Structural Monitoring With GPS
The Global Positioning System (GPS) is a rapidly evolving technology that is changing the way many navigation and surveying tasks are performed. The Autumn 1995 issue of Public Roads included an article, "Navigating the Future," which introduced readers to some of the concepts of GPS navigation applications. Now, in this article, we will look at another type of GPS positioning, phase-based surveying, and the emerging GPS application of structural deformation monitoring. Recent advances in GPS technology have made it a cost-effective tool for monitoring safety and performance of bridges.
Aging of our national bridge inventory and the fact that many bridges are carrying greater average loads than predicted during their design have significantly increased the need over the past few years to monitor bridge performance. Internationally, situations such as the Kingston Bridge in Scotland, which is currently undergoing a major repair and retrofitting program to fix critical deficiencies, and the recent collapse of the Korar-Babeldaob Bridge in Palau _ in the Caroline Islands group of the western Pacific Ocean _ have made this an issue that transcends national boundaries.
Both for traveler safety and for maintenance and repair planning, bridge monitoring is becoming increasingly important to transportation authorities around the globe. What's more, structural deformation and deterioration problems faced by bridge authorities are very similar to those faced by dam and railroad authorities. Most large American hydroelectric and flood-control dams were built from the 1930s through the 1950s; thus, our national dam inventory is also aging. To regain cost competitiveness, trains are being designed for significantly higher speeds, requiring that the tracks conform to tighter standards for rail alignment.
Sensor systems which monitor the geometry and deformations of large civil structures are not new. Accelerometers, strain gauges, linear variable displacement transducers (LVDT _ a distance measuring device), and total stations are familiar tools to many professionals involved with structural monitoring. The concept of using GPS for surveying large structures is itself not completely new. It has been known throughout the GPS community for several years that these satellite receivers are capable of determining positions to subcentimeter-level accuracies under appropriate conditions. What is new for structural monitoring applications is that recent advances in GPS receiver technology and data-processing software have made GPS a much more cost-effective tool, which can be integrated into an automated continuously operating system.
In addition to other structural deformation monitoring projects, the Applied Research Laboratories, University of Texas (ARL:UT) has developed a prototype general-purpose GPS-based bridge monitoring system for the Federal Highway Administration (FHWA) nondestructive testing and evaluation program. This system has been successfully tested on two major highway bridges in the United States, and its design is currently being customized for a number of other bridge monitoring projects.
GPS Positioning: Navigation vs. Surveying
In "Navigating the Future," GPS concepts were described for navigation applications that typically require positioning accuracies from one meter to 100 meters. GPS is also used for surveying, producing positions accurate to the centimeter and millimeter levels. The primary difference between navigation and surveying with GPS lies in how ranges to the satellites are computed. All GPS surveying is done using differential techniques with a reference receiver at a known site.
GPS Signal Components and Satellite Ranging
Each GPS satellite transmits on two frequencies called L1 (1.5 gigahertz [GHz]) and L2 (1.2 GHz). L1 is the primary signal used for most civilian applications, and L2 is used for computing ionospheric corrections in some cases. The L1 signal can be divided into three components: carrier wave, tracking codes, and navigation message. Information about the satellite positions (or orbits) is contained in the navigation message. Tracking codes (C/A-code and the encrypted Y-code) are correlated by the GPS receiver with an internally generated replica. This, along with the orbit information, allows the receiver to determine the time of travel of the signal and, thus, the receiver s range to each satellite being tracked.
Both the navigation message and tracking codes are modulated on the carrier wave, a continuous radio signal at the L1 frequency. As with the tracking codes, the GPS receiver can correlate to the carrier wave to gather ranging information to each satellite. For both codes and carrier waves, the receiver can correlate with an accuracy of about 0.3 percent of one cycle, or code chip. Because the frequency of the carrier wave is much higher than the frequency of the code chipping rate (2 megahertz [MHz] for the C/A-code vs. 1.5 GHz for L1), its cycles are much shorter, and thus, it is possible to make more accurate satellite ranging measurements using the carrier wave component of the satellite signal. However, this method is significantly more complex.
