⚡ electrical engineering

Operational Modal Analysis of Aeronautical Structures via Tangential Interpolation

This paper proposes and validates the NExT-LF method, a novel frequency-domain Operational Modal Analysis approach that combines the Natural Excitation Technique with the Loewner Framework to efficiently and accurately identify modal parameters in large aeronautical structures, demonstrating superior performance in noisy conditions compared to traditional SSI and NExT-ERA techniques.

Original authors: Gabriele Dessena, Marco Civera, Oscar E. Bonilla-Manrique

Published 2026-03-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Gabriele Dessena, Marco Civera, Oscar E. Bonilla-Manrique

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have a giant, complex musical instrument, like a massive airplane wing or a helicopter blade. You want to know exactly how it vibrates when it sings its own song. In engineering, these "songs" are called modes, and the notes they sing are frequencies, dampings (how quickly the sound fades), and shapes (how the object bends while vibrating).

Usually, to learn these songs, engineers have to hit the instrument with a giant hammer (a shaker) and listen to the response. This is called Experimental Modal Analysis (EMA). But in the real world, you can't always bring a hammer to a flying airplane or a spinning helicopter. You have to listen to the structure while it's just sitting there, vibrating from wind, traffic, or its own engines. This is called Operational Modal Analysis (OMA) or "listening to the ambient noise."

The problem? The "noise" in the real world is messy. It's like trying to hear a violin solo in a crowded, noisy cafeteria.

The Problem with Old Methods

The paper discusses two old ways to solve this "noisy cafeteria" problem:

The Time-Traveler (NExT-ERA): This method tries to reconstruct the "pure" sound by looking at how the vibration changes over time. It's good, but if the cafeteria is too loud (noisy data), it gets confused and misses some notes.
The Statistician (SSI-CVA): This method uses heavy math to find patterns in the noise. It's very accurate but can be slow and computationally heavy, like trying to count every grain of sand on a beach to find a specific one.

The New Solution: NExT-LF

The authors propose a new hybrid method called NExT-LF. Think of it as a smart filter that combines the best of both worlds.

Here is the analogy:

NExT (The Translator): First, the method takes the messy, noisy vibrations from the real world and translates them into a "ghost impulse." Imagine if you could magically convert the chaotic noise of the cafeteria into a single, perfect "clap" that the structure would have made if you had hit it with a hammer. This is the "Impulse Response Function."
LF (The Sculptor): Once you have that perfect "ghost clap," the second part of the method (the Loewner Framework) steps in. Think of the Loewner Framework as a master sculptor who looks at that single clap and instantly carves out the exact shape of the sound waves. It is incredibly fast and efficient at turning data into a clear picture of the structure's vibrations.

By combining the Translator (NExT) and the Sculptor (LF), the team created a method that is fast, handles noise well, and doesn't need a hammer.

The Experiments: Two Test Cases

To prove their new method works, they tested it on two very different "instruments":

The Giant Wing Spar (XB-2):
- The Setup: A large, flexible wing model made of aluminum, tested in a wind tunnel. They used a shaker to hit it (simulating a controlled environment).
- The Result: The new method (NExT-LF) found the "notes" (frequencies) just as well as the old methods. However, it was much better at figuring out the "damping" (how fast the sound fades) than the old time-traveler method. It was like hearing the song clearly even when the room was slightly echoey.
The Helicopter Blade (H135):
- The Setup: A real, unserviceable helicopter blade. This was the hard test. They didn't use a hammer; they just let the blade sit in a lab and listened to it vibrate from tiny, natural movements (Ambient Vibration Testing). The signal was very weak and noisy.
- The Result: This is where the new method shined. The old "Time-Traveler" method (NExT-ERA) got lost in the noise and missed several "notes." The new NExT-LF method, however, managed to hear three extra notes that the others missed! It found hidden vibrations that were previously invisible.

The Catch (Limitations)

The paper admits that while the new method is great at finding the pitch (frequency) and the shape of the vibration, it still struggles a bit with the volume fade (damping), especially in very quiet, noisy environments. It's like being able to tell you exactly what note a singer is hitting, but having a hard time guessing exactly how long the note will last if the room is very noisy.

Also, the method needs a good "reference point." For the wing, one sensor was enough. For the helicopter blade, they needed two sensors to catch the vibration in different directions, otherwise, the translation got a bit garbled.

The Bottom Line

This paper introduces a new, super-efficient way to listen to the "songs" of airplanes and helicopters without needing to hit them with a hammer. It combines a clever translation trick with a fast sculpting technique to find hidden vibrations that other methods miss. It's a significant step forward for keeping our aircraft safe and efficient by understanding how they vibrate in the real world.

1. Problem Statement

Operational Modal Analysis (OMA) is essential for extracting modal parameters (natural frequencies, damping ratios, and mode shapes) from structures under ambient or operational conditions without controlled input. While widely used in civil engineering, applying OMA to aeronautical structures presents specific challenges:

Noise Sensitivity: Aeronautical datasets are often contaminated by operational noise.
Computational Cost: Traditional time-domain methods like Stochastic Subspace Identification (SSI) can become computationally cumbersome for large, complex systems.
Robustness Issues: Existing frequency-domain methods like NExT-ERA (Natural Excitation Technique combined with Eigensystem Realization Algorithm) often struggle with noise contamination, leading to unstable damping estimates and difficulty in identifying higher-order modes.

The paper addresses the gap in applying advanced frequency-domain identification techniques, specifically the Loewner Framework (LF), to aeronautical structures using output-only data.

