vibration frequency of the Humber Bridge in the ... bridges with an overall dimension exceeding 100m. .... bridges such as the London Millennium Bridge.
HYBRID SENSOR SYSTEM FOR BRIDGE DEFORMATION MONITORING: INTERFACING WITH STRUCTURAL ENGINEERS Xiaolin Meng, Alan Dodson, Gethin Roberts, Emily Cosser Institute of Engineering Surveying and Space Geodesy (IESSG), The University of Nottingham, NG7 2RD, UK Abstract. Analytical finite element (FE) modelling, modal testing and FE model updating are approaches which are employed to detect structural dynamic characteristics in the mechanical and aerospace engineering communities. Their application to civil engineering structures is more recent. To successfully apply these approaches, field data acquisition plays a very important role in validating and updating the analytical model. As one of the feasible data collection tools for modal testing, GPS has been employed with other sensors by the IESSG at The University of Nottingham to monitor large structural deformations. While demonstrating promising advantages in positioning precision and flexibility in instrumentation installation, the requirements from the structural engineering perspective need to be well defined before this technology can be fully implemented. In this paper prototypes of hybrid sensor systems coupled with a computational FE model are introduced. Data collected from a test bed bridge are processed and analysed to demonstrate the feasibility of proposed systems. The results reveal that it is possible to achieve 3D millimetre positioning precision for detecting high dynamic structural deformations, and obtaining a highly accurate FE model for the purpose of structural health monitoring (SHM). Keywords. GPS, Finite Element (FE) Model, Structural Deformation Monitoring
1 Introduction A new science of structural health monitoring has emerged to assist the monitoring of structural integrity in recent years. One of the major tasks in SHM is the determination of structural dynamic models. Based on the parameters identified such as natural vibration frequencies, mode shapes and damping ratios, structural faults and fault locations can be detected through collecting precise field
testing data and updating an analytical model such as a finite element (FE) model. Two different methods are available to experimentally identify the structural dynamics: forced vibration testing (FVT) and ambient vibration testing (AVT). With FVT, known weight shakers are used to excite certain vibration modes. The output responses are measured by sensors such as accelerometers and more recently by GPS. The recorded results are then used to update analytical FE models. However, it is very difficult to find a device or method that can excite and detect the structural vibration frequencies of less than 2 Hz (http://www.rcidynamics.ch/), which are evident for large suspension bridges. For instance, the vertical vibration frequency of the Humber Bridge in the UK is 0.116 Hz (Brown et al. (1999)). AVT is actually an output-only approach. It represents many real world scenarios where only response data, such as 3D geometric deformations or forces on the structures are available but the actual loading conditions (inputs) are totally or partially unknown. Typical advantages of AVT are that no expensive dynamic force generator is required, tests can be conducted without interrupting traffic, and it can be used for the bridges with an overall dimension exceeding 100m. However, compared with more mature FVT approach, AVT is still in the research phase. The feasibility of GPS-based large structural deformation monitoring has been demonstrated in many GPS engineering applications in recent years (Fujino et al. (2000); Roberts et al. (1999)). In the presence of sufficient satellites with strong geometry, GPS can provide continuous, all-weather, automated and highly accurate measurements. The displacements measured with GPS can be compared with numerically predicted values from an established FE model either to update the model or to detect any potential structural problems. The defects of GPS positioning compared to more traditional techniques for such applications have been investigated by Meng (2002). For instance, GPS signal availability is always problematic due to
surrounding obstructions, such as dense bridge cables, high rising bridge towers and buildings, etc. Also, due to the 55º inclination, satellite availability depends on geographical latitude and GPS satellites are not uniformly distributed around the globe. As a consequence, the positioning precision of the two horizontal components will not be at the same level, and it will cause misinterpretation of the actual structural deformations (Meng et al. (2003b)). Multipath, residual tropospheric delay and cycle slips are other major errors that need to be addressed when GPS is used for structural deformation monitoring. Slow sampling rate greatly limits the efficiency of a GPS alone system for this purpose. For example, the currently available 20 Hz sampling rate is insufficient to monitor bridge deformations with vibration frequencies higher than 10 Hz. Ways of improving the signal availability and augmenting satellite geometry, as well as how to increase the overall sampling rate of a monitoring system are the research focus. The correct extraction of structural dynamics from field testing data is another important issue from the perspective of structural engineers. A hybrid deformation system, consisting of dual frequency GPS receivers and triaxial accelerometers, has