve for the poles of the system directly, yielding precise estimates of natural frequencies, damping ratios, and complex modal amplitudes without constructing a conventional Fourier spectrum.

6. Validation on Test Structures – The methodology is demonstrated on two benchmark problems. First, a square aluminium plate with a pair of “twin” modes (frequency separation < 2 Hz) is analyzed; the high‑resolution approach cleanly separates the two modes, whereas a standard FFT‑based spectrum shows a single merged peak. Second, the same plate is examined over a range of modal overlap values from 30 % to 70 %. In each case the extracted modal parameters agree with those obtained from an improved Rayleigh method (a physics‑based analytical approximation) to within 2 % relative error.

Key Advantages

• Robustness to High Overlap: Accurate mode identification is maintained up to μ ≈ 70 %, far beyond the 30 % limit of Fourier methods.
• Low SNR Tolerance: Because the sub‑space estimators exploit the structure of the data rather than amplitude peaks, reliable results are obtained even when the signal‑to‑noise ratio is modest.
• Automatic Model Order Selection: The order‑detection step eliminates the need for the analyst to guess the number of modes a priori.
• Simplified Experimental Setup: By deconvolving the known pseudo‑impulse, the need for separate transfer‑function measurements is removed, reducing test time and equipment complexity.

Limitations and Future Work
The high‑resolution algorithms involve large matrix operations, which increase computational load, especially for very long records or many frequency bands. At extreme overlap (μ > 90 %), order detection can become ambiguous, and the method may require additional regularization or prior information. Precise calibration of the pseudo‑impulse force is also essential to avoid bias in the deconvolution step.

Future research directions suggested by the authors include: (i) development of parallel or GPU‑accelerated implementations to enable near‑real‑time processing for large‑scale structures; (ii) extension of the framework to non‑linear or non‑stationary systems; (iii) integration of multi‑input/multi‑output (MIMO) measurements for full‑field modal identification; and (iv) automated coupling with structural health monitoring platforms to provide continuous, high‑resolution diagnostics in operational environments.

In summary, the paper delivers a rigorously validated, high‑resolution modal analysis technique that bridges the gap between low‑frequency, low‑overlap regimes and high‑frequency, high‑overlap regimes. By combining pseudo‑impulse excitation, impulse‑response reconstruction, band‑wise order detection, and sub‑space parameter estimation, the authors achieve accurate extraction of modal frequencies, damping ratios, and complex amplitudes even when modes are heavily overlapped. The close agreement with an improved Rayleigh analytical method confirms the physical fidelity of the approach, positioning it as a powerful tool for vibration analysis, modal testing, and advanced structural health monitoring.