Structural Health Monitoring using Wireless Sensor Networks - Mobile and Pervasive Computing in Construction - page 225

Figure 11.3 A typical node-level SHM system
algorithms are the AR-ARX algorithm (Sohn and Farrar, 2001), the DLAC
algorithm (Messina et al ., 1996, 1998) and the CDED algorithm (Wang
et al ., 2009). These algorithms can be directly embedded into each wireless sensor
node. In this approach, each sensor node performs local processing and only the
damage information, mainly corresponding to the areas where it is located, is
transmitted back to the base station. In the base station, generally a decision-fusion
technique is implemented and a more reliable and accurate damage information is
obtained. A typical node-level SHM system is illustrated in Figure 11.3.
Implementing node-level SHM algorithms in a WSN is relatively simple and
straightforward. Moreover, since no data-level communication is required, the
wireless communication is kept as low as possible. However, it should be noted is
that some of the node-level SHM algorithms are computationally intensive. For
example, in the DLAC method, the curve fitting stage is time consuming. It is quite
possible that the energy consumed during this stage is even larger than transmitting
the raw data. To address this problem, these time-consuming tasks may be more
appropriately assigned to the central station. Therefore, it is sometimes necessary to
decompose a SHM algorithm into several stages and determine by evaluating the
corresponding benefit and cost whether each stage should be implemented at
sensor nodes or should be offloaded onto the base station. This task assignment also
depends on the computational capability of the wireless sensor node and the
network condition. An example of this task assignment can be found in Hackmann
et al . (2008), where curve fitting and optimization tasks in DLAC are offloaded on
to the base station. Sensor nodes first transmit the required data to the base station,