2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020
Paper N° C000854
Registration Code: S-A01094

IMPACT OF VISCOUS DAMPING MODEL ASSUMPTIONS ON THE

NONLINEAR ANALYSIS OF MULTI-SPAN BRIDGES
D.R. Martinez(1), M.J. Kowalsky(2)
(1) Ph.D. Candidate, North Carolina State University, [email protected]
(2) Professor-Structural Engineering, North Carolina State University, [email protected]

Abstract
In the performance-based design of bridges (PBD), the accurate determination of deformation demands is essential to
ensure that structures achieve their intended performance. An integral part of advanced PBD approaches are design
verification techniques such as Nonlinear Time-History Analysis (NLTHA). One assumption that has proven essential
but is not well understood by engineers is the impact that viscous damping might have on the nonlinear response of
structures. Other researchers have shown that for Single-Degree-of-Freedom (SDOF) systems, the displacement time
histories from shake table tests compare much better with a Stiffness-Proportional Damping Model using a Tangent
Stiffness rather than an Initial Stiffness approach. Conversely, the problem is that the Initial Stiffness option is normally
the damping method that is included in most computational platforms, and many engineers are not aware of the
consequences that might have on the predictions of displacement response. In the case of Multi-Degree-of-Freedom
(MDOF) bridges, little research has been done on this problem, and experimental verification is difficult due to the cost
and complexity of tests at such a scale. Hence, in this study an extensive sensitivity analysis using NLTHA is performed
for the transverse direction of three configurations of multi-span reinforced-concrete bridges, considering parameters such
as the directivity of the ground motion, abutment restraint, and a concentrated-plasticity model in Ruaumoko3D. Seven
damping models, such as the Modal Damping, Rayleigh (Mass and Stiffness terms), Stiffness-Proportional (Initial and
Tangent) and Zero Viscous Damping, were considered. The results show that Modal Damping leads to the minimum
seismic demands in all the models, while the Tangent Stiffness-Proportional model and Zero Viscous Damping lead to
the largest displacements. The range of difference in the transverse displacements between the Modal Damping (lower
limit) and the Zero Viscous Damping (upper limit) is from 32% to 72% for bridges with free abutments. In contrast, a
difference from 42% to 69% was seen for structures with pinned abutments. Finally, the impact of the selection of a
viscous damping model on the response of an existing bridge is presented. The Anchorage Port Access Bridge (Alaska)
is a 700-meter long steel bridge that was analyzed using several damping models. Its hysteretic response was calibrated
from two quasi-static tests on half-scale models conducted at the Constructed Facilities Lab at NC State. A difference in
some piers of more than 100% among the damping models was observed when the bridge was subjected to the 2018
Anchorage earthquake that was spectrally matched to the AASHTO design spectrum. Thus, the results of both analyses
highlight the profound influence that the choice of the viscous damping model may have on the inelastic response of
bridges.
Keywords: Performance-Based Design, NLTH Analysis, Viscous Damping, Damping Models, Port Access Bridge.

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

1. Introduction
The philosophy of Performance-Based Design (PBD) has gained popularity with engineers as it allows
structures to be designed to achieve prescribed performance levels for defined seismic hazards. In this context,
Non-linear Time History Analysis (NLTHA) is the most sophisticated method currently available for design
verification. In inelastic analyses, Rayleigh damping is the viscous damping model that is preferred by
practitioners as it uses a combination of the Mass and Stiffness matrices to construct the damping matrix.
However, researchers have found problems with Rayleigh damping models for nonlinear analysis of Multi-
degree of Freedom Structures (MDOF).

Many researchers have studied this phenomenon. Chrisp [1] and then Bernal [2] were the first to identify the
high damping forces that can be produced in an inelastic analysis that uses initial proportional Rayleigh
damping. Similarly, Leger et al. [3] studied the effects of various viscous damping models (Rayleigh, mass-
proportional and stiffness-proportional damping) on MDOF structures. They recommended a range of periods
for the structure where it is suitable to use either the initial or the tangent stiffness. Priestley & Grant [4]
suggested using the tangent stiffness proportional damping (with special care in the selection of the damping
associated with the first mode as the structure will act as if initial stiffness proportional damping has been
prescribed, even if tangent stiffness is specified in the analysis). Consequently, Kowalsky & Hasgul [5] studied
the impact of initial and tangent stiffness proportional damping on SDOF systems using various hysteretic
rules obtaining a significant difference between both models. On the other hand, Hall [6] stated that the mass
proportional term of the Rayleigh approach cannot possibly exist for analysis with dynamic loads and proposed
to limit the high damping forces by capping the stiffness-proportional factor, eliminating the mass-proportional
damping contribution. In response, Charney [7] recommended to eliminate the use of viscous damping and to
rely only upon hysteretic damping for nonlinear analysis. More recently, Hall [8] performed an analysis
studying the influence of different damping models on the inelastic response of moment-frame buildings. He
stated that initial proportional damping should not be used for inelastic seismic analysis.