been proposed initially by the researchers at the IESSG of The University of Nottingham (Dodson et al. (2001); Roberts et al. (2000a); Roberts et al. (2000b)). Hardware and software have been developed and validated in a series of bridge trials in the UK (Meng (2002); Roberts et al. (2000a)). With this hybrid system, dramatic improvements in productivity, system reliability and overall performance were achieved. However, the data quality of this hybrid system largely relies on the quality of raw GPS measurements. The relative displacements from a triaxial accelerometer can only bridge very short gaps due to the drifts caused by the instrumental biases and scale factor errors. For example, the vertical position error can accumulate to 5m within 40 seconds using the force measurements collected by a Kistler triaxial accelerometer. Assuming that multipath and other systematic and random errors could be effectively modelled and mitigated, and that the sampling rate can be increased to track high structural dynamics, the defect caused by the GPS orbit design cannot be overcome by this system. Pseudolites are ground based transmitters of GPSlike code and carrier phase signals on either L1 or L2, and can be used to enhance GPS by providing increased accuracy, integrity and availability (Elrod and Dierendonck (1996)). They are ideal for augmenting GPS constellation for precise
deformation monitoring of bridges with limited movements (Dai et al. (2001); Barnes et al (2003)). In these applications, properly installed pseudolites can be used as another ranging source, which can improve the dilution of precision (DOP) and hence the overall quality of the solution. However, error sources relevant to pseudolites based positioning such as tropospheric delay and multipath, as well as time synchronisation and near-far problem need to be further investigated. In this paper, a suspension footbridge is used as a test bed for the proposed research. Predicted results from an FE model developed by project partners at Cranfield University, are compared with the structural dynamics extracted from the field testing data. Discussions are then made regarding the selection of the optimal sensor locations, and signal extraction for updating the analytical model. The research involved in these aspects form the interface between geodesists and structural engineers. Aimed at monitoring structural health, different hybrid sensor integration approaches are introduced. The authors demonstrate that by using the proposed sensor systems it is possible to achieve 3D millimetre positioning precision for continuous structural deformation monitoring.
2 A Test Bed for Structural Deformation Monitoring The research presented in this paper is funded by the Engineering and Physical Sciences Research Council (EPSRC) in the UK. The overall objective of the project is the creation of a system employing advanced computational tools such as FE modelling, coupled with GPS, pseudolites and triaxial accelerometer sensors, to remotely monitor the health of operational bridges without on-site inspection. The Wilford suspension bridge over the River Trent in Nottingham was chosen as a test bed for this project because of its high dynamics under normal loading conditions and convenience to access. It is a 66m long and 3.5m wide footbridge. This bridge consists of two sets of suspension cables restrained by two masonry anchorages (Figure 1). A steel deck covered by a floor of wooden slats forms the bridge deck. Three water/gas pipes are laid underneath the wooden deck for supplying water and gas to the local residents. In the last four years, a number of trials were conducted on this bridge to test the algorithms and hardware. Different instruments were installed on
the bridge, which included dual/single frequency GPS receivers, triaxial accelerometers, pseudolites, surveying total stations and meteorological meters. A complete procedure for bridge deflection monitoring was established, and successfully applied to the deformation monitoring of other bridges such as the London Millennium Bridge (Roberts et al. (2000b)). As shown in Figure 1, two closely setup reference stations were used to mitigate multipath (Roberts et al. (2001)). To detect bridge dynamics, the number and locations of the sensors are crucial. The characterisation of the dynamic behaviour of a structure, or the identification of the presence of structural faults, is possible only under certain conditions. This requires having a minimum amount of information collected or a minimum number of sensors installed. Therefore, the sensors should be strategically placed in order to provide a clear “image” of the structural behaviour. However, due to economical issues, usually only a limited number of sensors are placed on some accessible locations on the real structures. Many approaches can be employed to find optimal sensor locations. Figure 2 shows the sensor locations determined by an optimal algorithm, EFfective IndependenceDriving-Point Residue (EFI-DPR) (Meng et al. (2003a)). Details of this approach are beyond the scope of this paper.
fact that there were no detailed material and design descriptions about this old bridge. At this stage, the initial FE model is not accurate enough for health monitoring purposes. Therefore updating the FE model with precise field measurements is required. GPS/FEM approach could be then used to monitor the changes of structures and detect possible structural faults, through collaborative research between two universities with complementary expertise.