2. Methodology

The authors propose a hybrid method termed NExT-LF, which couples the Natural Excitation Technique (NExT) with the Loewner Framework (LF) via tangential interpolation.

Natural Excitation Technique (NExT):
- Used to convert output-only time-domain acceleration data into Impulse Response Functions (IRFs).
- Assumes ambient excitation is a stationary, broadband random process.
- Computes cross-correlation functions between sensor channels, which theoretically share the same functional form as free-decay responses.
Loewner Framework (LF):
- A frequency-domain system identification technique based on rational interpolation.
- It constructs a state-space realization directly from frequency response data (in this case, derived from the IRFs).
- Workflow:
  1. Acquire time-domain output data.
  2. Apply NExT to generate equivalent IRFs.
  3. Transform IRFs to the frequency domain via Fast Fourier Transform (FFT) to obtain equivalent Frequency Response Functions (FRFs).
  4. Feed the FRFs into the LF algorithm to extract system matrices ( $A, E, B, C$ ) and subsequently the modal parameters.
Comparison Baselines: The proposed NExT-LF method is benchmarked against:
- NExT-ERA: The standard time-domain approach.
- SSI-CVA: Stochastic Subspace Identification with Canonical Variate Analysis (a robust time-domain method).

3. Experimental Benchmarks

The study validates NExT-LF on two distinct aeronautical experimental datasets:

eXperimental BeaRDS 2 (XB-2) Flexible Wing Spar:
- Structure: A dynamically scaled main spar of a high-aspect-ratio wing (1.45 m span).
- Excitation: Controlled shaker excitation (random input, 2–400 Hz).
- Data: 8 accelerometer channels.
- Goal: Test performance under high signal-to-noise ratio (SNR) conditions.
Airbus Helicopters H135 Bearingless Main Rotor (BMR) Blade:
- Structure: A full-scale rotor blade (4.99 m span) with complex geometry (flapping, lagging, torsion).
- Excitation: Ambient Vibration Testing (AVT) in a lab environment (no external shaker).
- Data: 9 triaxial accelerometers (18 channels total).
- Goal: Test performance under low SNR and complex structural dynamics.

4. Key Results

A. Natural Frequencies ( $\omega_n$ )

High Accuracy: NExT-LF demonstrated excellent agreement with benchmark data for natural frequencies in both test cases.
- XB-2: Errors were minimal (e.g., < 0.2% for higher modes).
- H135 Blade: Relative differences compared to reference Automated OMA (AOMA) results were generally below 3%.
Mode Discovery: NExT-LF identified two additional modes (at 19.69 Hz and 59.14 Hz) in the H135 blade that were missed by both the NExT-ERA method and previous literature benchmarks. These were identified as coupled modes.

B. Damping Ratios ( $\zeta_n$ )

High Variability: As is typical in OMA, damping estimates showed significant dispersion across all methods (NExT-LF, NExT-ERA, SSI-CVA).
Comparison:
- XB-2: NExT-LF provided damping estimates closer to the benchmark than NExT-ERA, though SSI-CVA performed best overall.
- H135 Blade: All methods struggled with low-frequency damping accuracy due to low excitation levels. NExT-LF showed large relative errors (common in OMA) but remained consistent with the physical behavior of the structure.

C. Mode Shapes ( $\phi_n$ ) and Stability

Stabilization Diagrams: NExT-LF produced significantly more stable stabilization diagrams than NExT-ERA, particularly for damping estimates. NExT-ERA failed to stabilize damping for many modes in the XB-2 case.
Modal Assurance Criterion (MAC):
- For dominant modes in the low-to-mid frequency range, NExT-LF achieved high MAC values (>0.95) against benchmarks.
- Limitation: Above ~53 Hz (H135 blade), the correlation dropped (MAC < 0.7). This is attributed to the lack of sufficient excitation energy in the AVT data for higher frequencies and the absence of torsional measurements, rather than a failure of the LF algorithm itself.

5. Key Contributions

First Aeronautical Application: This is the first documented application of the NExT-LF method to aeronautical structures, extending its use beyond civil engineering.
Superior Mode Extraction: The method successfully identified a richer set of modes (13 modes for H135 vs. 9 for NExT-ERA) within the frequency band of interest.
Robustness to Noise: The coupling of NExT with LF proved more robust than NExT-ERA in handling noisy, output-only data, particularly in recovering stable damping estimates and higher-order modes.
Validation of Workflow: The study validates a workflow where time-domain correlations are transformed to the frequency domain for Loewner interpolation, offering a computationally efficient alternative to pure time-domain subspace methods for large datasets.

6. Significance and Conclusion

The paper concludes that NExT-LF is a viable and effective tool for the Operational Modal Analysis of aeronautical structures. It offers a compelling alternative to traditional methods by:

Providing higher computational efficiency for large systems.
Delivering more reliable mode stabilization, especially for damping ratios where NExT-ERA often fails.
Successfully identifying complex coupled modes in helicopter blades that other methods missed.

Limitations noted: The accuracy of mode shapes and damping ratios in high-frequency ranges is heavily dependent on the excitation conditions (broadband energy) and sensor layout (specifically the need for torsional measurements). The method's performance degrades when ambient excitation is insufficient to activate higher-frequency modes, a limitation inherent to the NExT preprocessing step rather than the Loewner Framework itself.

Overall, the study suggests that with careful selection of reference channels and frequency bands, NExT-LF can significantly enhance structural health monitoring and model updating for complex aerospace systems.