Other investigators such as Chopra & McKenna [9] and Carr [10] [11] have recommended using the
superposition of modal damping matrices (Wilson-Penzien) for nonlinear analysis of buildings since this
approach completely eliminates the spurious damping forces when the structure yields. Nonetheless, Petrini et
al. [12] found that the analytical response of a concentrated plasticity model for a SDOF reinforced concrete
column with tangent stiffness proportional damping was the closest to the results of a shaking table test. This
was obtained through the comparison of concentrated and plasticity models (using initial and tangent stiffness
proportional damping models) with the experiments. Despite this, Smyrou et al. [13] point out that for MDOF
systems the stiffness proportional damping model overestimates the damping for higher modes. They
suggested to model elastic damping in NLTHA for cantilever Reinforced Concrete (RC) walls using Modal
Damping with an artificial lowering of the damping ratio in the first mode as proposed by Priestley and Grant
[4]. In this case, the damping in the first mode will have similar behavior as the tangent stiffness proportional
damping obtained by Petrini et al.

Based on the existing literature, there are many options to model viscous damping in NLTHA. Nevertheless,
most of the recommendations are given for buildings, where non-structural components contribute to damping.
In the cases of bridges, where the structural system approaches a “bare-frame”, research is limited. In the
present paper, the aim is not to recommend the most appropriate damping model for nonlinear analysis of
structures, since that conclusion only can be reached through experimental verification. Instead, we present
the impact on the deformation demands for bridges due to the selection of various damping models for
nonlinear analysis. This will be carried out through the study of viscous damping models based on both
Rayleigh and Modal Damping. First, the results of NLTHA for these damping models will be evaluated for
three four-span RC bridges. Then, the impact of viscous damping will be shown through the seismic
assessment of the Anchorage Port Access Bridge, which is a 787-m long, irregular and highly complex steel
bridge, situated in Anchorage, Alaska.

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

2. Damping Models
2.1 Rayleigh Damping

ξ =ξ = ξ

(a) Physical Meaning (b) Mass and Stiffness Terms

Fig. 1 – Rayleigh Damping Model (adapted from [14]).
This damping model is commonly known as Proportional damping and is expressed as follows:
c = m+k (1)
Where the coefficients α and β are calculated by selecting two modes of vibration, m corresponds to the mass
matrix of the structure, and k is the stiffness matrix. The meaning of Rayleigh damping is illustrated in Figure
1(a), where a 2-story frame with beam and column stiffness Kb and Kc, respectively, is shown. The coefficient
βk simulates internal viscous dampers and acts in parallel with the elements. On the other hand, the term αm
corresponds to viscous dampers that connect the masses M1 and M2 to external fixed points [14]. This model
simulates the damping associated with the mass and stiffness of the structure and from Figure 1(b) is seen that
for the lower frequencies, the damping ratio will be higher and will depend strongly on the mass coefficient
term (dashed green line). In contrast, for the higher modes, the stiffness proportional term will dominate the
calculation of the damping ratio. Special care has to be taken for higher modes, where the damping ratio grows
linearly and there is low apportion of the mass term as in these cases a very high value of damping might be
assigned to these upper modes.

This stiffness matrix (k) can be computed considering the initial or the current stiffness of the structure. This
is known as the initial and tangent stiffness, respectively. The latter approach can be calculated using a tangent
or a secant damping matrix. In the case of the tangent stiffness with the tangent damping matrix, the stiffness
has hysteresis and the damping actions are not zero when the velocities are zero [11]. In contrast, the option of
the tangent stiffness with the secant damping variation does not have those issues. Therefore, this is the
recommended approach to model the tangent stiffness for NLTHA and the one considered in this paper.

2.2 Modal Damping

Wilson and Penzien (1972) showed that the damping matrix considering only the N mode shapes of free
vibration will produce a Caughey damping model more easily. This procedure is based on the
orthogonalization of the lateral mode shapes, resulting in a fully populated damping matrix obtained by
summing individual matrices for each mode. This approach was not initially popular as it increased the
processing time of the response history analysis due to the fully populated matrix. However, given the power
of modern computers, it has recently gained popularity and is the method recommended by some academics
to conduct nonlinear analysis [9] [11]. Equation 2 shows the representation of this approach.
𝑛
2∙ 𝑖 ∙ 𝑖
𝑐= 𝑚 ∙ 𝜙𝑖 𝜙𝑖𝑇 ∙ 𝑚
𝑀𝑖
𝑖=1 (2)

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

In Equation 2, N corresponds to the number of damped modes, Tn is the mode period, ξn is the fraction of
critical damping, m is the mass matrix, Фn is the mode shape, and ωi is the circular natural frequency.