Fig. 2 Rover Station Layout.
Fig. 3 3D FE Model for the Wilford Bridge. Table 1 Estimated Natural Frequencies by FE Model.
Fig. 1 The Wilford Bridge and Dual Reference Stations
The construction of a finite element model was undertaken using the SAFESA method, an FE algorithm developed by Cranfield University. The 3D FE model developed for the Wilford bridge is shown in Figure 3. With the recommended instrument layout, sufficient information about bridge dynamics can be obtained. Table 1 lists the relevant natural vibration frequencies estimated by the established FE model. While the first natural frequencies extracted from the accelerometer and GPS measurements are identical using Fast Fourier Transform (FFT), the frequencies estimated by the analytical FE mode do not match well. This could be partially due to the
Frequencies 1 2 3 4 5 6
1.3797 2.6949 4.4851 6.5006 8.6137 10.72
3 A Hybrid Sensor System Over the last couple of years, different hybrid sensor systems have been investigated. Initially, the research focus was on the expansion of the measurable sampling rate to detect higher structural dynamics. The research then evolved to develop a system which can be used to identify both small and large dynamic deformations with millimetre 3D positioning precision. Both systems are compared
in the following sections, using sample data sets collected from bridge trials. 3.1 Integration of GPS/Accelerometers The vertical vibration frequencies of the Wilford bridge could reach 20 Hz which are out of the detectable range of a GPS receiver gathering data at a rate of 20 Hz (see Table 1 and also Figure 4). Figure 4 shows 3D vibration frequencies in a bridge coordinate system (BCS), a defined computational reference system with its longitudinal axis along the bridge major axis, lateral axis perpendicular to the major axis and vertical axis coinciding with the plumb line. The frequencies are extracted from a sample data set collected by a Kistler triaxial accelerometer during a GPS and accelerometer based trial on the Wilford bridge on 20 June 2002 (Figure 5). The sampling rate in this trial was 100 Hz. Five obvious vertical movements in Figure 5 were excited by force. From Figure 4, it can also be seen that more vibration frequencies were evident in the lateral direction of the bridge due to strong lateral wind loading and the external force applied on the bridge (people jumping on the bridge when tests were made), which excited more mode shapes than in the vertical or longitudinal directions.
estimated from 10 Hz GPS positions and 100 Hz relative displacements by double integration of the vertical accelerations. It can be found that through GPS and accelerometer data integration the sampling rate can be expanded to the same level as an accelerometer, and more vibration frequencies can be detected by the FFT algorithm (see the low graph in Figure 8).
Fig. 5 3D Accelerations Excited by Force.
Fig. 6 3D Mean Displacements by GPS.
Fig. 4 3D Frequencies for the Wilford Bridge.
Figure 6 shows 3D mean centered displacements estimated from GPS measurements at the same time when the acceleration data were collected as illustrated in Figure 5. It is difficult to see the clear vertical movements caused by excitations, due to the high level of multipath and other receiver related random noises. Most vibration frequencies which are still measurable by GPS are buried undetectably by the noise spectra, as shown by Figure 7. To increase the sampling rate and also improve positioning precision, a hybrid system consisting of dual frequency GPS receivers and triaxial accelerometers was proposed. The algorithms and data processing procedures are documented in Meng (2002). Figure 8 shows an integrated vertical position result for a period of five minutes,
Fig. 7 3D Spectra of GPS Data.
Fig. 8 Integrated Positions and Frequency Distribution.
