Research Report FHWA Contract DTFH61-03-C-00104 Research Task No.

1 Wind Response and Health Monitoring of the Tacoma Narrows Bridge

Frequency Analysis of the Existing Tacoma Narrows Bridge

by Rafik Itani and Adel Al-Assaf Washington State Transportation Center (TRAC) Washington State University Department of Civil & Environmental Engineering Pullman, WA 99164-2910

Prepared for The Federal Highway Administration U.S. Department of Transportation Hamid Ghasemi, Project Technical Manager

June 2006

DISCLAIMER
The contents of this report reflect the views of the authors, who are responsible for the facts and accuracy of the data presented herein. The contents do not necessarily reflect the official views or policies of the Federal Highway Administration. This report does not constitute a standard, specification, or regulation.

TABLE OF CONTENTS
Introduction ...................................................................................................................................................................1 Scope .........................................................................................................................................................................1 Benefits......................................................................................................................................................................1 Study Significance.....................................................................................................................................................3 The Tacoma Narrows Bridge.........................................................................................................................................4 History .......................................................................................................................................................................4 Bridge Description.....................................................................................................................................................5 Finite Element Model ....................................................................................................................................................8 Towers .......................................................................................................................................................................9 Stiffening Truss .......................................................................................................................................................10 Floor beams .............................................................................................................................................................10 Main Cable ..............................................................................................................................................................11 Hangers....................................................................................................................................................................16 Material....................................................................................................................................................................17 Section Properties ....................................................................................................................................................17 Boundary Conditions...............................................................................................................................................17 Nonlinear elements ..................................................................................................................................................19 Load and Mass Estimation.......................................................................................................................................21 Estimation of Mass ..................................................................................................................................................22 Frequancy Analysis .....................................................................................................................................................24 Ambient Study.........................................................................................................................................................24 Eigenvalue Analysis ................................................................................................................................................25 Model Calibration....................................................................................................................................................25 Results .........................................................................................................................................................................26 Conclusion...................................................................................................................................................................30 Appendix A .................................................................................................................................................................32 Eigenvalue Analysis ................................................................................................................................................32 Modal Participation Masses (%)..............................................................................................................................33 Mode Shape Identification.......................................................................................................................................34 References ...................................................................................................................................................................36

ii

LIST OF TABLES
Table 1: Comparison between the Experimental and the Analysis Frequencies .........................................................26 Table 2: Modal Frequency Identification of the TNB .................................................................................................27

iii

LIST OF FIGUERS
Figure 1: Section of the Second Tacoma Narrows Bridge Suspended Structure..........................................................5 Figure 2: Exiting Tacoma Narrows Bridge Elevation View..........................................................................................6 Figure 3: The Detailed Finite Element Model of the Tacoma Narrows Bridge.............................................................8 Figure 4: TNB Main Tower Elevation and Side view. ..................................................................................................9 Figure 5: Tacoma Narrows Bridge Floor Truss...........................................................................................................10 Figure 6: Cable Profile Comparison ............................................................................................................................13 Figure 7: Difference between Catenary Profile and Parabolic Profile Along the Main Span Length .........................15 Figure 8: Normalized Tension in Main Cable .............................................................................................................15 Figure 9: Hanger Force Model ....................................................................................................................................16 Figure 10: Hanger Force..............................................................................................................................................17 Figure 11: Modelling of Windshoe top view at deck level.......................................................................................20 Figure 12: First Mode Period vs Normalized Density. ................................................................................................21 Figure 13: Effect of Eliminating Tower Contribution to Frequency Content..............................................................23 Figure 14: Mode shapes of Identified Modes ..............................................................................................................30

iv

INTRODUCTION

Scope This report is for the Wind Response and Health Monitoring of the Tacoma Narrows Bridge project. It summarizes the procedure taken to develop and calibrate a detailed finite element model for the existing bridge and includes the frequency analysis results.

The project was to design a health mentoring and wind response program for the existing Tacoma Narrows Bridge and to proceed with the initial stage of implementation on the existing structure. This required the modelling of the bridge to identify its structural properties and behaviours. The results will then be employed in the wind response study.

Benefits There are two ways in classical wind engineering, to conduct a wind response study of long-span bridges. Wind tunnel testing, which requires scaled prototype of the structure, and analytical methods, developed by Scanlan and Jones in 1971, which require experimental data extracted from wind tunnel testing for a representative section of the superstructure. Both approaches estimate the flutter condition and could be used for design or retrofit.

Recently, solving the Navier-Stokes equations became possible due to the advanced processing power of modern computers. This opened a new possibility to conduct a wind response study using finite element procedure, where the fluid (air) and the solid (structure) are

simultaneously coupled and solved in one matrix form. Although this approach is very promising, it is still under investigation, Brar et. al.

Health monitoring became a contemporary approach in wind response analysis. While it is not a design tool, it is useful to predict the need of retrofit and to assure the structural safety and serviceability. This involves installing system of sensors that measure the vibration and the deflections in the structure, and measure environmental parameters, such as, wind speed and temperature. It also might include surveillance cameras and traffic survey devices. Such instrumentation is then connected to a real-time computer system, the data acquisition system, that collects data based on certain time pace. The data is then continuously transmitted to anther computer system for filtration and storage. A specialized computer system, called global central computer system, manipulates the filtered data and translates them to tangible information, that vary from technical quantitative results, such as stress accumulation in the structural members, to general qualitative results, such as operation condition.

In order to create a relevant global central computer system, the behaviour of the structure under consideration should be programmed. Several techniques are possible to implement the system. This varies from stochastic models to deterministic models, which might be integrated among each other, and incorporated with the collected data. For bridge health monitoring, a detailed finite element model is a very fundamental requirement that could be used part of the decision making tools or could be used to train other models the system.