3. Numerical Study for Three Four-Span Bridges

The impact of viscous damping model assumptions will be assessed through evaluating the nonlinear response
of three reinforced concrete bridges using various damping models. The structures selected represent typical
geometric configurations from Alaska and were based on the work done by Kong [15]. The geometry of these
bridges can be seen in Figure 5. The structures have four circular columns of different heights (8m, 12m, 14m,
and 17m) but same diameter (2m). The materials of the structure have a concrete compression strength (f’c)
of 30 MPa and yield stress for the longitudinal and transverse steel of 420 MPa. The dead load from the
superstructure is 120 kN/m, which produces an axial load ratio of 7% for the columns. A 1% of steel ratio was
considered for the longitudinal and transverse steel reinforcing ratio of the columns. Also, the abutments were
assumed to behave free or restricted to translate in the transverse direction, while for the longitudinal direction
a bilinear elastic hysteresis loop with a GAP of 0.075m was considered. It is assumed that the superstructure
is supported on elastomeric bearings, therefore the columns will behave in single bending in the transverse
direction. The yield displacements using the approach proposed by Priestley et al. [16] for the columns of 8m,
12m, 14m, and 17m are 0.06m, 0.13m, 0.18m, and 0.26m, respectively. The Mander model [17] was used to
calculate the confined properties of the sections and later was employed to obtain the moment-curvature
relationship with an equivalent yield moment of 14160 kN*m. Using this, the cracked section moment of
inertia was calculated to be 0.21m4 for the elements of the substructure. The nonlinear time-history analyses
were performed using Ruaumoko3D [11], where a lumped plasticity model based on Giberson’s one-
component element was used for the modeling of the bridges. It was assumed that the inelasticity was
concentrated only at the ends of the column, and the superstructure was modeled as elastic. Lastly, a bi-
directional analysis using the two horizontal and perpendicular components of the ground motions was
considered.

3.1 Demands
The analyses of the structures were carried out using a suite of unscaled ground motions from the Pacific
Earthquake Engineering Research Center (PEER). Seven near-field pairs of ground motions were selected
considering moment magnitudes (Mw) between 6.5 and 7.5 and epicentral distances less than 20 km. The main
characteristics of these records are summarized in Table 1.

Table 1 – Characteristics of Ground Motions

Eq Earthquake Name Station Name Magnitude Mechanism Rrup (km) Vs30 (m/sec)
1 Loma Prieta Gilroy Array #3 6.93 Reverse Oblique 12.82 349.85
2 Northridge-01 Canyon Country - W Lost Cany 6.69 Reverse 12.44 325.60
3 Duzce_ Turkey Bolu 7.14 Strike slip 12.04 293.57
4 El Mayor-Cucapah_ Mexico Michoacan de Campo 7.20 Strike slip 15.91 242.05
5 El Mayor-Cucapah_ Mexico Riito 7.20 Strike slip 13.71 242.05
6 El Mayor-Cucapah_ Mexico El Centro Array #11 7.20 Strike slip 16.21 196.25
7 Darfield_ New Zealand DFHS 7.00 Strike slip 11.86 344.02

The demands on the bridges were calculated using an AASHTO design response spectrum, which has a
probability of exceedance of 7% in 75 years (approximately a return period of 1000 years). For this scenario,
the bridges are situated in Anchorage, Alaska, and have a PGA of 0.538g, Ss of 1.183g, S1 of 0.458g, and a
soil site class D. Note that the records selected from the PEER have a shear velocity at a depth of 30 meters
between 180 and 360 m/s, which make them compatible with the soil class D.

Figure 2(a) shows the acceleration response spectrum for the individual components of the seven pairs of
earthquakes selected for the analysis, and they are contrasted with the AASHTO design spectrum for the

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

bridges. The AASHTO design spectrum is obtained using the geometric means, which is essentially the same
as the RotD50 definition, which corresponds to the median of the response spectrums of SDOF oscillators (See
Boore et al. [18]). The RotD50 for the seven pairs of ground motions can be seen in Figure 2(b). These records
were selected because their RotD50 representation was the best fit for the design spectra used in the analysis.
2.50 2.50
DESIGN SPECTRA

Spectral Acceleration, pSa (g)

Spectral Acceleration, pSa (g)

