Structural Health Monitoring using Wireless Sensor Networks - Mobile and Pervasive Computing in Construction - page 222

Figure 11.2 A hybrid sleep and wakeup triggering system
in-field training. After an initialization process, all the sensor nodes are put into the
sleep mode. Once a sentry node is woken up by its vibration-trigger unit, it
communicates with other sentry nodes to confirm the presence of the event. Once it
is confirmed, all the sentry nodes broadcast a wakeup radio signal to trigger its
neighbors and initiate the sampling process. The structure of this hybrid triggering
approach is illustrated in Figure 11.2. Compared with the approach where only the
vibration trigger is used, this approach has a high probability of achieving
synchronized wakeup. Compared with the approach where only the radio-triggered
wakeup is used, this approach is faster, since all the sensor nodes can be triggered by
neighboring sentry nodes.
11.2.2 In-network processing
One of the most energy consuming operations in a WSN is wireless data
transmission. To overcome this, computational power on the wireless sensor
node is used. Instead of streaming the sampled data directly to a central unit,
the collected data are processed and only the processed information, which uses
fewer bits than the original, is transmitted. From this perspective, in-network
processing can be an effective way of decreasing the energy consumption.
In-network processing techniques used in WSN-based SHM can be largely
divided into three categories: compression, filtering and SHM algorithms.
Compression techniques use fewer bits to express the original data and generally
include procedures such as coding and decoding. Traditional data compression
techniques, either lossless or loss, can be directly used in WSN. In some examples
(Lynch, J., et al ., 2003; Mizuno et al ., 2008), Huffman coding, wavelet compression
is implemented at each sensor node on the sampled vibration data. However, direct
implementation of traditional compression techniques has some limitations, such
as low computing speed and the requirement of auxiliary memory. Therefore,
compression is usually implemented offline, which impedes use in online data
compression applications. To solve the problems, some WSN-tailored data com-
pression techniques have been proposed; for example, the lifting scheme wavelet