Positioning With Phase Data
GPS tracking codes are designed so that when the code is successfully correlated in the receiver, an unambiguous range to the satellite is derived. The receiver knows exactly how far through the long code sequence it currently is. For carrier wave tracking, it is not so simple. Every wavelength looks exactly like every other one, and each is only about 20 centimeters long (one C/A-code chip is 300 meters long). The receiver can determine the phase angle of the current cycle _ but not how many cycles lie between it and the satellite. Because the receiver directly measures the phase angle of the carrier wave, satellite ranging measurements derived from this procedure are called "phase" data. (Note: Correlating to the tracking codes produces "range" or "pseudorange" data. Most low-cost GPS receivers used for navigation applications produce only range data and positions derived from these measurements. Higher cost surveying and geodetic GPS receivers produce range and phase data.)
Whereas, range data directly gives an unambiguous distance between the GPS receiver and the satellite, phase data has an ambiguity in terms of the correct number of cycles between receiver and satellite. Therefore, the key to highly accurate GPS positioning using the phase data is called "ambiguity resolution." All GPS data processing software packages that compute positions to better than 5-centimeter accuracy perform some form of phase ambiguity resolution.
The first step to resolve phase cycle ambiguities is to compute, using the range data, a rough position estimate (good to a few meters). Then, a combination of redundant measurements and changes in satellite geometry over time are used to solve for the unknown number of cycles for each phase measurement. (Note: Only four satellites need to be tracked to solve for a three-dimensional position, but most modern GPS receivers will track up to eight or 12 satellites simultaneously, producing an overdetermined solution _ or redundant measurements.) Once this set of cycle ambiguities is solved, the phase data can directly produce a position for each data point, as long as the receiver does not lose its lock on the signal. If a temporary blockage of signals occurs, the phase cycle ambiguities must be re-resolved.
Conventional GPS phase-based positioning software requires about 30 minutes to one hour of data to allow sufficient changes in satellite geometry to robustly resolve phase cycle ambiguities. A new type of software called On-The-Fly (OTF) ambiguity resolution can perform this function in a few minutes and, under certain conditions, with just one data point.
GPS for Structural Monitoring
There are two system architectures for structural monitoring with GPS, one based on a fixed network of sensors and the other based on mobile sensors. The following sections describe these two implementations and discuss how recent advances in GPS receiver technology are making such systems cost-effective.
Fixed Networks: Continuous Monitoring
Most conventional bridge monitoring systems rely on a fixed network of sensors that transmit their data back to a central site for processing and analysis. This is also a useful architecture for GPS-based systems. In the ARL:UT bridge deformation monitoring system (BDMS), sensor nodes are mounted on the structure at sites of interest. For measuring long-term movement _ such as foundation settlement, creep, stress relaxation, and others _ the sensor nodes are mounted over the bridge piers. For measuring shorter term motion, such as that caused by wind or traffic loading, the sensors are mounted between piers.
Each BDMS sensor node consists of a GPS receiver, microcontroller, and data radio. The GPS receiver tracks the satellite signals and computes the range and phase measurements. These measurements are transmitted to the central processing site by the data radios. (Note: The ARL:UT system uses spread-spectrum radios operating in the 902- to 928-megahertz (MHz) frequency range. These radios provide high data-transfer rates (up to 115 Kbps) and do not require licensing under Federal Communications Commission regulations.) The microcontrollers perform temporary data buffering, provide receiver control interfaces, and manage network communications flow control. The central processing site consists of a data radio and personal computer (PC) with software for system control, data communications and management, data-quality checking, position computation, and movement analysis.
Once the system is initialized, data is automatically collected by the sensor nodes and transmitted to the central processing site. At this stage of system development, data processing and analysis functions are done interactively, with automation planned for future versions.