Study Significance Several studies were conducted on the existing Tacoma Narrows Bridge by various engineering firms. OPAC Consulting Engineers conducted a finite element modelling study in 1993 [3]. The study includes complete calculations of the geometric properties of the structural elements and estimations of the initial forces in the cables. The finite element model was then developed using SAP 2000.

Arvid Grant Associates and OPAC-Geospectra conducted a study, in 2003, on seismic evaluation of the Tacoma Narrows Bridge [1]. The study included identification of seismic hazards at the bridge site, identification of response frequencies of the structure, analysis of the bridge under ground motion and identification of structural deficiencies. Ambient vibration measurements, which provide experimental frequencies and estimations of the structural damping, are incorporated. These studies were found to be very useful to compare the results obtained from this research.

The formulation and analysis used here are to provide methodologies for assessing wind response of bridges. The model is used to assess the impact of structural alterations such as closing the open-grate segments along the deck without adversely affecting the wind response characteristics. While these alterations affect the aerodynamic characteristics the structural properties, such as frequency content, remain significantly unchanged.

For health monitoring and wind response, the structure operates within the elastic limit. Therefore, the analysis assumes linear elastic properties of the material and only geometric nonlinearity is considered in here.

The following report summarizes the detailed model used for the analysis of the existing Tacoma Narrows Bridge and provides discussion and conclusions of the research. Complete analysis results are found in the appendix.

THE TACOMA NARROWS BRIDGE

History The Second Tacoma Narrows Bridge was built after the catastrophic failure of its predecessor in 1940. When constructed, both were the worlds third longest span, surpassed by the Golden Gate Bridge and George Washington Bridge. The failure of the first bridge brought the attention to the approach of aerodynamic in the design of suspension bridges. Wind tunnel was used to assess the failure of the first bridge and in the design of the existing structure, see Farquharson 1954. Different design measures were considered in the new design to guarantee its sustainability in wind. The bridge, unlike the first one, has higher torsional frequency, which contributed significantly to its stability in wind to present time. Wind tunnel testing is now a standard design tool of long-span bridges.

Bridge Description Tacoma Narrows Bridge is a suspension bridge with main span of 2800 ft and two side spans of 1100 ft each, spanning the eastern shores of Puget Sound and Kitsap Peninsula with a four-lane roadway, with total width of 60 ft. The bridge is a segment of the state highway number 16, Washington State, travelling east-west.

The structure is made of steel, except for the roadway deck and the anchors, which are made of concrete. The superstructure is made of stiffening trusses which are 33 ft. deep and are supplemented by diagonal bracings. Series of floor beams, running along the span, carry the roadway deck and transfer the load to the hangers. Figure 1 shows the main structural elements of the superstructure, demonstrating the mentioned components. Some elements are eliminated to avoid view congestion. The roadway is not modelled in our study as explained in the modelling section of this report.

Figure 1: Section of the Second Tacoma Narrows Bridge Suspended Structure.

The two main towers, 467.04 ft high, support the main cable, which sags 280 ft. They stand on top of the retrofitted piers of the collapsed structure, which is dug embedded 200 ft in the narrows basin.

Figure 2 shows elevation view of the structure. Note that the structure consists of (viewed from west to east, i.e. from left to right), the following: one 162.5 ft west anchorage, three 150 ft steel deck girder approach span, 1,100 ft cable suspended steel side span, 2,800 ft cable suspended steel main span, 1,100 ft cable suspended steel side span, 45-2-1/2 reinforced concrete T-beam approach span, 42 5 reinforced concrete T-beam approach span, 45 reinforced concrete T-beam approach span, 45 reinforced concrete T-beam approach span and 185 ft east anchorage. The total length of suspended structure is 5,000 ft.

Figure 2: Exiting Tacoma Narrows Bridge Elevation View.

The deck width, curb-to-curb, is 46-8 1/8 which includes four 9ft lanes separated by 2.75 ft slotted wind grates and 1-7 wind grates separating the roadway from the sidewalks. In addition there are two 3.5 ft sidewalks one on each side of roadway and the width between the suspension cables is 60 ft. The two main cables are 20.25 ft in diameter.

Figure 4 shows the elevation and the side views of the main tower. The towers total length is 467.04 ft, measured from the pier face. The towers legs are made of steel segments, which are made of a built-up section of five rectangular champers arranged in cross shape. Tower legs are tapered. The first 141.5 ft has a parabolic tapering, 0.001x2 (ft), and the other segments are linearly tapered up to the top of the tower. The legs are connected with lateral and diagonal bracings.

The stiffening trusses are connected to the main tower legs at two points, as shown in the side view in Figure 4. A diamond-shaped truss element assembly embraces a giant damper, which is embedded inside each tower leg and connected to upper cord of the stiffening truss. Another assembly of truss elements connects the lower cord of the stiffening truss to the face of the tower leg, via a viscous damper as well. Both assemblies are designed to dissipate any excessive excitation along the longitudinal direction of the stiffening truss. The stiffening trusses are also connected to the middle part of the tower via the in plane upper cord bracings, which also connect the stiffening trusses together. A windshoe is designed at the point where the upper cord bracings are connected to the main tower lateral beam. The windshoe is simply a gap element which allows along the bridge deck and around the tower axis rotational movements; that is in the vertical direction, and restrains the other degrees of freedom.

The side spans have a grade of 3% , The main span is parabolic with 21 ft increase in elevation from the main tower centerline to the mid span, and it is made of 88 panels spaced at 30-10 5/8 on center. The spacing of panels in the side spans are changed to 30-3 3/8.

