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

Civil Engineering Reference

In-Depth Information

of sensor nodes, how to deploy these sensor nodes is a very fundamental as well as

important problem. About this problem, engineers from civil engineering and

computer science engineering have different concerns.

Civil engineers mainly consider whether the deployed sensor nodes are able to

achieve a good damage detection capability. A widely accepted criterion to evaluate

the deployment quality is the determinant of Fisher information matrix (FIM).

The larger the FIMdeterminant of the deployed sensor nodes, the higher the quality

of deployment (Kammer, 1991). Correspondingly, various sensor deployment

algorithms have been proposed (Worden and Burrows, 2001; Yao et al ., 2002;

Guo et al ., 2004) to maximize the FIM so as to achieve better damage detection

capability.

However, when designing the deployment algorithms, civil researchers all

assume a centralized wire-based SHM system. When wireless sensor nodes are

used as the platform for SHM, some practical issues arise. Deploying wireless sensor

nodes according to FIM only gives a disconnected WSN. Even if it is connected,

the energy consumption of delivering the data using the WSN can be significantly

high or unbalanced, limiting the practical use of the system. These topics are among

the main concerns of researchers in computer science engineering.

From the discussion above, sensor deployment inWSN-based SHM is a problem

requiring knowledge fromboth civil engineering and computer science engineering.

This multidisciplinary problem was first considered by Li et al . (2010), where an

algorithm called p-SPEMwas proposed to deploy wireless sensor nodes. Iteratively,

p-SPEM tries to maximize the ratio of the FIM determinant and the energy

consumptionwhile at the same time ensureingnetwork connectivity. The simulation

results (Li et al ., 2010) indicate that deployment usingp-SPEMperformsmuchbetter

(about 2-5 orders of magnitude) than the two uniform approaches and stays very

close to the EFI method in which only the FIM is considered.

Although p-SPEM has shown good performance, it assumes a static WSN

and the distance of two wireless sensor nodes is used as the only criterion to

evaluate whether they can communicate with each other. However, these two

assumptions are not realistic for practical applications. In practice, the wireless

propagation models at different candidate locations in the structure should be

developed by in-filed radio communication tests. These models, along with the

chosen MAC and routing protocols, should be considered in a sensor placement

algorithm to achieve some pre-defined objectives, such as damage detection

capability, battery lifetime, delay or degree of route redundancy. After deployment,

the statistics about the link quality in the WSN can also be used to update the

deployment. Practical sensor deployment problem in WSN-based SHM requires