Along-Wind Vibration of a Suspension Bridge Tower and Girder: Full-Scale Measurement


Dionysius M. Siringoringo1, and Yozo Fujino2 Department of Civil Engineering, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan 2 Department of Civil Engineering, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, Japan Email:,

ABSTRACT: Monitoring and permanent instrumentation of long span suspension bridges is an excellent opportunity to gain insight on the actual bridge vibration under various loading conditions. Vibration data collected during seismic, strong wind events, and ambient vibration has been used to monitor the bridge condition. This paper presents a study of a particular type of vibration observed on a suspension bridge tower. Monitoring system revealed a unique tower in-plane vibration recorded under moderate wind velocity between 13m/s and 24m/s. At this wind velocity range the downstream leg of the tower oscillates on its strong axis. The along-wind oscillation is characterized by a single-frequency harmonic-like response that resembles the in-line vortex shedding response. Influence of the tower in-plane vibration also appeared on the girder, where significant increase in girder lateral vibrations was recorded within the identical wind velocity range. While vortex shedding of bridge tower is not uncommon during construction when tower is in a free-standing stage, its occurrence on a tower of a completed bridge especially on its strong axis is rarely observed. KEY WORDS: wind-induced vibration, full-scale measurement, suspension bridge, tower vibration, vortex-induced vibration, along-wind vibration. 1 INTRODUCTION are made of steel box and connected with welding. Both towers are 132.5m high and 21m wide, giving a sag ratio of 1:10. The tower leg has a dimension of 5m x 3.6m on the base and gradually tapered to 3.2m x 3.6m on the top. The three bridge spans are discontinuous, connected by bearings and extension devices, and simply supported at the towers.

The recent trend on monitoring and permanent instrumentation of long-span bridges has given an excellent opportunity to gain insight on the real bridge performance under various conditions. One of such bridges is the Hakucho Suspension Bridge (HSB). The bridge is instrumented permanently with wind and seismic monitoring system and the monitoring system has been in operation since its opening in 1998. Vibration data collected during seismic and strongwind events as well as ambient excitation has been used to monitor the bridge condition. This paper presents the observation results on wind-induced vibration during six strong-wind events, where a wide range of wind velocity up to 30m/s was recorded. During these occasions, significant tower in-plane and girder lateral vibrations were measured. Characteristics of the tower in-plane and the girder lateral vibration and their relationships with wind velocity are studied in detail in this paper. 2 DESCRIPTION OF BRIDGE AND MONITORING SYSTEM

HSB is the largest suspension bridge in the eastern part of Japan. The bridge, located in Muroran Gulf Hokkaido Prefecture northern part of Japan, connects the Muroran Port in the south and Muroran City in the north. Total length of the bridge is 1380m consisting of 720m center span and two symmetric side spans of 330m each (Figure1). Bridge towers

Figure 1. Hakucho Suspension Bridge and Sensors layout for monitoring system The girder is a streamlined steel box with the width of 23m and maximum web height of 2.5m. The bridge was opened to public on June 13, 1998. Since located at the entrance of a bay, the bridge is exposed to relatively strong easterly wind especially from the Uchiura Bay on the west side. The bridge has permanent wind and seismic monitoring system that consists of 27 channels of vibration sensor placed on fourteen locations. Vibration sensors consist of twenty-two channels of uniaxial accelerometer, two uniaxial displacement sensors, and a triaxial free-field strong-motion accelerometer. To monitor velocity and direction of the wind, two ultrasonic anemometers (DA-600, Kaijo Denki) were installed on the center of mid-span and on the top of north tower (they are referred to as F1 and F2, respectively hereafter). All sensors measured the responses simultaneously and the data were recorded every 10 minutes with the sampling frequency of 20Hz. Six data sets of wind and bridge response are analyzed in this paper. Of the six sets two data sets are from measurement in 1999; namely March 6, 1999 (990306), and March 22, 1999 (990233), and the other four are from measurements in December 2005;25,26,27,and 28)(i.e. 051225, 051226,051227, and 051228). The wind velocity and direction recorded on the six occasions are measured by anemometer. The averages wind direction for most of the strong wind is around 250-300o. These easterly winds blow in the direction that almost orthogonal to the bridge deck (the bridge transverse axis is about 275o from the north). 3 GLOBAL STRUCTURAL BEHAVIOUR In order to clarify the global behavior during wind-induced vibration, the modal identification is conducted. For this purpose the Natural Excitation Technique (NExT) Eigensystem Realization Algorithm (ERA), and Empirical Mode Decomposition (EMD) - Hilbert-Huang Transform (HHT) are employed. Previous research [1] has shown the merit of the NEXT-ERA system identification for the structure with closely-spaced modes such as a suspension bridge. In addition, the EMD-HHT approach is implemented in this study to enhance the performance of modal identification amid the presence of nonstationarity response during strong wind excitation.