For a typical bridge instrumentation scenario, the accuracy of each position computed lies at the centimeter level. Averaging positions for each site over time increases accuracy to the millimeter level. Under good tracking conditions, 1-millimeter accuracy is achievable. Assuming that the GPS receivers produce high-quality phase data and that distances between the reference GPS receiver and those on the bridge (baselines) are kept short, positioning accuracy is mainly a function of the averaging time used. This, in turn, depends on the type of motion that is being monitored.
Foundation settlement, creep, and other movement generally occur over relatively long periods of time. Therefore, averaging times of a few hours can be used, producing positions with millimeter-level accuracy. Motion due to wind loading is cyclic with frequencies of a few hertz. Because, therefore, little or no averaging can be done, positioning accuracy is at the centimeter level. This restricts the utility of GPS for monitoring short-term motion to more flexible structures. Also, to measure short-term effects, sampling rates must be significantly increased, inducing heavier processing and data communication loads.
GPS receiver technology and signal processing techniques are rapidly advancing, and in the near future, positioning accuracies for individual data points will improve significantly. This will be particularly useful for monitoring short-term motion such as wind and traffic loading effects where averaging is not possible.
Kinematic Surveys: Periodic Surface Profiles
There is another implementation of GPS for bridge monitoring that does not apply to short-term motion but can provide a very cost-effective means of periodically measuring long-term deformations. This method consists of performing kinematic surveys of the bridge deck, using a combination of GPS and analog sensors. Achievable accuracies are in the centimeter range for each data point, and a very high spatial density of positions over the deck surface is produced. This allows for generation of three-dimensional surface profiles that will show pier settlement and vertical deformations of the superstructure.
For this implementation, one or two GPS receivers are mounted on the side of a vehicle above its roof. A wheel is mounted directly below the GPS antenna with an LVDT inserted into the coupling between the wheel and the antenna. This will measure the effect of the vehicle s suspension system so it can be deleted from the position solutions. Inertial sensors such as accelerometers and gyros can be added to the GPS antenna assembly to further increase positioning accuracy. (Note: GPS and inertial sensors have been combined for high-dynamics military navigation applications for some time, and commercial systems for decimeter-level kinematic mapping have recently become available. ARL:UT is currently in the proof-of-concept testing stage for millimeter-level kinematic positioning using GPS combined with inertial sensors.)
With the stationary reference GPS receiver collecting and storing its data, the instrumented vehicle is driven over the bridge several times in each traffic lane, collecting and storing its data for post-processing. Data from the GPS receivers and other sensors are combined for position computation and error analysis. The discrete positions are then filtered, and a three-dimensional surface profile is generated, representing the current shape of the bridge deck. Comparing successive profiles over time will show the extent and geometry of settlement or bending.
This technique derives its high cost-effectiveness from the fact that a few sensor packages can be used to monitor a large number of bridges. The GPS reference receiver and instrumented vehicle could periodically survey every bridge of interest in a wide region.
Low-Cost, High-Accuracy GPS Technology
GPS has been tested for surveying large structures since the late 1980s. However, until recently, the receivers that could achieve subcentimeter positioning accuracy cost around $30,000 to $40,000 per unit. This made the prospect of deploying a fixed network of sensors along a bridge prohibitively expensive.
Today, the GPS receivers employed in the ARL:UT structural monitoring systems cost around $5,000, and with some additional development, it will be possible in the near future to integrate receivers costing as little as $2,000. (Note: GPS receivers are the primary sensor and dominant hardware cost component for such systems. The overall system cost-effectiveness has also greatly benefited from recent advances in technology and reductions in cost for spread-spectrum data radios and microcontrollers. A discussion of these related technologies is beyond the scope of this article; however, they are also key components of a fixed network GPS-based bridge monitoring system.) Within the next few years, complete GPS receivers need not be used at all. Rather, GPS chipsets could be directly integrated onto microcontroller boards along with spread-spectrum radio chipsets to produce a sensor plus data communications package costing a few thousand dollars.