MEAN RECORDS
2.00 2.00

1.50 1.50

1.00 1.00

0.50 0.50

0.00 0.00
0.01 0.10 1.00 10.00 0.01 0.10 1.00 10.00
Period, T (Sec) Period, T (Sec)

(a) As-recorded unscaled individual records (b) RotD50 As-recorded pair of GM

Fig. 2 – Acceleration Response Spectrum of Ground Motions Selected for Analysis.
The calculation of the demands for the bridges was performed using the as-recorded unscaled records described
previously and the ground motions were then spectrally matched to the design spectrum for a range of periods
between 0.10 sec and 5.00 sec, which corresponds to the interval of elastic and inelastic periods of the
structures. This was done using the Matlab Code ArtifQuakeLetII [19] and is shown in Figure 3(a). Conversely,
the RotD50 for the seven pairs of ground motions were spectrally matched to the design spectra and are shown
in Figure 3(b). Note how the RotD50 response spectrum for each one of the records matches relatively well
with the AASHTO spectrum.

1.80 1.80
Upper Limit
Spectral Acceleration, pSa (g)

Lower Limit Upper Limit Lower Limit

Spectral Acceleration, pSa (g)

1.60 1.60
1.40 1.40
1.20 1.20
1.00 1.00
0.80 0.80
0.60 0.60
0.40 0.40
0.20 0.20
0.00 0.00
0.01 0.10 1.00 10.00 0.01 0.10 1.00 10.00
Period, T (Sec) Period, T (Sec)
(a) Spectrally Matched Individual Records (b) Spectrally Matched RotD50
Fig. 3 – Acceleration Response Spectrum of Spectrally Matched Records.

3.2 Damping Models

Modal and Rayleigh Damping were selected to conduct the nonlinear analysis. In the case of Rayleigh
damping, numerous variations were used to see its influence on the results. The seven types of damping models
used in this paper are summarized in Table 2.

Table 2 shows that DM1 corresponds to Modal Damping, which has a viscous damping (ξ) value of 5% for all
the modes of the structures. DM2 is calculated using a Rayleigh damping approach for modes 1 and 2 using
Initial Stiffness. Generally, this model represents the most common damping model in practice. DM3 and DM4
are calculated using modes 1 and 10 for the initial and tangent stiffness with secant damping variation,
respectively. The selection of mode 10 represents the upper limit at which 100 % of the mass will be achieved
for all the structures (see Figure 4a). DM5 and DM6 correspond to the stiffness proportional damping models

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

using the initial and tangent stiffness, respectively. In both cases, the coefficient β was calculated using a 5%
of viscous damping for modes 1 and 2 and eliminating the mass term. This produces that the first modes are
underdamped, while the upper modes are not as overdamped as it would be the case assuming 5% for the first
mode of the structures. Finally, DM7 represents the upper limit of the nonlinear response where zero viscous
damping is considered. The distribution of viscous damping for the seven damping models used in this study
are shown in Figure 4(b) for Bridge 1.

Table 2 – Characteristics of Damping Models used in the analysis.

DAMPING ξ1 ξ2
NOMENCLATURE TYPE OF DAMPING Mode 1 Mode 2
MODEL % %
DM1 MOD Modal (Wilson-Penzien) 1 5 2 5
DM2 M+K0-1&2 Rayleigh M&K- Initial Stiffness 1 5 2 5
DM3 M+K0-1&10 Rayleigh M&K- Initial Stiffness 1 5 10 5
DM4 M+KS-1&10 Rayleigh M&K- Tangent Stiffness (Secant Variation) 1 5 10 5
DM5 K0 Rayleigh- Initial Stiffness Proportional Damping 1 5 2 5
DM6 KS Rayleigh- Tangent Stiffness Proportional Damping (Secant Variation) 1 5 2 5
DM7 ZERO Non Viscous Damping 1 0 2 0

100 30.0
Cum. Mass Participation (%)

90 27.5
25.0
Viscous Damping (%)

80
22.5
70 20.0
60 17.5
50 15.0
12.5
40
10.0
30 7.5
20 5.0
10 2.5
0 0.0
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0
1 2 3 4 5 6 7 8 9 10
Mode of Vibration Circular Frequency, f, (Hz)
DM1 (MODAL) DM2 (M+K0-1&2)
BRIDGE 1, FREE BRIDGE 2, FREE BRIDGE 3, FREE DM3 (M+K0-1&10)/DM4 (M+KS-1&10) DM5 (K0)/DM6 (KS)
BRIDGE 1, PINNED BRIDGE 2, PINNED BRIDGE 3, PINNED DM7 (ZERO)

(a) (b)
Fig. 4 – Distribution of Viscous Damping for Nonlinear Analysis; (a) Mass Participation in Transverse
Direction for the three bridges with free and pinned abutments; (b) Viscous Damping per Damping Model
for Bridge 1 with pinned abutments.

