Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

EARTHQUAKE ENGINEERING & STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2016; 45:2235–2238


Published online 26 May 2016 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eqe.2762

AUTHOR'S REPLY

Response to John Hall’s Discussion (EQE-16-0008) to Chopra and


McKenna’s paper, ‘Modeling viscous damping in nonlinear
response history analysis of buildings for earthquake excitation’

Anil K. Chopra*,† and F. McKenna


Department of Civil and Environmental Engineering, University of California, Berkeley, CA, USA

Received 18 April 2016; Accepted 24 April 2016

KEY WORDS: response to discussion; Rayleigh damping; nonlinear analysis; response history analysis;
damping models; buildings

We welcome the Discussion of our paper by John Hall who has published important papers on the
elusive problem of modeling damping in nonlinear response history analysis (RHA) of structures.
This topic has become very important, now that nonlinear RHA is increasingly employed in
professional practice. We have organized our response to the several points raised in the Discussion
and subsequent email exchanges in four sections: (i) Rayleigh damping based on tangent stiffness;
(ii) penalty elements to model plastic hinges; (iii) limitation of linear viscous damping; and (iv)
dashpot models for the damping matrix.

1. RAYLEIGH DAMPING BASED ON TANGENT STIFFNESS

The Discusser has offered additional reasons, on which we concur, in support of our recommendation
that the use of the tangent stiffness matrix in Rayleigh damping be avoided. Our recommendation was
based on the assertion that defining damping proportional to the tangent stiffness lacks a physical basis
and has conceptual implications that are troubling; several arguments justifying this assertion were
presented. We take this opportunity to present response results to complement those arguments.
We compare the response of an elastoplastic single-degree-of-freedom (SDF) system for two
damping models: constant damping coefficient, c = ζ (2mωn) = ζ (2k/ωn) —the model used since the
inception of research in earthquake response of inelastic systems (Ref. [14] +, Chapter 7)—and
damping coefficient proportional to the tangent stiffness, that is, c = ζ (2kT/ωn). We will refer to these
models as initial-stiffness-proportional damping and tangent-stiffness-proportional damping,
respectively. The second damping model is identical to the first for the elastic branches of the force–
deformation curve, but its damping coefficient is zero for the plastic branches. The SDF system
considered has the following properties: initial, small vibration period Tn = 0.5 s; damping ratio
ζ = 5 %; and normalized yield strength f y ¼ 0:125 (or yield strength reduction factor = 8).
The deformation response of this system for two damping models is presented in Figure 1.1 Observe
that tangent-stiffness-proportional damping leads to larger deformations because it dissipates no energy
through viscous damping while the system is yielding. For this damping model, although the
deformation response appears to be reasonable, the variation of damping force with time seems

*Correspondence to: Anil Chopra, Civil & Environmental Engineering, University of California.

E-mail: chopra@ce.berkeley.edu

Copyright © 2016 John Wiley & Sons, Ltd.


10969845, 2016, 13, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/eqe.2762 by University of Ljubljana, Wiley Online Library on [24/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2236 A. K. CHOPRA AND F. MCKENNA

Figure 1. Response of an elastoplastic SDF system due to El Centro ground motion; Tn = 0.5 s, ζ = 5%, and
f y ¼ 0:125: (a) deformation; (b, c) damping force, (d) damping force–velocity relation; and (e) damping
force–deformation relation. Results are presented for two damping models: initial-stiffness-proportional
damping and tangent-stiffness-proportional damping.

unacceptable: the damping force drops suddenly to zero each time yielding occurs, as observed in
Figure 1(b). To further understand the implications of this damping model, we focus on a single
cycle of response in Figure 1(c)–(e). Evident are (i) the sudden drop in damping force and (ii)
nonlinear damping force–velocity relation with triangular loops (in the first and third quadrants),
behavior that lacks a plausible physical basis. As shown in Figure 1(e), this damping model does
not display complete elliptical loops in the fD  u plot that are characteristic of the classical damping
model. Another unacceptable feature of this damping model is that it implies a negative damping
coefficient when the member stiffness at large deformations becomes negative.
Because of these conceptual problems, use of the tangent-stiffness-proportional damping model is
not recommended, although it greatly reduces the spurious damping forces, as demonstrated in
Section 4 of our paper. The two preferred approaches to eliminate the spurious damping forces
developed in the traditional Rayleigh damping model are (i) MODAL damping, which was the focus
of the original paper, and (ii) removing damping from penalty-type elements to model plastic hinges [4].

2. PENALTY ELEMENTS TO MODEL PLASTIC HINGES

Penalty-type elements, commonly used to impose constraints at contact or sliding surfaces in finite
element modeling of structures, are also useful in modeling plastic hinges. Because penalty elements
provide only constraints, they should not contain damping. Therefore, stiffness contributions from
these penalty elements are not included in the a1k term of the Rayleigh damping matrix [4]. Referred
to as modified Rayleigh damping (Mod-RI), this approach eliminates the spurious damping forces [4, 9].
This is demonstrated by the earthquake response of a concentrated plasticity model of the building
(described in the original paper) presented for two damping models: (i) Mod-RI, modified Rayleigh
model without damping in plastic hinges, and (ii) MODAL, superposition of modal damping
matrices, Eq. (10), with damping specified in all 20 modes of dominantly lateral vibration of the
20-story building. In the first model, a damping ratio of 2% was specified at the first and third
modal frequencies to determine a0 and a1 in Eq. (2); the damping ratios in other modes are defined
by Eq. (3), which will give smaller damping in the second mode and larger damping in modes 4 and
higher (Figure 1). If the structure remains within the linear range, its response is essentially
unaffected by these differences in the damping models (results not presented for brevity). In the
inelastic range, the two damping models lead to similar, but not identical, results for earthquake