There are two primary attributes required of a GPS receiver system to produce subcentimeter-level positioning accuracy. First, the receiver itself must generate high-quality coherent phase data. (Note: Coherence ensures that the phase and range data are generated simultaneously.) Second, the GPS antenna assembly must have a precise electrical phase center and adequate multipath rejection capabilities. (Note: Phase multipath is currently best mitigated at the antenna. Several GPS manufacturers are currently developing signal-processing techniques that will reduce multipath effects within the receiver. When these are commercially available, cheaper antenna assemblies can be used, further reducing the system cost.)
Because of imperfections in antenna element design and manufacturing processes, the range to satellites in different parts of the sky will be measured from slightly different points on the GPS antenna. The antenna s electrical phase center is the region to which satellite signals coming from different elevation and azimuth angles get referenced. A good survey-quality GPS antenna will have a phase center size of 5 millimeters or less.
Structural monitoring applications involve relatively short GPS baselines, generally 2 kilometers or less. Under these conditions, multipath, or signal reflection from nearby objects into the GPS antenna, is the dominant source of error. (See "Navigating the Future" for an explanation of why other GPS error sources are highly correlated and thus cancel over short baselines.)
Because phase multipath is zero mean (range multipath is not), its induced errors can be successfully reduced by averaging over time. However, if the antenna assembly does not provide a high degree of multipath rejection, averaging times of four to 24 hours may be needed to produce a single position with subcentimeter accuracy. The most effective device for multipath mitigation at the antenna is a choke ring. These groundplanes with concentric circular troughs reduce multipath through successive rounds of destructive interference as the signal nears the antenna element.
Due to lack of requirements for most applications, the advancement of GPS antenna technology in terms of low cost coupled with high accuracy has not kept pace with receiver advances. Antennas suitable for structural monitoring currently cost around $1,500 to $3,000, making them almost as expensive as the receivers themselves. This situation is mainly because very few networks of low-cost, high-accuracy GPS positioning systems have been deployed. Few, if any, other applications require this level of antenna precision in high numbers. As more GPS structural monitoring systems are developed, market pressure will cause manufacturers to focus more attention on the development of low-cost, high-accuracy GPS antennas, and prices should then be reduced.
Blackwater Bridge deck surface profile generated from a GPS kinematic survey. Plot on top shows current deck elevation minus as-built elevation curve. Region 1 indicates pier settlement, confirmed through conventional surveys. Region 2 is erroneous apparent lifting caused by driving next to a crane, which caused high multipath in GPS signals. Plot on the bottom shows error characteristics of GPS data (through "phase residuals") over the bridge deck surface. Spike in Region 2 indicates erroneous results, but with GPS alone, there is no way to correct for these. Integration of inertial sensors will correction of GPS errors.
For structural deformation monitoring, and all other businesses, panaceas are mythical things. GPS, like any other technology, has limitations for measuring deformations of large civil structures. However, it is becoming a powerful and cost-effective tool for monitoring some types of structural deformation and performance. The past few years have seen dramatic reductions in hardware cost coupled with significant increases in performance.
Such efficiency improvements as seen in GPS are also occurring in key related technologies such as spread-spectrum data radios and embedded microprocessors. These technologies and advances in data processing techniques, such as GPS On-The-Fly (OTF) ambiguity resolution and inertial sensor data integration, are combining to give transportation authorities a new type of system that can measure structural deformations. In the near future, structural monitoring will continue to increase in importance and, with these new tools, will decrease in cost.
The authors thank the Texas Department of Transportation for their assistance in testing the ARL:UT bridge deformation monitoring system on the Hartman Bridge and the Florida Department of Transportation for their assistance with testing on the Blackwater River Bridge. We also thank Harold Bosch at FHWA's Turner-Fairbank Highway Research Center for his technical assistance in the development of the system.
Keith Duff is the program manager for structural monitoring projects in the Geomatics Systems Division of the Applied Research Laboratories, University of Texas at Austin. He received a bachelor's degree in mathematics from the University of Texas and has eight years of experience in GPS-related research and development.
Michael Hyzak is a project engineer for structural monitoring systems development at ARL:UT. He received a bachelor's degree in civil engineering from the University of Texas and is currently finishing his master's thesis on GPS technology applied to structural monitoring applications.