FINITE ELEMENT MODEL

A detailed finite element model is developed using MIDAS-Civil. The SAP 2000 was initially used. However, both the dynamic and the static analyses were not accurate. It is found that SAP2000 does not support explicit feature to handle the effect of the initial tension on the elements stiffness. Prior conducting any frequency or dynamic analyses, it is required to conduct a shape-finding analysis to determine the unstrained profile, and then a static stage construction analysis is conducted to build up the nonlinear geometric stiffness. MIDAS-Civil has very specialized features for cable elements and modelling of suspension bridge. As-builtdrawing is used with pre-calculated initial tension to account for the geometric stiffness and to conduct the dynamic analysis.

Figure 3: The Detailed Finite Element Model of the Tacoma Narrows Bridge

Figure 3 shows a general view for the finite element model with pull-outs for the main tower, the west tower and the superstructure. Note that the approaches, shown in Figure 2, are not included. The main cable is extended to where it terminates at the anchorage. The V-shape element, in the middle of the main tower, is a rigid element connecting the top cord of the superstructure with the tower. All the shown elements pass through the centre of mass of their corresponding structural elements.

Towers The cross sectional properties of each segment is calculated and assigned to the finite element model. The software used to model the bridge, MIDAS-Civil, supports tapered section
Windshoe location Damper

calculation. The cross section properties calculations proposed in the OPAC Consulting Engineers study, 1993, is adopted in the development of our model.
Tower Face Damper

All elements are modelled based on centreline. Therefore, offsets are Figure 4: TNB Main Tower Elevation and Side view. assigned to diagonal and lateral beam

bracings to account for the rigid end effect.

Stiffening Truss The stiffening truss, as described before, is divided into three spans, a main span with 2800 ft in length and two side spans with 1100 ft in length. The stiffening truss elements are modelled as beam elements in the analysis. The properties of the cross sections are calculated and compared with OPAC study [3] and are found to be accurate.

Floor beams In suspension bridges, floor beams connect the stiffening trusses and supports the roadway deck. In the Tacoma Narrows Bridge, the floor beams are, in fact, trusses. Previous studies condensed the structural properties of floor truss into a single beam element or a spine element, which is used in the global model. Figure 5: Tacoma Narrows Bridge Floor Truss.

While this reduces the number of the degree of freedom, the accuracy of the solution is also slightly diminished. Figure 5 shows the geometric configuration of the floor truss.

10

In this study the floor beams are considered in details. For simplicity in modelling and to guarantee stability in the numerical procedure of the finite element analysis, the floor truss elements are treated as beam elements. This assumption is verified using a separate model, such that, the maximum difference in axial load in each element due to the sustained loads as measured. The difference was found to be quite small and less than 8%. Lateral loads are assumed to be carried by the stiffening trusses, which are sufficiently braced by the floor trusses and the diagonal braces.

The roadway stringers, which carry the roadway slab and transfer its load to the floor truss, are eliminated. It is assumed that their contribution to the superstructure stiffness is negligible. However, they contribute to the gravity load and the mass. Removing these elements from the finite element model substantially reduces the computational time. Their weight and mass however are lumped with the upper cord of floor truss.

Main Cable Three main issues are related in modelling the main cable. These are: sectional property, shape or profile finding and internal forces. The last two issues are interrelated, since the internal tension force is a function of the cable sag, which in turn determines the profile of the cable.

The main cable is made of bundles of stands with each bundle consisting of several wires. The total cross sectional area of the cable is found by summing up the total area of the wires in the cross section. A 1.7 ft^2 net cross sectional area is used to model the main cable, with a

11

modulus of elasticity of 29,000 ksi. An alternative approach is to use the gross area and modify the modulus of elasticity of the cable element. This approach is also pursued and the effective modulus of elasticity is found to be 27,000 ksi, with a gross area of 1.8 ft^2. The results of modal analysis of both approaches are found to be the same. A parametric study is also conducted to examine the effect of cable cross sectional area on the modal response. Negligible differences are found in the modal responses and frequency contents with a 20% variation in cable cross sectional area.

Cable profile is the most important issue to be determined. In this study the final sag is know to be 280 ft. Thus, the internal forces due to the sustained loads in each of the cable element should be calculated accordingly. There are two types of cable profiles, catenary profile and parabolic profile. While it is known that a cable element exhibits a catenary profile, it has been suggested that cable element has a parabolic shape, as described by the earlier theory of suspension bridge text books, Pugsley 1950. In this study both profiles are calculated and are compared as shown below.

The profile, horizontal and tension forces along a parabolic cable can be completed by the following expressions:

4.s 2 x L2 w.L2 H= 8.s y=

Equation 1-a Equation 1-b

2 2

w.L L 2 x T= + 2 4.s L

Equation 1-c

12

where, y is the vertical distance, H is the horizontal force component, T is the tension force, s is the sag, L is the span length of the parabola, w is the total gravity load distribute uniformly all over the cable length, and x and y are the horizontal and vertical distances, measured according to the reference point shown in Figure 6.

The catenary shape is a hyperbolic sinusoidal function, or its equivalent exponent function, as shown:
e x / a + ex / a y = a.(cosh( x / a) 1) = a. 2 1

Equation 2

where, a is the shape parameter, which needs to be calibrated to obtain the profile shape. Note if the target profile is known, iterative procedure is required to obtain a. The coordinate system, x and y, is set up as shown in Figure 6.
Parabolic Profile Catenary Profile L/2

In order to obtain the shape parameter a in equation 2, a simple trial and error procedure is employed and the parameter is found to be 3545.708. Therefore, for the Tacoma Narrows Bridge, equations 1a and 2 are reduced to:
s

Figure 6: Cable Profile Comparison

1 x2 7000 y = 3545.708.(cosh( x / 3545.708) 1) y=

Equation 3 Equation 4

where, x and y are measured in feet.