3.3 Nonlinear Analysis for Bridges with Pinned Abutments

The means of the maximum transverse displacements calculated using the seven pairs of ground motions
spectrally matched to the design spectrum and for the seven damping models are shown in Figure 5. Figures
5(a), 5(b) and 5(c) correspond to the transverse nonlinear response for Bridges 1, 2 and 3, respectively.

Figure 5 shows that for bridges with pinned abutments there are three main behaviors. The first tendency is
that the damping models Modal, M+K0-1&2, M+K0-1&10 and M+KS-1&10 have similar inelastic
displacement patterns. All these responses are relatively close among them and this could be explained since
these models are based on a classical Rayleigh damping approach. In these cases, the damping matrix is
calculated using the mass and the stiffness proportional terms of Equation 1. For these scenarios, the mass
proportional term (m) is dominating the inelastic response. Thus, regardless of what type of stiffness is used
in the analysis (initial or tangent), the response will be similar for a Rayleigh damping model. These findings
are in agreement with the results obtained by Priestley and Grant [4].

The second tendency corresponds to the behavior of the bridges with the stiffness proportional damping model
(K0-1&2 and KS-1&2). In this situation, the KS-1&2 model has larger deformation demands than the K0-1&2
approach with differences varying between 3% and 6%. This is expected since the KS model is calculated

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

using the current stiffness of the structure instead of the initial elastic properties. Also, note that there is a
displacement variation between the stiffness proportional model with tangent stiffness and the modal damping
(lowest response) of 39%, 34%, and 44%, for the central columns of bridge 1, 2 and 3, respectively. The
explanation of this behavior is not only that the stiffness proportional damping model is not using the mass
term of Equation 1, but also the1 value of viscous damping is lower than 5% for the
Bridge first
Bridge 2 modes of the bridges.
0.50 0.35

Trans. Displacement (m)

Trans. Displacement (m)

0.30
0.40
0.25
0.30 0.20

0.20 0.15
8m
0.10 18m
12m
0.10 14m 14m 14m
0.05
35m 40m 40m 35m 35m 40m 40m 35m
0.00 0.00
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Length (m) Length (m)
MODAL M+K0-1&2 M+K0-1&10 M+KS1-10 MODAL M+K0-1&2 M+K0-1&10 M+KS1-10
K0 KS ZERO K0 KS ZERO
(a) Bridge 1 Bridge 3 (b) Bridge 2
0.50
Trans. Displacement (m)

0.40

0.30

0.20
8m
0.10 18m 18m

35m 40m 40m 35m

0.00
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Length (m)
MODAL M+K0-1&2 M+K0-1&10 M+KS1-10
K0 KS ZERO
(c) Bridge 3
Fig. 5 – Transverse Displacement Profile with Pinned Abutments.
Lastly, the upper bound occurs when no viscous damping is deployed. As stated previously, some researchers
[4] [7] believe that the non-viscous damping model is closer to the reality for nonlinear analysis of bridge
structures because the amount of viscous damping that participates in the inelastic response is small. There are
differences in the deformation demands between the lowest and highest responses for Bridges 1, 2 and 3, of
60%, 46%, and 69%, respectively. All these results demonstrate the high impact that the definition of the
viscous damping model has on the deformation demands of bridges.

3.4 Nonlinear Analysis for bridges with Free Abutments

The same bridges were analyzed considering abutments that are free to translate. This is a more realistic
scenario for bridges that are supported on elastomeric bearings and it is the normal assumption made by
practitioners. Caltrans recommends in their design guide specifications [20] to model the abutments in the
transverse direction with an elastic-plastic behavior having a yield displacement of 0.05m and a force of 30%
of the vertical force on the abutments. For this case study, the conservative hypothesis of fully free translation
was made. Bridge 1 Bridge 2
0.50 0.70
Trans. Displacement (m)

Trans. Displacement (m)

8m
0.60 18m
12m
0.40
0.50
35m 40m 40m 35m
0.30 0.40
0.20 0.30
14m 14m 14m 0.20
0.10
35m 40m 40m 35m 0.10
0.00 0.00
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Length (m) Length (m)
MODAL M+K0-1&2 M+K0-1&10 M+KS1-10 MODAL M+K0-1&2 M+K0-1&10 M+KS1-10
K0 KS ZERO K0 KS ZERO

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

(a) Bridge 1 Bridge 3 (b) Bridge 2

0.50

Trans. Displacement (m)