Table 1 shows the complete results of identification from girder accelerations. It is well-known that the natural frequencies and damping ratios vary with respect to wind velocity as shown in the previous research [1] so that the results listed in Table 1 are representative example from one data set recorded at wind velocity 24m/s. The system identification generates 12 modes in the frequency range of 0 – 1.5Hz, where the girder mode shapes dominate the overall bridge mode shapes. In this frequency range, the modes consist of one lateral mode, eight vertical modes and three torsional modes. The lowest mode was found to be the lateral mode at 0.089Hz, while the first vertical and first torsional mode was 0.12Hz and 0.502Hz, respectively. In general natural frequencies identified from strong wind responses agree well with those analytically predicted by FEM. In addition to global modes, the system identifications also give two modes with local mode shapes with tower dominant lateral modal displacement identified at 0.6Hz and 0.8Hz. As will be shown later on, the two modes correspond to the occurrence of single-frequency in-plane response of the tower. The maximum modal displacements of these modes are located in-plane on the top of the tower. Both modes seem to be pure tower local in-plane mode, accompanied by small coupling with girder in lateral direction. Table 1. Identified global modal parameters under wind velocity 24 m/s Modes Frequency (Hz) Damping (%) NEXTNEXTHHT FEM HHT ERA ERA 1-L-S 0.089 0.090 0.09 4.07 2.31 1-V-S-B 0.119 0.121 0.126 3.00 2.82 1-V-A-B 2-V-S-B 2-V-A-B 3-V-S-B 1-T-S 3-V-A-B 4-V-S-B 2-T-A 5-V-S-B 0.148 0.218 0.318 0.444 0.502 0.570 0.734 0.813 1.105 0.153 0.220 0.319 0.449 0.504 0.584 0.737 0.813 1.110 0.151 0.220 0.322 0.439 0.472 0.570 0.722 0.778 1.160 3.27 2.12 1.61 1.50 1.05 0.48 0.37 0.77 0.49 4.56 2.88 1.29 2.22 1.91 1.63 1.43 1.65 1.01

3-T-S 1.200 1.197 1.290 0.19 1.06 Note: V: Vertical, L: Lateral, T: Torsion, A: Asymetric, S: Symetric, B: Bending 4 WIND-INDUCED VIBRATION OF TOWER

The bridge towers are made of steel box and connected by welding. Both towers are 132m high, with the width of 21m on the base and 18m on the top. Each tower column has dimension of 5m x 3.5m on the base and gradually tapered to 3.2m x 3.5m on the top. Of the two towers, only the south tower is instrumented with accelerometers, whereas the north tower is instrumented with anemometer. For this reason response and wind characteristics at the same tower locations cannot be directly compared. However, investigation has shown that wind characteristics on both anemometers are very

similar. Accordingly, wind property of anemometers on the center of midspan (F1) at elevation 62m is utilized to characterize the wind in this analysis. Furthermore, considering that wind in direction normal to the bridge axis is dominant in most of occasions, this direction is taken as the standard wind direction in that bridge response normal to bridge axis is denoted as along-wind and the response parallel to bridge axis is denoted as crosswind. Acceleration responses analyzed in this study are obtained from two levels, namely, the deck level at elevation 42m (node AM2), and on the top of the tower (node AM3 and AM4). Of all accelerometers, only AM4 is placed on the west leg, the rests are on the east leg. To quantify bridge response, root-mean-square (RMS) of acceleration responses was computed for every ten minutes. The tower responses were analyzed separately in two directions: 1) out-of-plane, in which the tower oscillates in its weak axis parallel to the bridge axis due to crosswind, and 2) in-plane, where the tower oscillates on its strong axis due to along-wind in direction perpendicular to the bridge axis. 4.1 Relationship between wind velocity and tower out-ofplane acceleration