13

Figure 6 shows a schematic sketch for the two cable profiles, showing one half of the main cable profile. The difference between the parabolic and the catenary profiles are very small and scaled sketch barely depict the discrepancy. Figure 7 shows the difference between the two profiles. It is assumed that the Catenary cable is the base line profile and the difference, in feet, is calculated along the span length as shown in Figure 7. The maximum difference between the two profiles is 0.905 ft at a distance of 0.7071 of half span length. If the maximum difference of the two profiles is normalized based on the sag, then it could be concluded that the difference between the two profiles is negligible.

A single panel of the superstructure was modelled with simple supports to estimate its weight. Based on the panel analysis, it is found that the average uniformly distributed dead load (w) is 10.4 kips/ft, that is 5.2 kips/ft per cable. Thus the total dead load carried by each cable is 15,162 kips. The Maximum tension expected based on Equation 1-c is 19,602 kips. From the finite element analysis it is found that the maximum tension is 20,277.1, which is 3% greater than the estimated value. The horizontal forces are estimated to be 18,200 based on Equation 1-b. From the finite element analysis it is found to be 18,948.72, which is 4% greater than the estimated value.

The error found between the estimated values from Equations 1 and the finite element analysis is considerable. The source of the error is due to using a catenary profile in the finite element analysis, while a parabolic profile is assumed in the derivation of Equations 2. Moreover, the used distribute load w in Equations 1 is based on averaged value, however, the finite element analysis considers all the variation in load distribution. This verifies the cable

14

finite element analysis and confirms that the implemented internal forces are accurate. The estimation of the cable internal forces has significant effect on the stiffness of the cables and thus the overall accuracy of the frequency analysis. Figure 8 shows the distribution of the ratio of the tension force in the cable to the total weight of the superstructure along half of the cable span.

While in suspension bridges, gravity loads are transferred to the main cable via discrete hangers, and hence applied as concentrated load, it is assumed here that load carried by the hanger is uniformly distributed gravity load in order to determine the shape of the cable.

1 Difference Between Profiles (ft)

1.36

Normalized Tension
Force / Total Structural Weight

0.8 0.6 0.4 0.2 0 0 200 400 600 800 1000 1200 1400 Main Span (ft)

1.34 1.32 1.3 1.28 1.26

Normalized Tension
1.24 0 200 400 600 800 1000 1200 1400

Distance ft

Figure 7: Difference between Catenary Profile and Parabolic Profile Along the Main Span Length

Figure 8: Normalized Tension in Main Cable

All elements are modelled based on members centreline. Offsets can be used to model rigid joint effect.

15

Hangers

Hanger force is estimated by iterative finite element analysis, using detailed finite element model. Figure 9 shows a schematic sketch for a section along the bridge of the model used to estimate the hanger forces. The main cable is removed and the hangers are supported with hinges along the cable profile. Initially the hangers are assumed to be truss elements. A linear static analysis is conducted based on that assumption. The truss elements are then converted to cable elements and the internal forces estimated in the truss elements are used as initial forces in the cable elements.

Iterative procedure is conducted to estimate hanger forces and then comparing them with the assumed initial hanger forces. Figure 10 shows the values and the distribution of the hanger forces. The two solid lines in the graphs resemble the location of the main towers. It could be seen that the final iteration has smoother force profile along main span. These

Hangers

Stiffening truss

Figure 9: Hanger Force Model values are assigned to the hangers in the global detailed finite element model.

16

160 150 140 Force (kips) 130 120 110

300 250 200 150 100 50

100 90 -1200

Force (kips)

-200

800

1800 Distance (ft)

2800

3800

0 -1200

-200

800

1800

2800

3800

Distance (ft)

(a) First Iteration For Hanger Force Figure 10: Hanger Force

(b) Final Hanger Force

Material

According to the Tacoma Narrows Bridge specifications the used steel is ASTM A7-46. The modulus of elasticity used is 29,000 ksi for all elements except the main cable and the hangers (suspenders).

Section Properties

The calculations done by OPAC [3] are adopted. The OPAC model considers spine element for the floor truss, and lump its sectional properties using hand calculation. This study, however, includes each floor truss element with its actual properties modelled.

Boundary Conditions

The boundary conditions used include the support system of the towers and the equivalent stiffness of the approaches.

17

The main towers are fixed on two massive pile foundations, which are imbedded in the basin of the narrows. In the finite element model, the towers are assumed to be ended at the surface of the pier. A general spring, with six degree of freedom, is assigned at each tower leg to simulate the soil-structure interaction. The values of the initial foundation stiffness are obtained from the study done by Geospectra, in August 1993.

The stiffness of soil-structure interaction in the Tacoma Narrows Bridge is found to be high and insignificant in frequency analysis of the Tacoma Narrows Bridge, if compared with fixed support. Analysis was conducted for the fully fixed and partially fixed condition. The main towers are fixed and frequency analysis is conducted. Then it is compared with the analysis done using the linear elastic springs. Neither the frequencies nor and the mode shapes were changed significantly. However, the modal participation factor of the first 20 models changed by around 8%. This might have slight effect on the coupling of modes in wind analysis. For the accuracy of the analysis the linear equivalent springs for the initial foundation stiffness was used in the model.

The effect of the approaches is modelled as linear elastic springs. The equivalent stiffness of the west approach is added as three linear springs distributed on the top beam of tower number 3. The approach is modelled separately and static forces are applied to determine the equivalent stiffness values, which were estimated at 226570 kips/ft and 73.2 kips/ft in for the longitudinal and the transverse directions, respectively. The east tower is very short in length and made of concrete. The east approach is assumed sufficiently rigid and it is modelled as hinge support. The

18

east approach is also excluded from the model. This will not have a significant effect on wind analysis, and will not influence the frequency content.

