1. Structural Health Monitoring of Chulitna River Bridge
J. Leroy Hulsey1, J. Daniel Dolan2, Feng Xiao1 1University of Alaska Fairbanks, 2Washington State University
Introduction
There are total of 73 sensors installed on this bridge. The sensor arrangement (Figure 2) provides information about changes in the load distribution in the girders and the trusses. Most of sensors located in the places which have low load rating factors, others are used to indicate the load distribution of the bridge.
Results show several frequencies that we were able to characterize. These were 1.500, 2.190, 2.846, 3.224, and Hz [3]. The frequencies 2.856, 3.224, and Hz are the bridge vertical modes. The frequencies and Hz are the longitudinal modes. The measured results correlate with FEM with the error of -10.2%.
According to the updated results, the largest error the in the transversal direction reduced from % to -19.9%. There are only two roller bearings no in connect with the supports. Three roller bearings work as springs in the vertical directions.
In the global level, the updated FEM’s natural frequencies in three directions were compared with the measured results. The largest error is 8.9% in the longitudinal first mode. The updated FEM’s error is acceptable in the global-level.
In August of 2012, Alaska University Transportation Center (AUTC) was requested by the Alaska Department of Transportation & Public Facilities (AKDOT & PF) to install the structural health monitoring system on the Chulitna River Bridge along the Parks Highway outside of Trapper Creek, AK. The objective of this study is to provide important information for structural condition assessment of the Chulitna River Bridge.
The instrumentation of this project began on August 18th and finished on September 9th. Load test was performed on September 10th. There are 73 sensors, included strain sensors, rosettes, displacement sensors, temperature sensors, accelerometers and tilt meters, measuring static and dynamic load of the three AKDOT & PF dump trucks.
The bridge finite element model (FEM) was updated based on HDR’s FEM. There are 18 variables selected to be updated. There are 4 flange of girders, 6 flange of stringers, 3 composited truss lower chords, elastic modulus of concrete deck, cross frames and 3 spring supports.
The objective function is the error between the field measurement data and the FEM analysis data. The objective functions are divided into two categories: the global-level and the local-level objective function. The global-level objective functions are the errors between the measured frequencies and the FEM analyzed frequencies in three different directions. The local-level objective functions are the errors between the measured stresses and the FEM analyzed stresses. The sensors were installed in the cross section of mid-span, the pier 3 and the pier 5. In mid-span, strain sensors were installed on the longitudinal members (the girder, the stringer’s flange and the composited trusses lower chord) [1] in order to check the bridge longitudinal behavior. In the pier 3 and 5, strain sensors were installed on the cross frame’s diagonal members and displacement sensors were installed on the roller bearings in order to check the bridge transversal behavior.
Conclusions
The FEM’s updating is divided into two stages. Firstly, FEM updating in the longitudinal direction. Secondly, FEM updating in the transversal direction. The updated FEM’s accuracy was check in both local- and global level.
The longitudinal member such as the girder flanges, the stringer flanges, the composited trusses lower chord’s cross area and the elastic modulus of concrete deck are selected to be updated. The three paralleled trucks mid-span loading’s measured results compared with the FEM analyzed results work as the objective functions. Updated variable were adjusted in the reasonable range to minimize the objective function. In the global-level, the vertical and the longitudinal natural frequencies also be checked.
In the transversal direction, the unconnected roller bearings and the cross frames were selected to be updated. Two trucks stopped at two critical cross sections in load test. Compared with the FEM analyzed results can generate a new objective functions. After model updating, both local- and global-level results shows lower errors in the transversal and the longitudinal direction.
In the local-level the largest error reduced from % to -19.9%. In the global-level, the largest error reduced from -10.2% to 8.9%.
The next stage of research is to use mathematical optimization method to update the FEM. Briefly, the FEM will be transferred into mathematical model. The more objective functions will be selected from both static tests and dynamic tests. More variables will choose to be updated. The objective functions and variables will be set in reasonable ranges. The optimized results will be calculated based on mathematical optimization methods. More object functions and variables will ensure the reliability of updated model. Optimized updated method can control the objective functions intelligently so it is reasonable to achieve lower error in the new optimized model. The new optimized model can indicate bridge as-is condition which will help to further evaluation this bridge.
Figure 4: Two Trucks Load Test
Figure 2: Sensor Layout
Structural Health Monitoring System
The structural health monitoring system is composited by five parts: the sensors, the sensor multiplexer, the sensor interrogator, the local computer and the remote computer (Figure 3). Sensor Multiplexer is located in the control panel at the bridge with controlled temperature and humidity. The sensor interrogator and the local computer are located in the control panel at Princess Hotel.
FE Model Updating
Model Updating in the Longitudinal Direction
There are 13 fiber optic strain sensors installed in the mid-span of bridge [1] to measure the three paralleled DOT truck’s dead load [4].