3.2 Integration of GPS and Pseudolites While the initially proposed sensor system could increase the data sampling rate and bridge the short
time gaps of GPS positioning, it cannot improve the data quality caused by poor satellite geometry and signal availability. One effective way to solve the problem is to introduce several ground-based pseudolites to form stronger geometry of signal transmitters. Simulated results reveal that it is possible to achieve uniform millimetre level 3D positioning accuracy. Selection of optimal pseudolite locations is critical. Criteria for this have been investigated by Meng (2002) using a DOP simulator, and further validated by the actual pseudolite configurations. Figure 9 shows the instrument layout for the bridge deformation monitoring trial jointly conducted with The University of New South Wales in Australia on 16th October 2002. In this trial, three Integrinautics’ IN200 pseudolite generators were setup on the northern side of the Wilford bridge to augment the GPS satellite geometry. Pseudolite PL1 was setup on a tripod on the west bank of the River Trent. PL2 and PL3 were set up on the east bank. Two Allstar single frequency receivers capable of receiving GPS and pseudolite signals were used as reference and rover stations. It should be noted that during the data collection the sampling rate was set to 1 Hz, which can only be used to assess the improvement in the positioning solutions due to the augmented system.
Using the multipath corrected phase measurements and a GPS/pseudolite post-processing software package developed by the SNAP group in The University of New South Wales, each epoch positions have been solved. Figures 11, 12 and 13 show the mean centred easting, northing and vertical coordinates for the pseudolite augmented system. Comparing the standard deviations between GPS and the augmented systems, improvements at 43.4%, 13.3% and 34.6% were achieved in easting, northing and vertical component respectively. Also the vertical positioning precision is higher than that of the northern component with augmented system. These results confirm the simulated solution calculated by Meng (2002). To improve the positioning accuracy in the northern component to the same level as the easting or vertical component, some pseudolites should be set above the horizon of the receivers. However, for this particular observation environment, it is not feasible to find such locations. 5-12(L1) mean 51.1 mm stdev
3.8 mm
0.1
0.05
0
-0.05
-0.1 0
500
1000
1500 No. of Samples
2000
2500
Fig. 10 Phase DD before Removal of Multipath. N
PL 3 PL 1 W Tower
East
stdev
2.6 mm
0.05
0.025
PL 2 E Tower
0
-0.025
Rover Ref
-0.05 0
500
1000
1500 2000 No. of Samples
2500
3000
Fig. 11 Mean Centred Easting Positions from GPS/PLs Fig. 9 Pseudolite Setup for a Bridge Trial.
Since there are only very small relative movements between pseudolites and receivers, multipath caused by the surroundings was assumed as to be constant bias. This can be identified by phase measurement double differencing. Multipath can then be removed from the raw measurements. Figure 10 shows the phase double differencing residuals between the base satellite PRN5 and PL1 (designated as PRN12). The mean value of these residuals is 51.1 mm. This could be due to signal obstruction from the arches of the west bridge anchorage. After removal of this constant bias from the phase measurements, using the approach recommended by Barnes et al. (2003), the mean residuals are reduced to -0.3 mm.
North
stdev
7.8 mm
0.05
0.025
0
-0.025
-0.05 0
500
1000
1500 2000 No. of Samples
2500
3000
Fig. 12 Mean Centred Northing Positions from GPS/PLs Vertical
stdev
6.8 mm
0.05
0.025
0
-0.025
-0.05 0
500
1000
1500 2000 No. of Samples
2500
3000
Fig. 13 Mean Centred Vertical Position from GPS/PLs.
Conclusions
Aimed at detecting structural dynamics for health monitoring using GPS, accelerometers and pseudolites, this paper has discussed the interface between structural engineers and geodesists. To achieve this, a test bed bridge for the project was introduced. An analytical FE model was developed for this bridge. Initially predicted results by the FE model were compared with structural dynamics extracted from field testing data. The needs to update this model were emphasised. Two sensor systems, for collecting data, were introduced with particular focus on the signal processing for extracting structural dynamics. The authors pointed out that with a hybrid sensor system consisting of GPS and accelerometers, the sampling rate can be increased to the same rate as the accelerometer and short time millimetre positioning precision can be assured given that there is a strong satellite geometry, and sufficient number of satellites. The authors then proposed a hybrid system comprising of ground based pseudolites and GPS satellites for improving the transmitter geometry. According to the data processing for an actual bridge trial, 3D positioning precision can reach the millimetre level. In trials with proposed system, an improvement in precision of up to 43% was achieved. This paper demonstrated through real examples that the integration of GPS, accelerometers and pseudolites can overcome the defects in GPS satellite geometry, signal availability, positioning accuracy and sampling rate.
