Compressed sensing embedded in an operational wireless sensor network to achieve energy efficiency in long-term monitoring applications

Abstract

Compressed sensing (CS) is a powerful new data acquisition paradigm that seeks to accurately reconstruct unknown sparse signals from very few (relative to the target signal dimension) random projections. The specific objective of this study is to save wireless sensor energy by using CS to simultaneously reduce data sampling rates, on-board storage requirements, and communication data payloads. For field-deployed low power wireless sensors that are often operated with limited energy sources, reduced communication translates directly into reduced power consumption and improved operational reliability. In this study, acceleration data from a multi-girder steel-concrete deck composite bridge are processed for the extraction of mode shapes. A wireless sensor node previously designed to perform traditional uniform, Nyquist rate sampling is modified to perform asynchronous, effectively sub-Nyquist rate sampling. The sub-Nyquist data are transmitted off-site to a computational server for reconstruction using the CoSaMP matching pursuit recovery algorithm and further processed for extraction of the structure's mode shapes. The mode shape metric used for reconstruction quality is the modal assurance criterion (MAC), an indicator of the consistency between CS and traditional Nyquist acquired mode shapes. A comprehensive investigation of modal accuracy from a dense set of acceleration response data reveals that MAC values above 0.90 are obtained for the first four modes of a bridge structure when at least 20% of the original signal is sampled using the CS framework. Reduced data collection, storage and communication requirements are found to lead to substantial reductions in the energy requirements of wireless sensor networks at the expense of modal accuracy. Specifically, total energy reductions of 10-60% can be obtained for a sensor network with 10-100 sensor nodes, respectively. The reduced energy requirements of the CS sensor nodes are shown to directly result in improved battery life and communication reliability.

Sub-Nyquist signal-reconstruction-free operational modal analysis and damage detection in the presence of noise

“…No acceleration signals reconstruction (i.e. retrieval of the x[n] input sequences to the considered array of sensors), and, more generally, no solution of any 1  optimization problem is undertaken for the task as required by other recent compressive sensing based OMA approaches proposed in the literature 6,10 .…”

“…However, despite the above advantages of WSNs, there are still practical issues to be overcome related to a need for local power supply at the sensors and restrictions to the amount of wireless data *Kyriaki.Gkoktsi.1@city.ac.uk transmission due to bandwidth limitations. To address the above issues, some recent research effort [6][7][8][9][10] was devoted in exploring compressive sensing (CS)-based techniques which achieve simultaneous signal acquisition and compression before transmission. In a nutshell, these techniques consider non-uniform random sampling schemes to acquire signals at an average rate below the Nyquist rate, by exploiting signal sparsity in some domain (e.g., the Fourier domain).…”

“…In this context, O'Connor et al 6 reported significant energy savings in a long-term field deployment of prototyped wireless sensors acquiring bridge acceleration response signals via a sub-Nyquist random sampling scheme based on the CS theory. Under the assumption of spectrally sparse measurements, the acquired signals are transmitted towards a base station and reconstructed to the uniform Nyquist grid by means of a standard CS signal recovery algorithm.…”

Condition Assessment of Reinforced Concrete Bridges: Current Practice and Research Challenges

“…Wireless sensors have alleviated the cost and labour associated with cabled monitoring systems. O'Connor et al [55] employed a wireless sensor network to measure bridge accelerations, strains and temperatures. However, limitations of using wireless sensors include constraints in power and transmission bandwidth.…”

“…Solar power supply, vibration, or wind could sustain long-term wireless sensor network operations, while less relevant communication bandwidth constraints could be made by conveying less data. For instance, O'Connor et al [55] introduced the compressed sensing data acquisition approach to achieve energy efficiency in long-term monitoring applications.…”