HDR FEM’s mid-span loading results and it indicates the FEM’s composite trusses lower chords have larger stresses than measured results [4]. The FEM’s composite trusses stringers are too soft than the as-is condition. In consider of those problems, 14 variables were selected to adjust the load distribution in the composite trusses and the girders.
According to the HDR FEM longitudinal behavior, the largest error exists in the lower chord member and reducing the cross section area of lower chord to 0.43 will decrease the local-level error below 50 %. Then the largest error transfer to the composite truss lower flange. To change the elastic modulus of concrete deck to 3,000 ksi will improve the local accuracy to 5%, however, the global error will come up to 15% and the error changes from too stiffer to too softer. In order to balance the error in the local- and the global-level, the elastic modules of concrete deck changed to 3,300 ksi and the stringer lower flanges’ cross section area changed to 2.5.
The largest error in the global-level reduced from -10.2% to 8.80 % and in the local-level reduced from -66.4% to %. The global measured data is from force balanced accelerometers in the ambient test; the local measured data is based on the 13 fiber optic strain sensors’ measurement in the mid-span.
Model Updating in the Transversal Direction
The stiffness of cross frame and supports condition determined the load distribution in the transversal direction. In the HDR’s report, there are 5 roller bearings not fully connect with the supports and the HDR FEM also removed those supports from the original FEM [5].
There are 5 displacement sensors were installed in those locations to measure the movement of roller bearing in the vertical direction and 8 strain sensors were installed in the diagonal members to measure the reaction of the supports and the cross frames.
According to the displacement sensor’s results, three roller bearings have limited movement in the vertical direction. others are more flexible in the vertical direction.
In order to show the reaction between the cross frames at supports. There are 8 strain sensors installed on the cross frames in those five unconnected roller supports locations [1].
Comparing the field measurement and the FEM, there are large errors exist in the cross frame stress results. The largest error is -43.4% in the cross section of pier 3. In the cross section pier 5, the largest error is %.
Displacement sensors indicate the bearing 1, 3 and 5 have limited movement in vertical direction. So it may work as a semi-rigid supports at those locations. In model updating, there are 3 spring supports added on those locations. The stiffness of spring supports and the cross frames updated according to reduce errores in the objective functions.
Bridge Description
The Chulitna River Bridge, built in 1970, is located at Historic Mile Post on the Parks Highway between Fairbanks and Anchorage, Alaska. The original bridge was a 790-foot long, 5 spans continuous bridge with two exterior steel plate girders and three sub- stringers. It had a cast-in-place concrete deck 34 feet wide. In 1993, the bridge deck width was increased to 42 feet 2 inches by replacing the original cast-in-place deck with precast concrete deck panels. To accommodate the increased loads, the two original exterior plate girders were strengthened, three new longitudinal steel trusses were installed utilizing the original stringers as top chords, and the steel bracing was added to the piers (Figure 1).
References
Figure 3: Structural Health Monitoring System
[1] Hulsey, J. L., Brandon, P. and Xiao, F. (2012), “Structural Health Monitoring and Condition Assessment of Chulitna River Bridge: Sensor Selection and Field Installation Report”, Alaska Department of Transportation Research, Development, and Technology Transfer, report, Fairbanks, Alaska.
[2] Hulsey, J. L., Brandon, P. and Xiao, F. (2012), “Structural Health Monitoring and Condition Assessment of Chulitna River Bridge: Load Test Report”, Alaska Department of Transportation Research, Development, and Technology Transfer, report, Fairbanks, Alaska.
[3] Xiao, F., Chen, G. S. and Hulsey, J.L. (2012), Experimental Investigation of a Bridge under Traffic Loadings, International Conference on Frontiers of Mechanical Engineering, Materials and Energy (ICFMEME 2012), Dec. 20,  Beijing, China. 
[4] Hulsey, J. L., Xiao, F. (2013), “Structural Health Monitoring and Condition Assessment of Chulitna River Bridge: Mid-Span Loading Report”, Alaska Department of Transportation Research, Development, and Technology Transfer, report, Fairbanks, Alaska.
[5] HDR, Load Rating and Structural Assessment Load Rating Report-Bridge No. 255: ChulitnaRiver Bridge, 2011.
Load Test & Ambient Test
Two and Three Trucks Mid-span Loading
We conducted a total of 17 different tests (trials) using various combinations of the test trucks. The testing program content both the static tests and the dynamic tests [2]. The bridge response was also recorded under normal daily traffic between the tests.
The mid-span loading report [4] is documented both the experimental and the analytical results in the mid-span loading condition. Figure 4 shows two trucks load test.
Ambient Tests
A set of dynamic field tests were conducted on the Chulitna River Bridge using an ambient free-decay response approach. This approach was used to estimate the dynamic properties of the bridge. The identified natural frequencies were characterized by the Fast Fourier Transform (FFT) methods.
Figure 1: Chulitna River Bridge
SHM System
Sensor Layout
The sensors layout addresses specific issues, i.e. overstress of the plate girder, load transfer through the cross frame, load distribution in the girders and the trusses and the truss bearing not in contact with the supports.