0.40

0.30

0.20
8m
18m 18m
0.10
35m 40m 40m 35m
0.00
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150
Length (m)
MODAL M+K0-1&2 M+K0-1&10 M+KS1-10
K0 KS ZERO
(c) Bridge 3
Fig. 6 – Transverse Displacement Profile with Free Abutments.
Figure 6 displays the transverse displacement profile for the three bridges with free abutments. For the three
structures, the dispersion in the results is higher compared with the pinned abutment case. This is predictable
since the inelastic behavior is concentrated in the columns, while with pinned abutments, the elastic damping
of the superstructure participates in the seismic response. The Modal damping approach is always giving the
lowest response. Also, notice how the Rayleigh damping models with mass and stiffness proportional terms
are close among them, again showing that the mass term is dominating the inelastic response. A larger
dispersion is seen for the stiffness proportional damping models with initial stiffness and tangent stiffness. In
this case, a fairly constant difference of 10% for Bridge 1 is observed. For Bridge 2, there is a variation in the
results from 8% to 46%, while Bridge 3 has differences between 6% and 13%. Comparing the extreme
responses, a maximum difference in the results of 32%, 72%, and 76% is noted. These significant differences
in the nonlinear results highlight the importance of a careful selection of a viscous damping model for
structures that are modeled to translate freely in the transverse direction.

4. Anchorage Port Access Bridge

(a) Photo of Anchorage Port Access Bridge

Mass 1 Mass 2 Mass 3

Steel Section 2,
Inelastic Cap Beam, Elastic

Steel Section 2,
Elastic

Steel Section 1,
Inelastic

Pedestal, D≈0.76m
Elastic

Fixed Supports

(b) 3D Modeling of APA Bridge (c) Lumped plasticity for two-column bent

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

Fig. 7 – Global Model of APA Bridge.

The last part of this paper shows the impact of viscous damping model assumptions on a real, irregular, large,
and complex bridge: the Anchorage Port Access (APA) Bridge, a 787m long steel bridge. This structure was
built in 1972 and currently is being assessed by the Alaska Department of Transportation (AKDOT) for seismic
deficiencies. The main problem of the bridge is that its columns have slender and non-compact sections,
allowing them to fail prematurely due to buckling and low displacement ductilities. Therefore, two half-scale
tests of the piers 11 and 16 were constructed and tested at the Constructed Facilities Lab at NC State. The Piers
11 and 16 were selected because they have a D/t ratio of 80, and 44, respectively.

A 3D lumped plasticity spline model of the Anchorage Port Access Bridge was developed in Ruaumoko 3D
to conduct nonlinear time history analyses. Figure 7(b) shows the model of the bridge. Some of the
assumptions used in this model include: (1) the abutments are free to translate in both directions, (2) the
columns are fixed at their base, and (3) the superstructure is continuous. In addition, Figure 7(c) exhibits the
inputs used to conduct the nonlinear analysis. The plastic hinges are located at the ends of the columns, while
the cap beams and the superstructure are modeled as elastic elements. This figure also displays the mass
discretization. In this case, the recommendations given by Priestley et al. [16] were followed considering one-
third of the column self-weight for the analysis.

4.1 Experimental Tests

z’
y’
0.76m 3.20m x’ 1.37m

2 1 3 4
3 4 5
Top Steel Section Lp= 0.76m
(0.76mx0.010m), Cap Beam, Elastic
Inelastic (0.58mx0.91m)
x’
2.03m
2
y’
z’
OPENING
5
Bottom Steel Section CLOSING
1
(0.81mx0.016m), 1.09m
Elastic
6

Y
X Fixed Support
Z

(a) Schematic of Test 1 in Ruaumoko3D (b) Experiment vs Analytical Model Force-

Displacement Response
Fig. 8 – Model schematic and Force- Displacement response for Test 1.

The nonlinear analysis of the APA bridge was conducted using the hysteresis that best fit the experimental
results. Figure 8(a) indicates the representation of Test 1 using a lumped plasticity approach in Ruaumoko 3D.
In this case, the cap beam was modeled as an elastic element. In addition, Figure 8(b) exhibits a comparison
of the vertical component of the force-displacement response given by the chosen hysteretic rule for the
analysis (Ramberg Osgood with strength and cycle degradation) and the experiment. The hysteresis model
effectively captures the linear and nonlinear behavior observed in the experiment up to a displacement ductility
of 2.0. After that point, it is assumed that the column will buckle and will have rapid strength degradation.