lower wind velocity (less than 13m/s) and higher wind velocity (more than 24m/s), the relationship is somewhat similar to that of the crosswind motion, where the acceleration increase proportionally with wind velocity following a quadratic equation. However, for wind velocity between 13 and 24m/s there is a sudden acceleration increase. To clarify the responses characteristics of the three wind regions, time and frequency domain characteristics of the responses are analyzed. The results are as follow. 1. In the first wind velocity region (<13 m/s), accelerations are quite random with relatively small amplitude. Frequency spectra are characterized by multiple peaks dominated by four frequency peaks within the range of 0-2Hz (i.e. 0.48 Hz, 0.6 Hz, 0.8Hz and 1.01 Hz). In the wind velocity region 3 (>24 m/s), the accelerations amplitude become larger than that of region 1 but generally smaller than those in region 2. The timedomain responses have similar random characteristics as in wind region 1. The frequency domain response is characterized by multiple frequency peaks similar to the peaks that appear in the first wind region. In short, similar the first region, the response can be typically described as buffeting (Figure 4). In the wind region 2, where the wind velocity is between 14 and 24 m/s, the responses are characterized by the dominant single frequency peak at either 0.6Hz or 0.8 Hz. Figure 5 shows representative examples of the response. In this region the time domain acceleration response is relatively constant for several minutes and the responses of AM3Y and AM2Y are in-phase with constant amplitude proportion.

Figure 2 shows the relationship between tower out-of-plane response and the across-wind velocity. From the three locations of accelerometers, a clear trend was observed in which the tower out-of-plane acceleration increases as the wind velocity increases. Evaluation of the trend shows that the relationship between wind and tower out-of-plane motion is similar to what commonly expected in buffeting response.


Figure 2. Tower out-of-plane acceleration w.r.t. wind velocity. 4.2 Relationship between wind velocity and tower in-plane acceleration

Figure 3. Tower in-plane acceleration w.r.t. wind velocity. The characteristics of tower in-plane motion for all available record are examined and the results are categorized into four categories (Figure 6): 1) random-like response with multiple frequency peaks, 2) harmonic-like response with single-frequency peak at 0.6Hz, 3) harmonic-like response with single-frequency peak at 0.8Hz, and 4) response codominated by two frequency peaks (0.6Hz and 0.8Hz). It was observed that:

Figure 3 shows the relationship between tower in-plane acceleration and the alongwind mean velocity in term of RMS. Similar to the out-of-plane acceleration the in-plane acceleration increases as the wind velocity increases. Note however, there are three regions with noticeable patterns. At





The in-plane motions with random-like multiple-peaks response characteristic generally have smaller RMS compared to other responses in the same wind velocity region. Relationship between the RMS of acceleration with respect to wind velocity of the random-like response can be described using quadratic equation similar to the tower out-of-plane motion. As explained previously, in wind region 2 (13-24 m/s), the responses were characterized by harmonic-like single-frequency response at 0.6Hz or at 0.8Hz. Response with latter frequency appear at the higher wind velocity range (17-24 m/s), while the former appear at lower wind velocity range (13–17 m/s). The response with single-frequency peak 0.6Hz and 0.8Hz occur at the reduced wind velocity Vr = V / fD around 7-8. All responses with 0.8Hz single-frequency peak were caused by wind originated from west-southwest quadrant and confined within 10 to 30 degree away from the bridge transverse axis. Meanwhile, most of the responses with 0.6Hz single-frequency peak were caused by wind originated from west-northwest within 20 degree of the bridge transverse axis.

