Civil Engineering Reference
In-Depth Information
Figure 11.4 The architecture of cluster-based SHM algorithms
computation, and even different capabilities to detect structural damage. How to
find a clustering strategy which is optimal in one or some of these aspects is
important as well as challenging.
Besides clustering, another technique to design WSN-based SHM algorithms is
called “model-based data aggregation” (Nagayama and Spencer, 2008). For some
centralized SHM algorithms it is found that, although they use the raw data of
sensor nodes, the collection of these raw data is, however, unnecessary. Instead, the
raw data can be aggregated according to the SHM algorithms adopted using much
less wireless communication without altering the algorithms themselves. The
broadcast property of wireless communication and the in-network processing
capability of sensor nodes are used during the aggregation to decrease wireless data
transmission. For example, in the traditional NeXT-ERA algorithm, a reference
node (usually a base station and cluster head) samples vibration measurement and
also collects the measurement from other sensor nodes. It then calculates the auto-
correlation function (ACF) of its own measured data and the cross-correlation
functions (CCFs) between itself and other sensor nodes. The ERA algorithm is then
implemented at the reference node based on the ACF and CCFs. The above
procedures can be implemented using the model-based data aggregation: instead
of transmitting the raw data to the reference node, the reference node broadcasts its
data to all the other nodes. After receiving the data, each sensor node calculates the
CCF and only transmits back the CCF to the reference node. In this way, the wireless
transmission is much lower than in the traditional approach (more detail can be
found in Nagayama and Spencer, 2008). In addition, some of the computation
originally implemented at the reference node is offloaded. The model-based data
Search WWH ::
Custom Search