Civil Engineering Reference
In-Depth Information
then receive the calculated results and continue the remaining tasks. The similar
idea of task assignment has also been applied in Lynch
et al
. (2004), where the task
of finding the ARX model is finished by a central server.
Another significant drawback of node-level SHM algorithms is their ineffec-
tiveness in detecting structural damage. In this approach, each sensor node
performs SHM algorithms based on its own measured data without any collab-
oration with others. Generally speaking, the SHM algorithms which use data from a
single node cannot produce reliable and accurate identification result; even if this
result is only used to evaluate the local area corresponding to the sensor node.
Without collaboration in the data level, an individual sensor node lacks the ability
to distinguish the actual damage from the input change and environment noise,
and therefore, is prone to generating false positive alarms (indication of damage
when none is present). Not surprisingly, most of the classic SHM algorithms, such
as finite element model updating methods (Friswell and Mottershead, 1995), state
space identification based algorithms (Farrar and Doebling, 1997), and eigen
realization algorithms (ERA) (Juang and Pappa, 1985), are all centralized.
11.3.2 Collaborative SHM algorithms
The ineffectiveness of node-level SHM algorithms to detect damage makes
collaborative SHM algorithms a promising technique inWSN-based SHM systems.
Collaborative SHM algorithms are generally modified from the traditional cen-
tralized SHM algorithms. Since centralized SHM algorithms are effective in
detecting damage but generally will incur excessive wireless communication and
computation, collaborative SHM algorithms try to find a way to reduce the wireless
communication and computation while still try to achieve the original damage
detection capability of centralized ones. In terms of how collaborative SHM
algorithms are modified from centralized SHM algorithms, they can be classified
as cluster-based SHM algorithms, model-based data aggregation, and networked
computing.
A straightforward way to modify centralized SHM algorithm is through clus-
tering. In this approach, the whole network is divided into a number of clusters.
Sensor nodes within one cluster are generally within a single hop communication
range of its cluster head (CH). The cluster head in each cluster is responsible of
collecting measurement from all the sensors in its cluster and performs classic
centralized SHM algorithms. Cluster heads can further communicate with each
other to obtain more reliable damage information. The architecture of cluster-
based SHM algorithms is illustrated in Figure 11.4. Compared with the traditional
centralized approach, the cluster-based approach limits the hops as well as the
number of sensor nodes in each cluster, thus limiting the wireless communication
as well as the intra-cluster computation. Compared with the node-level SHM
algorithms illustrated in Figure 11.3, cluster-based architecture uses multiple
sensor nodes to obtain local decision and, therefore, can provide a more reliable
and accurate damage identification result.
In cluster-based SHM algorithms, clustering itself is important. Different
clustering strategies have different intra/inter-cluster wireless communication,
Search WWH ::
Custom Search