Figure 5. Example of tower in-plane acceleration with single frequency peak characteristics measured at wind region 2: (a) 0.6Hz dominant-response and (b) 0.8Hz dominant response Most of the responses with single-frequency characteristics have low turbulence intensity (i.e. around 10%). The low turbulence intensity is attributed to the fact that all of the responses were excited by the wind that come from the open sea at Uchiura Bay.

Figure 4. Example of tower in-plane acceleration with multiple frequency peaks characteristics measured at: (a) wind region 1, (b) wind region 3. Figure 6. Tower in-plane acceleration w.r.t. wind velocity categorized based on frequency characteristics.(MF: multiple frequency peaks, SF 0.6Hz: Single Frequency peak at 0.6Hz, SF 0.8Hz: Single frequency peak at 0.8Hz, DF 0.6Hz&0.8Hz: Double Frequency peaks at 0.6Hz and 0.8Hz)

Figure 7. Girder lateral acceleration of the main-span w.r.t. wind velocity 5 WIND-INDUCED VIBRATION OF GIRDER 6

Figure 8. Girder lateral acceleration of the side-span w.r.t. wind velocity COMPARISON WITH FINITE ELEMENT MODEL AND TOWER DISPLACEMENT

The relationship between vertical girder acceleration and the wind velocity is similar to the response due to buffeting observed on other long-span bridges. The maximum rootmean-square (RMS) of vertical acceleration reached up to 16cm/s2, and 12cm/s2 for the midspan and side span, respectively. In lateral direction the girder acceleration increases with the increase in wind velocity. However, the increasing trend is also not proportional for all wind velocity range. There exist wind velocity regions, where abrupt change in acceleration was observed. For example, girder responses for the wind velocity region 14–24m/s and 17-24 m/s for AK1Y and AK5Y, respectively are significantly higher than girder response for other wind ranges. For the midspan acceleration (AK1Y), the girder lateral accelerations increase significantly after 13m/s and reach the peak at 17m/s. For the sidespan acceleration (AK5Y), the rms increases abruptly after 17m/s and reaches the peak at 20m/s. For both spans, the increase in acceleration is confined within the wind velocity of 14-24 m/s. This wind region coincides with the wind region where the South tower experiences the single-frequency dominant in-plane motion. Frequency spectra of the girder lateral acceleration during this wind region reveal the presence of frequency peaks at 0.6Hz and 0.8Hz. These peaks dominate the girder lateral response only at this wind velocity range, and do not significantly appear at other wind ranges, indicating an apparent relation to the tower in-plane motion.

Finite element model is utilized to explain the tower and girder lateral vibration. The model indicates that in the vicinity of 0.6Hz and 0.8Hz exist two local modes whose modal displacement are dominated by tower in-plane motion (Figure 9). In both modes, the highest modal mass ratio that indicates the largest modal displacement is from the tower, followed by suspension cables. The first mode is 0.603Hz mode, which is dominated by tower in-phase and in-plane motion with small participation of motions from other structural components. This mode is also characterized by noticeable lateral modal displacement in the midpoint of the main span. The other tower dominant mode is the 0.775Hz mode. It is a local tower mode characterized by tower out-of-phase motion. This mode has a stationary point (i.e. zero lateral modal displacement) in the middle of girder main span. Unfortunately, the in-phase and out-of-phase characteristics of the finite-element modes of the north and south towers cannot be confirmed by measurement since only the south tower is instrumented with accelerometers. The difference in girder lateral modal displacement at the midpoint of the main span between the 0.603Hz and the 0.775Hz mode explains the condition depicted in Figure 7. When wind excites the 0.6Hz mode (velocity 13-17m/s) significant increase is observed on the girder lateral vibration, since the mode has the lateral modal displacement at the midpoint of the main span. Meanwhile during the 0.8Hz motion (wind velocity 17-24 m/s), there is no significant increase on the girder lateral vibration, given that this mode has a zero modal displacement at the main span midpoint.

