The long-term deterioration of large-scale infrastructure is a critical national problem that, when left unchecked, is costly to remedy. In extreme cases it can lead to component failure. For example, 25% of the 600,000 bridges in the United States were rated as structurally deficient or functionally obsolete in 2007. Civil engineering structures collapse more frequently than people realize. From 1989 to 2000, for instance, more than 130 American bridges collapsed.1 Current bridge management practice relies on the use of visual inspection which is mandated by the National Bridge Inspection Program. With visual inspection both subjective2 and costly, there is a need for automated sensor technologies that can monitor bridges during regular operation.
Many new sensor technologies have been proposed for monitoring the health of bridges. In particular, technologies such as microelectromechanical systems (MEMS) offer low costs and improved sensitivities when compared to traditional macro-scale sensors. However, the data acquisition systems required by sensors installed in large civil engineering structures remain a technological bottleneck. Current monitoring systems typically use tens of kilometers of coaxial wire to connect sensors to a centralized data acquisition system. Tethered sensors are expensive to install and maintain, which erodes the cost benefits associated with MEMS units.
Wireless sensors have emerged in recent years as a viable alternative to traditional tethered boxes. These new units are an order of magnitude cheaper than tethered sensors, which often cost a few thousand dollars per sensing channel.3 At the University of Michigan, we developed the Narada wireless sensor prototype explicitly for use in large-scale infrastructure systems. It combines state-of-the-art embedded system technologies to produce a low cost (<$100 per node), high resolution (16-bit) data acquisition node that can interface with traditional sensors.4 Additional functional features include a powerful computational core and a long-range (300m) IEEE802.15.4 wireless radio.
Figure 1. (a) The Yeondae Bridge and (b) its first mode of response using acceleration response data from 50 sensors installed along the bridge. (c) Narada sensor. SRAM: Static random access memory. D/A: Digital-to-analog. A/D: Analog-to-digital.
To validate the performance of the sensor, we conducted a variety of field experiments in operational structures. For example, a network of interfaced MEMS accelerometers was installed in the Yeondae Bridge in the Republic of Korea. This 180m long bridge consists of a steel box girder that continuously spans across five supports (see Figure 1). We installed units in 50 locations along the length of the bridge to measure the vertical acceleration response. We conducted additional field studies on the Gi-Lu cabled-stay bridge in Taiwan using 20 Narada nodes. In both field studies, the sensor operated without data loss and synchronized autonomous units to within 3ms. The data was of such high quality that we could perform a modal analysis.
While wireless communications between sensors is driving interest in this new technology, the most significant innovation is the ability to locally process sensor data. Unlike tethered monitoring systems where computational authority is centralized to a single data repository, a wireless sensor network couples computational resources within each node. Sensor-localized data interrogation alleviates the burden on the broadcast bandwidth by converting high-bandwidth data streams, i.e. raw data, into low-bandwidth streams. Since communication between units consumes more power than any other function, reducing the demand translates into higher power efficiency of the nodes. In addition, sensor-localized interrogation facilitates scalable data management architectures for systems with high nodal counts. To date, distributed modal analysis and agent-based combinatorial optimization have been successfully implemented as part of a larger damage detection strategy.5,6 Our efforts are currently focused on creating parallel computational frameworks for within-network collaborative computing. This approach will minimize the need for data exchange between sensor nodes.
This work has been partially sponsored by the University of Michigan, the National Science Foundation (Grants CMMI-0528867, CMMI- 0726812, and CMMI-0724022) and the Office of Naval Research (Young Investigator Program). Collaborations with Kincho Law (Stanford University), Yang Wang (Georgia Institute of Technology), C. H. Loh (National Taiwan University), and Chung-Bang Yun (KAIST) are greatly appreciated.
Department of Civil and Environmental Engineering
University of Michigan
Ann Arbor, MI
Jerome Lynch is the director of the Laboratory for Intelligent Structural Technology. He is also an assistant professor in the Department of Civil and Environmental Engineering and the Department of Electrical Engineering and Computer Science. He serves on the program committees of numerous SPIE conferences, including ones on nondestructive characterization for composite materials, aerospace engineering, civil infrastructure, and homeland security.