Abstract
This paper assesses two different approaches for efficient output-only Vibration-based Structural Health Monitoring (V-SHM) in large-scale civil engineering structures, promoting the use of dense arrays of low-power wireless sensors. Firstly, a non-uniform deterministic sub-Nyquist multi-coset sampling scheme (PSBS) is considered to acquire ambient stationary structural response signals. This sampling scheme is coupled with a power spectrum blind sampling technique along with the frequency domain decomposition algorithm of operational modal analysis to obtain structural modal properties without necessitating signal reconstruction in the time-domain. Secondly, a spectro-temporal compressive sensing approach (STCS) is considered for cases where signal reconstruction in the time-domain is desired. The usefulness and applicability of these approaches are numerically demonstrated using field-recorded data from a bridge overpass and an operating on-shore wind turbine.
Introduction: The Challenge of Wireless V-SHM
Vibration-based Structural Health Monitoring (V-SHM) is a critical tool for ensuring the long-term safety and integrity of civil infrastructure, such as bridges and wind turbines. By analyzing how a structure vibrates under "ambient" loads—such as traffic flow or wind—engineers can identify its modal properties (natural frequencies and mode shapes). These properties serve as a unique "fingerprint" that changes if the structure sustains damage.
However, implementing V-SHM via wireless sensor networks faces a significant hurdle: the "data deluge." Standard sampling requires the Nyquist rate (sampling at twice the highest frequency of interest), which generates massive amounts of data. Transmitting this data wirelessly is energy-intensive, rapidly draining sensor batteries and necessitating frequent, costly maintenance. This research explores two innovative sub-Nyquist sampling methods designed to capture the essential structural information while reducing data transmission by up to 89%.
Methodology
The study compares two distinct approaches to sub-Nyquist data acquisition:
1. Power Spectrum Blind Sampling (PSBS)
This method utilizes a deterministic "multi-coset" sampling pattern. Instead of continuous sampling, the sensor captures a specific, pre-defined subset of data points.
- Innovation: Unlike many compressed sensing techniques, PSBS does not require the signal to be "sparse" (composed of only a few frequencies). It treats vibrations as stochastic processes and aims to reconstruct the Power Spectral Density (PSD) directly from compressed data. Crucially, it skips the computationally expensive step of reconstructing the original time-domain signal.
2. Spectro-Temporal Compressive Sensing (STCS-rwBPDN)
This approach employs random non-uniform sampling in time and relies on a "reweighted Basis Pursuit De-Noising" algorithm.
- Innovation: It uses a two-way communication protocol between the sensor and the server to estimate the signal's sparsity level. Unlike PSBS, this method reconstructs the full time-domain signal, providing engineers with the actual temporal vibration data if needed for further analysis.
Experimental Validation
The authors validated these methods using two real-world field datasets to provide a "ground truth" for comparison:
- Bärenbohlstrasse Overpass (Zürich, Switzerland): Vertical acceleration data from 18 sensors recorded while the bridge was open to traffic.
- On-shore Wind Turbine (Lübbenau, Germany): Acceleration response data from an operating turbine.
The performance of the sub-Nyquist methods (using 11% to 45% of the original data) was compared against the full 100% sampled datasets.
Key Results and Insights
- Significant Data Reduction: Both methods proved remarkably robust, maintaining high accuracy even when using only 11% of the original data. This represents an 89% reduction in data transmission requirements, which directly translates to extended battery life for wireless nodes.
- High Fidelity Identification: For the bridge overpass, both methods identified the first four natural frequencies with an error of less than 1.3% at an 11% compression ratio. The Modal Assurance Criterion (MAC) values—measuring the match between calculated and true mode shapes—were consistently above 0.94.
- Hardware Friendliness: The deterministic sampling used in PSBS is generally easier to implement on low-cost hardware than the random sampling required for STCS, making it highly attractive for mass-produced wireless sensor nodes.
- Frequency Limitations: While highly effective, the study noted that "anti-resonance" parts of the spectrum are the first to degrade as compression increases, particularly in the wind turbine test cases.
Conclusion
This research demonstrates that sub-Nyquist sampling is a viable path forward for sustainable, large-scale structural monitoring. By drastically reducing the power consumed by data transmission, these methods enable the use of denser sensor arrays for longer periods. Future work will focus on the hardware-level implementation of these algorithms on ultra-low-power microcontrollers to measure real-world battery life gains in the field.
Summary updated