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

4.2 Demands
1.80 1.80
Lower Limit Upper Limit DESIGN

Spectral Acceleration, pSa (g)

Spectral Acceleration, pSa (g)

1.60 1.60 RotD50

1.40 As Recorded/ 1.40 RotD100
Unscaled Records
1.20 1.20
1.00 1.00
Spectrally
0.80 Matched Records 0.80
0.60 0.60
0.40 0.40
0.20 0.20
0.00 0.00
0.01 0.10 1.00 10.00 0.01 0.10 1.00 10.00
Period, T (Sec) Period, T (Sec)
(a) As recorded ground motion spectrally (b) RotD50 and RotD100 of spectrally matched
matched to design spectrum records
Fig. 9 – Ground motion matching of as recorded motion of November 2018, Anchorage, AK earthquake.

The displacement demands on the bridge were calculated using the pair of ground motions recorded at the Port
Access Station (NP8043) of the Anchorage earthquake that occurred on the 30th of November, 2018. The two
components of the record were spectrally matched to a design spectrum, which was based on a previous Site-
Specific Seismic Hazard Analysis study the AKDOT conducted [21]. The program ArtifQuakeLetII was used
to perform the spectral matching of the records. Figure 9(a) shows the spectral match of the pair of ground
motions for a range of periods from 0.1 to 10 seconds. On the other hand, Figure 9b) displays the calculation
of the RotD50 and RotD100 of the spectrally matched records. In this circumstance, again, as in the case study
of the three reinforced concrete bridges, the RotD50 of the ground motion matches the design spectrum well;
thus, the two components of the matched earthquake do not need to be altered.

(a) Plan view of Port Access bridge

6
DM1. MODAL

5 DM2. M+KS-1&2
DM3. M+KS-1&20
Disp. Ductility

4 DM4. KS-1&2
DM5. KS-1&20
3

Pier Locations
(b) Displacement ductility per each pier for the 5 damping models considered
Fig. 10 – Results in transverse direction of NLTHA for Anchorage record spectrally matched to design
spectrum for a Fy= 290 MPa.
Figure 10 displays the nonlinear time history analysis for the bridge with the spectrally matched ground
motions for five damping models and yield stress (Fy) of 290 MPa. These results were obtained by calculating

10

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

the envelope of maximum displacements for the pair of ground motions rotated each 15°. Five damping models
were used to show the possible range of deformations at each pier. The following damping models were
analyzed: Modal, Rayleigh (M+KS) and Stiffness Proportional Damping (KS) for two modes of vibration:
1&2 and 1&20 and 5% of viscous damping ratio because 90% of the mass participation in the transverse
direction occurs before mode 20.

As shown in Figure 10(b), the maximum displacement demands on the bridge are concentrated between piers
five and twelve. Moreover, it can be seen that there is a bigger dispersion in the results of the damping models
in comparison with the three bridges analyzed in the previous section. This proves that the impact of viscous
damping models on irregular structures is higher than for regular structures. In most of the piers that behave
inelastically, the lowest response was associated with the Modal Damping. In addition, the data in the figure
exhibits the importance of the selection of the second mode in the Rayleigh Damping for irregular structures,
where the higher modes are essential. In this situation, there is a considerable variation in the results for
Rayleigh damping with modes 1&2 and 1&20. It is seen that the higher modes of the DM2 model are
overdamped. Furthermore, even for some piers such as 8 and 9, DM3 is very close to DM4 (stiffness
proportional damping), which is the approach that gives the second largest response. This is only behind DM5,
which is the damping model that used zero viscous damping. Finally, it is observed that in piers such as 6 and
7, there are differences higher than 100% between the extreme responses.

5. Conclusions
This study shows the profound impact that the selection of viscous damping has on the inelastic response of
bridges. This was done by studying the nonlinear behavior of bridges in the transverse direction, considering
three cases: (1) three RC bridges with pinned abutments, (2) three RC bridges with free abutments, and (3) the
Anchorage Port Access Bridge. In all the cases, various damping models based on Rayleigh and Modal
Damping were analyzed. For the RC bridges with pinned and free abutments, the mass term dominates the
responses of Rayleigh Damping Models. Modal Damping is always lower than Rayleigh damping, but very
similar in its responses. This can be explained because the higher modes had limited participation in the
response of these bridges. Damping models based on stiffness proportional and zero viscous damping give the
largest demands. In the case of the Port Access Bridge, the higher modes are important, producing considerable
differences in the deformation demands of all the damping models considered in this paper. Finally, the
numerical study conducted here highlights the importance of the selection of a proper damping model to obtain
realistic deformation demands of bridges using NLTHA. Since there is not a consensus on what is the proper
damping model that should be used in nonlinear analysis with lumped plasticity models, this topic is expected
to be studied further.