The two anchors, supporting the main cables and transferring their internal forces to the ground are modelled as hinges, at their corresponding location. Anchors may experience slight movement in case of earthquakes, but their displacements are negligible if wind loading is considered.

Nonlinear elements

According to field investigation, the tower dampers were found to be leaking. These dampers are never replaced and their replacement is impractical. It is difficult to estimate their response based on the provided specifications in the original drawings. The AGA study provided a force-velocity relationship for the tower damper, which is:

0.135V 2 , For V > 31.5 in/sec F = 135 , For 31.5 V 0.02 in/sec 566900 V 2 , For V > 0.02 in/sec

Equation 5

where F is in kips, V is the velocity of the piston in inch/second. The OPAC study utilized this relationship to establish the equivalent linearly elastic viscous damped truss element.

From the above relationship, the tower dampers will not be activated until relatively large change in displacements occurred along the stiffening trusses. Other dampers installed in the

19

structure are more sensitive to slower motion. Explicit modelling of these dampers is ignored in this study. It is assumed that the modal damping, measured in the ambient study, accounts for the contribution of these dampers.

The nonlinear damper is

main span stiffening truss

side span stiffening truss

modelled based on the relation shown in Equation * for V > 31.5 in/sec. Midas-Civil supports such quadratic damping relationship and a small stiffness, 1 kips/ft, is assigned to the visco-elastic damper
Hook element Floor truss
Distance = 0 ft

upper cord bracing Tower leg

Gap element Rigid link End Release

element. Figure 11: Modelling of Windshoe top view at deck level

The nonlinearity of the windshoe is modelled using a combination of gap and hook elements. Figure 11 shows a schematic top view of the meeting point of the stiffening truss and the main tower. The windshoe is connected to the bracing which ends at the mid-span of the lateral beam connecting the tower legs. There is one windshoe for each span, (i.e. two wind shoes at each tower). The hook element has zero stiffness unless it is stretched a distance of 1.25 ft then its stiffness will be engaged and it will act as rigid link due to the assigned high stiffness. The same mechanism works for gap element, but in the other direction.

20

Load and Mass Estimation

In this stage, only sustained loads are estimated. The sustained loads are due to the structural elements, such as, stiffening and floor trusses; and the non-structural elements, such as, the deck and handrails. Traffic load and live load are ignored in the frequency analysis.

Since all structural

15.5

elements are included in the model, and their gross crosssectional properties assigned, the self weight is then used to estimate the gravity. However, the actual structural elements are not all solid and thus the gross area is not suitable for body force
Fundamental Period (sec) 15 14.5 14 13.5 13 12.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 Normalized Density y = -10.978x4 + 44.136x3 - 65.923x2 + 46.84x R2 = 1

Figure 12: First Mode Period vs Normalized Density.

calculations. Therefore, the density of the structural elements might need to be optimized to obtain an averaged density of the structural elements.

For this purpose, a parametric study is conducted to find the effect of changing the material density on the dynamic response. A steel density of, 0.49 kips/ft3 is used as a baseline. Figure 12 shows the change in structural frequency due to the change in density. Although the numerical values shown in this relationship are for the first mode, other values of the other modes are within 10% of these values. It could be seen that change in structural period is not

21

strongly sensitive to the change in density. Thus the steel density is used for the gravity load and for the inertial masse computations in all degrees of freedom.

The superstructure slab, its supporting stringers and handrails are not explicitly modelled, as it is assumed that they have negligible contribution to the overall stiffness. The weight of these elements are averaged and added as concentrated loads and masses at the joints of the upper cord of the floor truss. Then the loads are calibrated, by trial and error procedure, to reach the experimental periods. It was found that the load on the inner and outer joints is 12.91 kips and 10.34 kips, respectively. These figures are close to the estimated values obtained based on quantity surveying of the deck. The gravity loads are used to calculate the equivalent concentred masses of the deck in the three translation directions.

Estimation of Mass

A mass is distributed based on tributary area. MIDAS-civil has the capability to lump the mass based on section properties and material density. Roadway slab mass is lumped at each node based on tributary area. Both Translational mass and rotational mass in the X, Y and Z directions are considered in the analysis.

22

10 9 8 7 6 5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 10 20 30 40 50 Mode Number 60 70 80 90 100

Figure 13: Effect of Eliminating Tower Contribution to Frequency Content

A parametric study is conducted to evaluate the effect of the tower on the frequency content. The tower mass is calculated based on the assigned density and cross sectional area. The density of the tower is set to zero, to eliminate the effect of the tower on the frequency content. Figure 13 shows the percentage difference between the frequency of the bridge without the tower contribution and with the tower contribution to the frequencies. It could be shown that the tower effect is less the 1%, up to the 30th mode. The effect of the tower on the modal analysis is, in fact, critical after the 100th mode, where the modal analysis is cut off. Moreover, it is found from the modal analysis that the frequencies related to the superstructure oscillation are found between the first and the fiftieth modes.

Frequancy Ratio %

23

The parametric study verifies that uncertainty in the tower modelling is not of concern for wind analysis of the bridge superstructure. It is evident that the frequencies governing the tower vibration are beyond these of the superstructure.

FREQUANCY ANALYSIS

Ambient Study

Ambient vibration measurements were taken by the Johns Hopkins University under supervision of Dr. Nicholas Jones in 1993. The study revealed the first 9 vertical, 1 torsional and 12 lateral natural frequencies. Mode shapes and damping ratio are derived from the data generated.

