Energy harvesting wireless sensor networks are a promising solution for low cost, long lasting civil monitoring applications. But management of energy consumption is a critical concern to ensure these systems provide maximal utility. Many common civil applications of these networks are fundamentally concerned with detecting and analysing infrequently occurring events. To conserve energy in these situations, a subset of nodes in the network can assume active duty, listening for events of interest, while the remaining nodes enter low power sleep mode to conserve battery. However, judicious planning of the sequence of active node assignments is needed to ensure that as many nodes as possible can be reached upon the detection of an event, and that the system maintains capability in times of low energy harvesting capabilities. In this paper, we propose a novel reinforcement learning agent which acts as a centralised power manager for this system. We develop a comprehensive simulation environment to emulate the behaviour of an energy harvesting sensor network, with consideration of spatially varying energy harvesting capabilities, and wireless connectivity. We then train the proposed reinforcement learning agent to learn optimal node selection strategies through interaction with the simulation environment. The behaviour and performance of these strategies are tested on real unseen solar energy data, to demonstrate the efficacy of the method. The deep reinforcement learning agent is shown to outperform baseline approaches on both seen and unseen data.

Measuring and analysing the vibration of structures using sensors can help identify and detect damage, potentially prolonging the life of structures and preventing disasters. Wireless sensor systems promise to make this technology more affordable and more widely applicable. Data driven structural health monitoring methodologies take raw signals obtained from sensor networks, and process them to obtain damage sensitive features. New measurements are then compared with baselines to detect damage. Because damage-sensitive features also exhibit variation due to environmental and operational changes, these comparisons are not always straightforward and sophisticated statistical analysis is necessary in order to detect abnormal changes in the damage sensitive features. In this thesis, an automated methodology which uses the one-class support vector machine (OCSVM) for damage detection and localisation is proposed. The OCSVM is a nonparametric machine learning method which can accurately classify new data points based only on data from the baseline condition of the structure. This methodology combines feature extraction, by means of autoregressive modeling, and wavelet analysis, with statistical pattern recognition using the OCSVM. The potential for embedding this damage detection methodology at the sensor level is also discussed. Efficacy is demonstrated using real experimental data from a steel frame laboratory structure, for various damage locations and scenarios.

When coherent light from a laser beam is passed through a transparent reduction of a variable‐density or variable‐area record section, the seismic signals act as an optical grating to produce a diffraction pattern which is the two‐dimensional Fourier transform of the section itself. With suitable lenses the diffraction pattern can be converted back into an image of the original section. By obstructing portions of the pattern corresponding to particular frequencies or dips on the section one can remove such frequencies or dips from the reconstructed image. The equipment developed for this processing incorporates special design features to combine high optical resolution, precise discrimination of moveouts and frequencies, limitation in the length of the overall optical path to permit the use of a short optical bench, and visual monitoring by use of a microscope or a closed‐circuit TV system. Filter elements consist of wedges mounted on a rotary stand for velocity rejection, wires of various diameters for band stop frequency rejection, and plates bounded by knife edges for low‐pass filtering. The technique is applicable to most problems encountered in seismic prospecting where spurious events obscure desired reflections. The most frequent application so far has been the removal of multiple reflections. The method has turned out to be highly useful for eliminating noise, regardless of origin, which interferes with reflections whenever the noise consists of traveling events, even though fragmental, which have different apparent velocities from the reflections. The method has also been effective in solving structural problems in tectonic areas by removing diffractions or, alternatively, by enhancing them at the expense of the reflections to delineate faults and other sources of diffraction. Ringing or reverberation can often be attenuated or eliminated in marine shooting by passing reflection frequencies that are less than the lowest observed harmonic of the fundamental reverberation frequency. Examples are shown of transforms and/or filtered sections illustrating these applications. A particularly valuable feature of this optical processing system is the ease of monitoring the results. The facility with which this can be done gives the technique distinct advantages over digital or analog methods, where the geophysicist loses contact with his results while processing is under way. Optical filtering also offers an intrinsically more economical approach to seismic data processing because hundreds of information channels can be handled n a single photographic operation.

Summary In this paper, a data‐based damage detection algorithm that uses a novel one‐class kernel classifier for detection and localisation of damage is presented. The demands of wireless sensing are carefully considered in the development of this fully decentralised and automated methodology. The one‐class kernel classifier proposed in this paper is trained through a faster and simpler to implement iterative procedure than other kernel classification methods, while retaining the same advantages over parametric methods, making it especially attractive for embedded damage detection. Acceleration time series at each sensor location are processed into autoregressive and continuous wavelet transform‐based damage‐sensitive features. Baseline values of these features are used to train the classifier, which can then classify features from new tests as damaged or undamaged, as well as outputting a localisation index, which can be used to identify the location of damage in the structure. This methodology is evaluated using acceleration data taken from a steel‐frame laboratory structure under various damage scenarios. A number of parametric studies are also conducted to investigate the effect of sampling frequency and baseline data sample size. Copyright © 2016 John Wiley & Sons, Ltd.

No abstract