Acknowledgements The authors would like to acknowledge EPSRC in the UK for providing the funding for this project (Ref: GR/R28218/01). The supportive research of the Cranfield University is gratefully acknowledged. The Australian Research Council International Linkage programme supported Dr Barnes’s travel to Nottingham to conduct the GPS/pseudolite based bridge trial and his assistance in the data processing is highly appreciated.
References Barnes, J., C. Rizos, R. Warnant, X. Meng, E. Cosser, A.H. Dodson, G.W. Roberts (2003). The Monitoring of Bridge Movements using GPS and Pseudolites. In: Proc of 11th International Symposium on Deformation Measurements, International Federation Surveyors (FIG), Commission 6 Engineering Surveys, Working Group 6.1, 25-28 May, Santorini, Greece. Brown, C.J., R. Karuna, V. Ashkenazi, G. Roberts, R.A. Evans (1999). Monitoring of Structures Using the Global Positioning
System. In: Proc Instn Civ. Engrs Structs & Bldgs, February, pp. 97-105. Dai L., J. Wang, C. Rizos and S Han (2001). Pseudo-satellites Applications in Deformation Monitoring. GPS Solutions, 5(3), pp. 80-87. Dodson, A.H., X. Meng and G. Roberts (2001). Adaptive Method for Multipath Mitigation and Its Applications for Structural Deflection Monitoring. In: Proc of International Symposium on Kinematic Systems in Geodesy, Geomatics and Navigation (KIS2001), Banff, Alberta, Canada, 5-8 June, pp. 101-110. Elrod, B.D. and A.J.V. Dierendonck (1996). Pseudolite. In Global Positioning System: Theory and Applications, edited by B.W. Parkinson, J.J. Spilker, P. Axelrad, and P. Enge, American Institute of Astronautics (AIAA), Inc., Washington, D.C., pp. 51-79. Fujino, Y., M. Murata, S. Okano, M. Takeguchi (2000). Monitoring System of the Akashi Kaikyo Bridge and Displacement Measurement Using GPS. In: Proc of Nondestructive Evaluation of Highways, Utilities, and Pipelines IV, Proceedings of SPIE, edited by A.E. Aktan, and S.R. Gosselin, pp.229-235. Meng, X. (2002). Real-Time Deformation Monitoring of Bridges Using GPS/Accelerometers. PhD thesis, IESSG, The University of Nottingham, Nottingham, UK. Meng, X., M. Meo, G.W. Roberts, A.H. Dodson, E. Cosser, E. Luliano, A. Morris (2003a). Validating GPS Based Bridge Deformation Monitoring with Finite Element Model. In: Proc of GNSS 2003, The European Navigation Conference, 22-25 April, Graz, Austria. Meng, X., G.W. Roberts, E. Cosser, A.H. Dodson, J. Barnes, C. Rizos (2003b). Real-time Bridge Deflection and Vibration Monitoring Using an Integrated GPS/Accelerometer/Pseudolite System. In: Proc of 11th International Symposium on Deformation Measurements, International Federation Surveyors (FIG), Commission 6 Engineering Surveys, Working Group 6.1, 25-28 May, Santorini, Greece. Roberts, G.W., A.H. Dodson, and V. Ashkenazi (1999). Twist and Deflection: Monitoring Motion of Humber Bridge, GPS World, 10 (10), pp. 24-34. Roberts, G.W., A.H. Dodson, C.J. Brown, R. Karuna, E. Evans (2000a). Monitoring the Height Deflections of the Humber Bridge by GPS, GLONASS and Finite Element Modelling. In Proc of IAG Symposia Geodesy Beyond 2000, edited by Schwarz, Springer-Verlag, Berlin, pp. 355-360. Roberts, G.W., X. Meng, C.J. Brown (2000b). The Use of GPS for the Measurement of Bridge Movements: A Viability Study. The University of Nottingham and Brunel University. Roberts, G.W., X. Meng, A.H. Dodson (2001). Data Processing and Multipath Mitigation for GPS/Accelerometer Based Hybrid Structural Deflection Monitoring System. In: Proc of ION GPS 2001, 14th Int. Tech. Meeting of the Sat. Div. of the U.S. Inst. of Navigation, 11-14 September, Salt Lake City, USA, pp. 473-481.