The damping ratio was determined to be in the range of 0.3% to 3.5% and the study concluded that the average damping is 1.3%, which is a reasonable number and close to the true value for the size of the Tacoma Narrows Bridge operating within the elastic range. The results of the natural frequency are shown with the frequency analysis results below.

It could be shown that a damping ratio of 1.3% is relatively small value. Although damping affects the amplitudes of vibration, its influence on the damped natural frequency is very small and could be considered equal to the undamped natural frequency.

24

Eigenvalue Analysis

The developed detailed model of this research has more degrees of freedom than needed for accurate frequency analysis. The model is desired in static analysis to produce accurate estimation for stiffness. However, it might results in undesired modes of vibration particularly when there are very flexible elements, such as cables. Using the traditional procedure to solve the Eigenvalue problem of the equation of motion will yield local mode of vibration in the solution, and thus large number of modal vectors should be solved to reach the desired set of global response.

In order to eliminate this issue, Ritz vectors procedure which is based on Reyleigh-Ritz method, is used. The Ritz method creates a mode shape vector using a normalized displacement vector, which is based on the inertia force distribution associated with the pervious mode shape vectors. The coupling between the modes comes from the stiffness and mass matrix. Adequate number of force vectors is assigned to guarantee accurate spatial distribution of the forces, which is used initially to obtain the first mode shape. In the analysis, 100 Ritz vectors are generated, that give the same number of frequencies. The mode shapes and frequencies obtained from the analysis are found to compare reasonably well with experimental studies as shown in Table 1.

Model Calibration

The following procedure is used to calibrate the model: 1. Pre-modelling Phase: a. Evaluation of the bridge detailed drawings.

25

b. Review previous studies on the bridge. c. Develop a 3-D CAD model for the bridge geometry. 2. Structural Modelling Phase: a. Transfer the CAD to structural analysis the programs, SAP 2000 and MIDASCivil. b. Calculate gross cross sectional properties. c. Compute the weights of non-structural elements. d. Estimate boundary condition of foundations and the approaches. e. Approximate the initial tension in main cable and hangers. f. Solve Eigen value problem 3. Model Calibration a. Calibrate the estimated sustained loads and re-evaluate the cable initial tension. b. Calibrate boundary condition. c. Calibrate selfweight, by tuning the material density. d. Evaluate the modal response and reiterate, as needed7.

RESULTS
The following lists the bridge frequencies. Experimental frequencies are obtained from the ambient vibration taken by the Johns Hopkins University in September 1993 [3]. The analytical results are found using MIDAS-civil. Ritz vector is used in the Eigenvalue analysis. Table 1: Comparison between the Experimental and the Analysis Frequencies
Longitudinal Modes Location Center Span Field Values Hz 0.187 Value - Hz 0.15648 Analytical Error -20% Mode Number 3

26

.262

0.20709

-26%

Transverse Modes Location Field Values Hz 0.074 0.33 Center Span 0.358 0.386 0.496 Value - Hz 0. 071021 0. 340087 0. 348071 0. 396217 0.482368 Analytical Error -4% +3% -3% +3% -3% Mode Number 1 18 19 21 29

Vertical Modes Field Values Location Center Lane 0.162 Hz Curb Lane 0.162 0.20 Side Span 0.262 0.30 0.362 Center Span 0.162 0.30 0.162 0.30 0.25 0.249038 Not Identified 0.324840 0. 40120 0. 156481 0. 324840 +8% +10% -4% +8% Value - Hz 0. 156481 Analytical Error -5% Not Identified -1% 7 15 25 3 14 Mode Number 3

Table 2 shows thirteen modes of vibration for the Tacoma Narrows Bridge. The identified modes are only those vertical and transverse modes of vibration. Appendix A shows complete identification of the bridge mode shapes and their locations. Periods after the 50th mode are found to be relatively very small, less than 1.5 second, for wind analysis. Table 2: Modal Frequency Identification of the TNB
Mode 1 2 3 4 1 2 3 4 Frequency (rad/sec) 0.446236 0.908061 0.983197 1.167195 Frequency (Hz) 0.0710206 0.1445223 0.1564806 0.1857648 Shape L_HS V_HS V_FS L_FS Location MS MS MS MS Symbol Key L V T HS Lateral Mode Vertical Mode Torsional Mode Half Sinusoidal

27

5 6 7 8 9 10 11 12 13

7 15 19 20 21 22 36 37 41

1.564752 2.04103 2.186997 2.372891 2.489505 2.520852 3.604311 3.721043 3.928639

0.2490380 0.3248400 0.3480713 0.3776573 0.3962170 0.4012060 0.5736439 0.5922223 0.6252623

V_(FS+HS) V_2HS L_FS T_HS (L+T)_(FS+HS) V_2FS V_(2FS+HS) (L+T)_FS T_FS

MS MS+SS C+S S S MS MS MS MS

FS MS SS SC MC TW C S

Full Sinusoidal Mid Span Side Span Side Span Cable Mid Span Cable Tower All Cables All Span

Mode 1: Lateral

Mode 20: Torsional

Mode 2: Vertical

Mode 21: Lateral-Torsional

28

Mode 3: Vertical

Mode 22: Vertical

Mode 4: Lateral

Mode 36: Vertical

Mode 7: Vertical

Mode 37: Lateral-Torsional

29

Mode 15: Vertical

Mode 41: Torsional

Mode 19: Lateral

Figure 14: Mode shapes of Identified Modes

CONCLUSION
The Tacoma Narrows Bridge has been successfully and accurately modelled for the purpose of dynamic analysis. The analytical results show good agreement with the experimental data. The transverse and the vertical frequencies are almost equal to the values obtained by the ambient testing. However longitudinal frequencies do not show accurate agreement with the experimental results.