Figure 9. FEM-generated modes with dominant modal displacement of towers that correspond to tower in-plane motions observed in the measurement. (Both mode shapes are viewed from top) Since the tower maximum in-plane accelerations are mainly caused by the single-frequency oscillation, the maximum displacement can be computed using the following relation:η = a /(2π fi ) 2 , where η is the maximum amplitude of displacement, a denotes the maximum amplitude of acceleration, and f i is the observed single frequency of the tower (i.e. f=0.6Hz and 0.8Hz). From observation, the maximum in-plane acceleration recorded was 28cm/s2 and 32cm/s2, which is due to the single-frequency oscillation at 0.6Hz and 0.8Hz, respectively. Therefore, the maximum displacement amplitude calculated by the above equation is about 2cm and 1.3cm for 0.6Hz and 0.8Hz, respectively. Considering the height of tower 132.5m the single-frequency in-plane oscillations only produce small in-plane tower inclination, which suggests that although the in-plane singlefrequency oscillation creates significant increase of acceleration, it does not produce an excessive displacement. 7 DISCUSSION

vibration of the downstream structure. As pointed out by Zdravkovich [3] for cylindrical structures and by Godwa et al.[4] for rectangular structures, the occurrence of this flowinterference oscillation depends on the ratio between longitudinal and transversal spacing between the structures. In the case of Hakucho Bridge tower the ratio between longitudinal spacing of the tower legs (L) and the width of the tower in direction normal to the wind (D) (L/D ratio) is around 5.5. In transverse direction, two spacings (T) may be considered since the 0.6Hz and 0.8Hz single-frequency oscillation are excited by winds from different directions, namely 290o and 250o, respectively. These give the ratio (T/D) of 1.3 and 2.1 for 0.6Hz and 0.8Hz single-frequency oscillation, respectively. With these ratios, the downstream tower leg may experience oscillation caused by wake interference. The downstream tower leg was submerged in the wake created by vortex-shedding behind the upstream tower leg, and this caused fluctuation of drag force. When the downstream tower leg is fully submerged in the wake the drag force decreases. Meanwhile the force increases when it is less submerged. This fluctuation is thought to cause significant streamwise in-plane oscillation of the tower and was transferred to the girder through the suspension cable. Further investigation is still carried out by wind tunnel testing and simulation on computation fluid dynamics. As a side note, this study shows that through monitoring and permanent instrumentation of long span bridges, one can gain insight into vibration of bridge component more comprehensively and examine various types of vibration that are not commonly observed. The lessons learned from monitoring are expected to enhance our understanding in vibration of long-span bridges. ACKNOWLEDGMENTS The authors gratefully acknowledge Mr. Shigehiro Fukaya from Chodai Engineering Consultant for discussion on the finite element model. REFERENCES
[1] [2] Siringoringo DM and Fujino Y (2008). System identification of suspension bridge from ambient vibration response, Engineering Structures Vol.30 (2), pp. 462–477. Larose GL, Zasso A, Melelli S, and Casanova D, 1998, Field measurements of the wind-induced response of a 254 m high freestanding bridge pylon, Journal of Wind Engineering and Industrial Aerodynamics Vol. 74-76,pp 891-902. Zdravkovich, MM, 1985, Flow-induced oscillations of two interfering circular cylinders, Journal of Sound and Vibration Vol.101, pp.511–521. Gowda BHL and Kumar RA, 2006, Flow-induced oscillations of a square cylinder due to interference effects, J. Sound and Vibration Vol.297, pp.842-864

The single-frequency harmonic-like excitation that occurs on a relatively moderate wind velocity is somewhat similar to the condition of resonance often observed during vortex-shedding oscillations. While vortex shedding of bridge tower is not uncommon during construction when tower is in a freestanding stage, its occurrence on a tower of a completed bridge, especially on its strong axis is not commonly observed. A similar phenomenon of single-frequency alongwind vibration of a long-span bridge tower is reported by Larose et al. [2] from observation on the Storebaelt Bridge tower during construction free-standing stage. Wake interference of the downstream structure due to staggered arrangement of the tower legs is thought as a possible mechanism behind the tower single-frequency oscillation. This phenomenon has been observed on the vibration of two tall cylinders placed in proximity, in which flow separation that occurs on the upstream structure excites

[3] [4]