Databases Reference
In-Depth Information
FIGURE 9.11: A continuum of L-shaped plates with embedded WSNs. (This
figure is a copyright of and reproduced with permission of Civil-Comp Ltd.
Previously published in [98].)
as a self-organizing (ad hoc) virtual network of processing nodes. Each node
executes the same copy of a very simple AM algorithm, which provides a nat-
ural framework for supporting parallelism. The algorithm is best suited for
immensely parallel systems, such as WSNs.
9.2.2.1.1 GN for stress pattern detection in a WSN
In SHM, an
arbitrary L-shaped plate with in-plane loading is used as an object of inves-
tigation. It is assumed that each of these plates is embedded with a WSN, as
shown in Figure 9.11.
Complex shapes can be formed using these simple L-shaped plates. The em-
bedded WSN can measure strain, stress, displacement, or any other parameter
of importance in the design of this continuum. These parameters are assumed
to be vectors orthogonal to the plane of the WSN. The in-plane stresses have
been selected as the orthogonal vector for this study. Two stress states, of the
six possible states under the horizontal and vertical load conditions, were arbi-
trarily selected to demonstrate the in-network pattern recognition capability
of the application. It is assumed that these two stress patterns are highly detri-
mental to the continuum and must be watched so that their occurrences are
detected in real time. These patterns can result for non-critical stress states.
However, the final determination of the pattern detected by the WSN is per-
formed outside of the network, where greater computational resources can
be made available for interpolating stress readings obtained from a relatively
coarse-grained WSN.
9.2.2.2
Parallel Adaptive Mesh Refinement
The patterns picked up by the WSN through in-network processing can only
represent binary level variations in the patterns. It is possible that the pattern