30

There are several factors that affect the frequencies of the structure. The most important factors include mass estimation and distribution, element stiffness and initial tension in the cables. Boundary conditions and soil structure interaction are of significance. Towers have small influence on the low structural frequencies, and their effects are significant only after the 50th mode. Thus, the superstructure and cable properties have the most impact in the lower frequency range.

Since the frequencies associated with the longitudinal mode shapes do not show high accuracy compared with the ambient study, calibration for the longitudinal boundary conditions, which are related to the condensation of the approaches stiffness, will be necessary. However, since longitudinal frequencies will not be excited by wind forces and they do not couple with any of the other mode of vibration, the results obtained here are considered acceptable. Further calibration of the longitudinal frequency needs to be conducted, if earthquake analysis is to be performed.

31

APPENDIX A
Eigenvalue Analysis
Mode No 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 Frequency (rad/sec) 0.4462 0.9081 0.9832 1.1672 1.3012 1.5073 1.5648 1.5988 1.8497 1.9034 1.9048 1.9052 1.9068 2.0315 2.0410 2.0913 2.1368 2.1479 2.1870 2.3729 2.4895 2.5209 2.6453 2.6681 2.6763 2.6847 2.9379 3.0175 3.0308 3.0338 3.1165 3.1500 3.3631 3.3785 3.4433 3.6043 3.7210 3.7792 3.8121 3.8671 3.9286 3.9916 Frequency (cycle/sec) 0.0710 0.1445 0.1565 0.1858 0.2071 0.2399 0.2490 0.2545 0.2944 0.3029 0.3032 0.3032 0.3035 0.3233 0.3248 0.3328 0.3401 0.3418 0.3481 0.3777 0.3962 0.4012 0.4210 0.4246 0.4259 0.4273 0.4676 0.4803 0.4824 0.4828 0.4960 0.5013 0.5353 0.5377 0.5480 0.5736 0.5922 0.6015 0.6067 0.6155 0.6253 0.6353 Period (sec) 14.0804 6.9193 6.3906 5.3831 4.8289 4.1686 4.0155 3.9300 3.3969 3.3010 3.2985 3.2979 3.2952 3.0929 3.0784 3.0045 2.9404 2.9253 2.8730 2.6479 2.5239 2.4925 2.3752 2.3549 2.3477 2.3404 2.1387 2.0823 2.0731 2.0711 2.0161 1.9946 1.8683 1.8598 1.8247 1.7432 1.6886 1.6626 1.6482 1.6248 1.5993 1.5741 Mode No 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 Frequency (rad/sec) 4.5488 4.6111 4.6783 4.7142 4.7865 4.8180 4.9101 5.0463 5.1340 5.2249 5.3256 5.3535 5.3751 5.7356 5.7517 5.8874 6.0325 6.2647 6.3285 6.5879 7.0892 7.1514 7.1681 7.5785 7.9893 8.1640 8.5155 8.9955 9.2453 9.3795 10.4664 10.7922 11.4814 11.8388 12.8915 13.9077 14.5303 15.8278 16.3625 18.7194 20.1433 22.8403 Frequency (cycle/sec) 0.7240 0.7339 0.7446 0.7503 0.7618 0.7668 0.7815 0.8031 0.8171 0.8316 0.8476 0.8520 0.8555 0.9128 0.9154 0.9370 0.9601 0.9971 1.0072 1.0485 1.1283 1.1382 1.1408 1.2062 1.2715 1.2993 1.3553 1.4317 1.4714 1.4928 1.6658 1.7176 1.8273 1.8842 2.0517 2.2135 2.3126 2.5191 2.6042 2.9793 3.2059 3.6352 Period (sec) 1.3813 1.3626 1.3430 1.3328 1.3127 1.3041 1.2796 1.2451 1.2239 1.2026 1.1798 1.1737 1.1690 1.0955 1.0924 1.0672 1.0416 1.0030 0.9928 0.9538 0.8863 0.8786 0.8766 0.8291 0.7865 0.7696 0.7379 0.6985 0.6796 0.6699 0.6003 0.5822 0.5472 0.5307 0.4874 0.4518 0.4324 0.3970 0.3840 0.3357 0.3119 0.2751

32

43 44 45 46 47 48 49 50

4.0022 4.0450 4.0913 4.1292 4.3491 4.4390 4.5064 4.5150

0.6370 0.6438 0.6511 0.6572 0.6922 0.7065 0.7172 0.7186

1.5699 1.5533 1.5358 1.5216 1.4447 1.4155 1.3943 1.3916

93 94 95 96 97 98 99 100

25.1582 28.7127 33.6017 39.9175 52.0059 70.6210 105.1965 379.7931

4.0040 4.5698 5.3479 6.3531 8.2770 11.2397 16.7425 60.4459

0.2497 0.2188 0.1870 0.1574 0.1208 0.0890 0.0597 0.0165

Modal Participation Masses (%)

