Structural Health Monitoring using Wireless Sensor Networks - Mobile and Pervasive Computing in Construction - page 223

transform (LSWT) method (Zhang and Li, 2006) and the adaptive linear filtering
compression (ALFC) algorithm (Kiely et al ., 2010). The common features shared in
these compression algorithms include fast implementation and integer arithmetic
operations without floating-point operations. These properties enables real-time
compression using resources-limited wireless sensor nodes. Spatial correlation in
data among multiple sensors in a WSN can also be exploited for data compression.
Related work can be found elsewhere (Chou et al ., 2003; Ciancio and Ortega, 2005).
Another type of independent in-network processing techniques is filtering. In
this approach, each sensor node examines the collected data and only transmits the
important portions to the base station whilst dropping others. Whether a data
portion is important is application-dependent. Important data are usually those
collected during a critical event, or those containing the structural damage
information. A WSN-based SHM system called 'Wisden' has been proposed
(Xu et al ., 2004; Paek et al ., 2005). Wisden uses a filtering technique which
eliminates quiescent periods in structural response and only transmits the data
portions where signal energy is above a pre-defined threshold. A system called
'Vango' has also been designed which contains series of filters (Greenstein
et al ., 2006). Sampled data are filtered through these filters and only the strongly
interesting portions are transmitted wirelessly. Applications users can configure
these filters for different application purposes. Experiment on 'Vango' has dem-
onstrated that the filters can differentiate interesting data from uninteresting data
before it is transmitted, and thus reduced the traffic by 78%. Despite of the
successful implementation of filtering techniques in these systems, some inherent
limitations should be noticed. Most of the SHM applications require synchro-
nously sampled data from sensor nodes. Filters in different sensor nodes, unless well
calibrated, can hardly guarantee that filtered data from these sensor nodes are
synchronized.
The third category of independent in-network processing techniques is the SHM
algorithms. Compared with the previously mentioned compression and filtering,
more domain knowledge of structural engineering is involved. The output of SHM
algorithm is the highly compacted information directly associated with structural
condition. How to design simple and effective SHM algorithms in WSN is a
challenging problem and will be described in detail in Section 11.3.
11.2.3 Energy harvesting
High energy efficiency and a sufficient energy source are two significant aspects in
prolonging the lifetime of a WSN-based SHM system. Techniques relating to
energy efficiency have been described in the previous sections. In terms of the
second aspect, Roundy et al . (2004) compared the power densities of available
harvesting sources, such as sunlight, vibration (e.g., human motion), thermal
gradient, and so on, as shown in Table 11.4. It can be seen that the power density of
the solar cells can generate comparably large power supply. Therefore, solar panels
and rechargeable batteries have been chosen in many WSN-based SHM systems.
Perhaps the first SHM system which integrates solar power into wireless sensor
nodes is described by Olund et al . (2007). The SHM system is deployed on a