6. Acknowledgements
The first author is pleased to acknowledge COLCIENCIAS and the Fulbright Association in Colombia for a
scholarship to pursue his Ph.D. studies in the United States. Financial support from NC State University and
the Alaska Department of Transportation and Public Facilities is also gratefully appreciated.

7. References
[1] D. Chrisp, Damping models for inelastic structures. Master’s Thesis, University of Canterbury, Christchurch,
New Zealand. 1980.
[2] Bernal Dionisio, “Viscous Damping in Inelastic Structural Response,” J. Struct. Eng., vol. 120, no. 4, pp. 1240–
1254, 1994.

11

© The 17th World Conference on Earthquake Engineering - 2d-0029 -

2d-0029 The 17th World Conference on Earthquake Engineering

17th World Conference on Earthquake Engineering, 17WCEE

Sendai, Japan - September 13th to 18th 2020

[3] P. Léger and S. Dussault, “Seismic-Energy dissipation in MDOF Structures,” J. Struct. Eng., vol. 118, no. 5, pp.
1251–1269, 1992.
[4] M. J. N. Priestley & D. N. Grant, “Viscous Damping in Seismic Design and Analysis,” J. Earthq. Eng., vol. 2469,
no. September, pp. 229–255, 2005.
[5] M. Kowalsky and U. Hasgul, “Impact of Viscous Damping Models on Nonlinear Response of SDOF Systems,”
10th U.S. Natl. Conf. Earthq. Eng., 2014.
[6] J. F. Hall, “Problems encountered from the use (or misuse) of Rayleigh damping,” Earthq. Eng. Struct. Dyn., vol.
35, no. 5, pp. 525–545, 2006.
[7] F. a. Charney, “Unintended Consequences of Modeling Damping in Structures,” J. Struct. Eng., vol. 134, no. 4,
pp. 581–592, 2008.
[8] J. F. Hall, “Performance of viscous damping in inelastic seismic analysis of moment ‐ frame buildings,” Earthq.
Eng. Struct. Dyn., no. June, pp. 2756–2776, 2018.
[9] A. K. Chopra and F. McKenna, “Modeling viscous damping in nonlinear response history analysis of buildings
for earthquake excitation,” Earthq. Eng. Struct. Dyn., no. September 2015, pp. 193–211, 2016.
[10] A. Carr, A. Puthanpurayil, O. Lavan, and R. Dhakal, “Damping models for inelastic time-history analyses - A
proposed modelling approach,” 16th World Conf. Earthq. Eng., vol. 1488, no. Abstract ID, p. 11, 2017.
[11] A. J. Carr, RUAUMOKO-3D - A program for inelastic time-history analysis. Department of Civil Engineering,
University of Canterbury, New Zealand. 2004.
[12] L. Petrini, C. Maggi, M. J. N. Priestley, and G. M. Calvi, “Experimental verification of viscous damping modeling
for inelastic time history analyzes,” J. Earthq. Eng., vol. 12, no. SUPPL. 1, pp. 125–145, 2008.
[13] E. Smyrou, M. J. N. Priestley, and A. J. Carr, “Modelling of elastic damping in nonlinear time-history analyses
of cantilever RC walls,” Bull. Earthq. Eng., vol. 9, no. 5, pp. 1559–1578, 2011.
[14] Perform 3D User Guide, “Nonlinear Analysis and Performance Assessment for 3D Structures,” User Man.
PERFORM 3D-Computer Struct., no. August, 2006.
[15] C. Kong, “Rapid Direct Displacement-Based Assessment Approach for Bridge Structures,” North Carolina State
University, 2017.
[16] M. J. N. Priestley, G. M. Calvi, and M. J. Kowalsky, Displacement-Based Seismic Design of Structures. 2007.
[17] J. B. Mander, M. J. N. Priestley, and R. Park, “Theoretical Stress-Strain model for Confined Concrete,” J. Struct.
Eng., vol. 114, no. 8, pp. 1804–1826, 1989.
[18] D. M. Boore, “Orientation-independent, nongeometric-mean measures of seismic intensity from two horizontal
components of motion,” Bull. Seismol. Soc. Am., vol. 100, no. 4, pp. 1830–1835, 2010.
[19] L. A. Montejo and L. E. Suarez, “An improved CWT-based algorithm for the generation of spectrum-compatible
records,” Int. J. Adv. Struct. Eng., vol. 5, no. 1, p. 1, 2013.
[20] Caltrans, “Seismic Design Specifications- Second Edition State of California Department of Transportation May
2016,” no. May, 2016.
[21] URS, “Anchorage Port Access Bridge Site-Specific Seismic Hazard Analysis and Development of Seismic
Design Ground Motions Report of Findings,” 2013.

12

© The 17th World Conference on Earthquake Engineering - 2d-0029 -