Mode 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 TRAN-X MASS 0 0 0.01 0 3.04 0 0 0 0 0 0 0 0 0 0.02 0 0 0 0 0 0 1.31 0 0 0.12 0 0 0 0 0 0 0 46.1 0.03 0 TRAN-Y MASS 23.68 0 0 0.01 0 15.12 0 9.64 0 0 0 1.99 0 0 0 0 0 0.37 0.03 3.05 3.07 0 0.04 0 0 0.03 0.08 0 1.86 0 0.01 0 0 0.89 0 TRAN-Z MASS 0 3.23 0 0 0 0 2.42 0 0 0 0 0 0 0 37 0 0 0 0 0 0 0 0 0 4.29 0 0 0 0 0 0 0 0.06 0 0 Mode SUM 0 3.23 3.23 3.23 3.23 3.23 5.64 5.64 5.64 5.64 5.64 5.64 5.64 5.65 42.65 42.65 42.65 42.65 42.65 42.65 42.65 42.65 42.65 42.65 46.94 46.94 46.94 46.94 46.94 46.94 46.94 46.94 47 47 47 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 TRAN-X MASS 0 0 0 0.12 0 0.89 0 0.05 1.96 0 0 0 0 0.16 1.44 0.67 0 0 0.05 0 0.83 0.06 2.41 0.37 0 0.06 0.65 0.1 3.21 0.05 0.56 0 0.33 0.01 0.01 TRAN-Y MASS 0.2 0.08 0.06 0 0.03 0 3.34 0.67 0 0 0 0 0 0.01 0 0 0.13 0 0 0.03 0 0.01 0.04 0 0 0 0.19 0 0.02 0.02 0 0.01 0.04 0 0.01 TRAN-Z MASS 0 0 0 0.03 0 0 0 0 0 0.01 0.02 0.01 0.01 0 0 0.04 0 0 1.38 0 0.68 1.67 0.19 0 0.08 0.15 0 0.52 0.01 0 0.05 0.01 0.02 0.45 0.06

SUM 50.72 50.72 50.72 50.76 50.76 50.76 50.76 50.76 50.76 50.77 50.78 50.79 50.81 50.81 50.81 50.85 50.85 50.85 52.23 52.23 52.91 54.57 54.76 54.76 54.84 55 55 55.52 55.53 55.53 55.58 55.6 55.61 56.06 56.13

33

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

0.9 0 0.26 5.9 0 0 0 0 0 0 0 0 0 0 0

0 0.02 0 0 0.94 0.02 0 0 1.67 0 0.07 0 0.68 0 0

2.42 0 0.28 0 0 0 0 0 0 0 0 0 0 0 1.02

49.41 49.41 49.7 49.7 49.7 49.7 49.7 49.7 49.7 49.7 49.7 49.7 49.7 49.7 50.72

86 87 88 89 90 91 92 93 94 95 96 97 98 99 100

0 0 0 0 0.02 0.11 0 0.07 0.09 0.05 0.24 0 0 0 0

0 0 0.01 0 0.01 0.02 0 0.01 0.01 0.01 0.09 0 0.01 0 0

1.08 0.75 1 6.64 0.5 0.22 5.46 0.51 0.84 1.16 0.61 9 1.12 13.95 0.69

57.21 57.96 58.97 65.6 66.1 66.33 71.78 72.29 73.13 74.29 74.9 83.9 85.03 98.98 99.67

Mode Shape Identification

Mode # 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Frequency (rad/sec) 0.446236 0.908061 0.983197 1.167195 1.301158 1.50725 1.564752 1.598782 1.849657 1.9034 1.904833 1.905209 1.906756 2.031485 2.04103 2.091287 2.136832 2.147853 2.186997 2.372891 2.489505 2.520852 2.645296 2.668146 2.676282 2.684675 Shape L_HS V_HS V_FS L_FS V_FS L_HS V_(FS+HS) L_HS L_HS L_HS L_HS L_FS L_FS L_FS V_2HS L_HS L_2HS L_HS L_FS T_HS L+T_HS V_2FS L_FS L_2HS V_HS V_HS Location MS MS MS MS SS SS MS SS SC SC MC MC+TW MC+MS SC+MC MS+SS SC C+S SC C+S S S MS SC SC SC SC Mode # 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 52 Frequency (rad/sec) 2.937856 3.017497 3.030805 3.033756 3.116474 3.15004 3.36308 3.378495 3.443337 3.604311 3.721043 3.779155 3.812122 3.867088 3.928639 3.991616 4.002207 4.045034 4.091282 4.129206 4.349058 4.438963 4.506398 4.515044 4.61109 Shape L_2FS L_4HS L_2FS L L L V_3FS L L V_(2FS+HS) L+T_FS V_(2FS+HS) V_FS L T_FS L L L L L_2FS T_HS T_HS V V T_HS Location C MC C C C SC S SC SC MS MS SS+MS SS C MS SC MC C C C+MS SS SS SS SS SS

34

Symbol Key L Lateral Mode V Vertical Mode T Torsional Mode HS Half Sinusoidal FS Full Sinusoidal MS Mid Span SS Side Span SC Side Span Cable MC Mid Span Cable TW Tower C All Cables S All Span

35

REFERENCES

1- AGA and OPAC Geospectra, Tacoma Narrows Bridge Seismic Vulnerability Study, 1994. 2- OConnor C., Design of Bridge Superstructures, Wiley-Interscience, New York, 1971. 3- OPAC Consulting Engineers, Tacoma Narrows Bridge Seismic Vulnerability Study:

Calculations, 1993.
4- Peterson S., Experimental Response and Analysis of the Evergreen Point Floating

Bridge, A PhD dissertation, Washington State University, 2002

5- Anil Chopra, Dynamics of Structures: Theory and Applications to Earthquake Engineering, Second Edition, Pearson Education Inc., India, 2001 6- Alfred Pugsley, The Theroy of Suspension Bridges, Second Edition, Edward Arnold Publishers LTD, London, UK,1968 7- P. S. Barar, R. Raul and R. H. Scanlan, Numerical Calculation of Flutter Derivatives Via

Indicial Functions, Journal of Fluid and Structures, 1996, vol 10, pp. 337-351.
8- F. B. Farquharson, Aerodynamic Stability of Suspension Bridges: with special reference to the Tacoma Narrows Bridge, The Structural Research Laboratory University of Washington, Seattle, 1954.

36