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2016; 45:2235–2238
DOI: 10.1002/eqe
10969845, 2016, 13, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/eqe.2762 by University of Ljubljana, Wiley Online Library on [24/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
RESPONSE TO DISCUSSION 2237

response. This is evident from the roof displacement history (Figure 2) and the peak values of floor
displacements, story drifts, and plastic rotations (Figure 3).
Having already discarded Rayleigh damping based on initial stiffness or tangent stiffness, the
preferred models are Mod-RI and MODAL. The first of these two models is easier to understand
and implement, whereas the second model eliminates the spurious damping forces even if rotational
springs that model plastic hinges include damping, as in some computer programs [3, 16].

3. LIMITATIONS OF LINEAR VISCOUS DAMPING

The Discusser has demonstrated that the damping forces can be unrealistically large if the Rayleigh
damping models, RI or Mod-RI, are used. Earthquake analysis of a 10-story shear building with 5%
damping resulted in peak value of total damping force as 7.2% of the weight of the building or 60% of
the yield strength of the building [4]. This problem arises because yielding of structural elements limits
their resisting forces, but no such mechanism exists for limiting the damping forces. Thus, the problem
may exist with any linear damping model, including the MODAL damping model, as the Discusser states.
However, the extent of the problem is not clear because the model analyzed in Ref. [4] was a shear
building, which is not applicable to modern moment-resisting frames based on the strong-column–
weak-beam design or modern coupled shear walls, a common structural system for tall buildings.
We have briefly explored the problem and obtained preliminary results. For this purpose, the
earthquake response of the 20-story building was computed for three damping models:
• Mod-RI, Rayleigh damping with no contribution from rotational springs that model plastic hinges,
• Mod-KI, stiffness-proportional damping, otherwise similar to Mod-RI, and
• MODAL.

Figure 2. Response history of roof displacement for two damping models: Mod-RI and MODAL.

Figure 3. Peak values. (a) floor displacements; (b) story drifts; and (c) plastic hinge rotations—at each floor,
it is the largest value among all hinges—for two damping models: Mod-RI and MODAL.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2016; 45:2235–2238
DOI: 10.1002/eqe
10969845, 2016, 13, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/eqe.2762 by University of Ljubljana, Wiley Online Library on [24/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2238 A. K. CHOPRA AND F. MCKENNA

Table I. Base shear Vbd due to damping expressed as a percentage of base shear Vbs due to stiffness, yield
base shear Vby, and total weight W for two buildings.
Damping model 10-story shear building 20-story SAC building

Vbd/Vbs Vbd/Vby Vbd/W Vbd/Vbs Vbd/Vby Vbd/W


Mod-RI 43a 56a 6.7 20 32 2.4
Mod-KI 86 102 12.2 65 93 7.0
MODAL 41 56 6.7 22 36 2.7
a
OpenSees results, obtained using RI and KI damping models, differ slightly from those in Ref. [4].

Modified Rayleigh damping is based on 5% damping in the first and third modes, Mod-KI on 5%
damping in the first mode, and MODAL on 5% damping in each mode; the choice of 5% damping
matches the value in Ref [4] but differs from the 2% value in our paper, which is representative of
current practice for tall buildings. The total damping force, interpreted as the base shear due to
damping, Vbd, is expressed as a percentage of the base shear due to stiffness, Vbs; yield base shear,
Vby; and total weight, W. Also included for comparison are the corresponding values for the 10-story
shear building defined in Ref. [4].
These results presented in Table I demonstrate that damping forces in the 20-story moment-resisting
frame building are much smaller compared with the 10-story shear building, indicating that the problem
seems less severe than claimed in Ref. [4]. However, the Vbd/Vbs ratio in the nonlinear 20-story frame is
much larger than in the corresponding linear system, suggesting that damping forces are still excessive.
Clearly, this problem needs to be investigated further considering a range of buildings and ground motions
to establish the severity and extent of the problem. Thereafter, the Discusser’s proposal for ‘capped’ viscous
damping, which imposes a limit on the damping forces, deserves further investigation and development.

4. VISCOUS DAMPER MODELS FOR THE DAMPING MATRIX

The Discusser comments that the spatial distribution of viscous dampers over a building that is consistent
with a damping matrix constructed according to the MODAL damping model is unrealistic; this assertion
is illustrated by an example. We agree, but the same concern is also valid for the most common—almost
pervasive—Rayleigh damping model. The mass-proportional part, that is, c = a0m, is difficult to justify
physically. If m is a diagonal matrix, the diagonal coefficients of c = a0m represent linear viscous dampers
that connect each degree-of-freedom (DOF) to a fixed support (Figure 11.4.1 in Ref. [14]). What physical
mechanism can cause such damping forces? It is difficult to come up with a meaningful answer.
Stiffness proportional damping, that is, c = a1k, represents a set of linear viscous dampers that
interconnect the DOFs in the same way as stiffness elements (Figure 11.4.1 in Ref. [14]). It is
perhaps the only model that appeals to intuition because it can be interpreted to model energy
dissipation arising from the velocities associated with member deformations. However, this damping
model implies that modal damping ratios increase linearly with frequency, a trend that is not
supported by experimental data on modal damping ratios determined from forced vibration tests or
from building motions recorded during earthquakes. In contrast, various damping models—Rayleigh
damping, Caughey damping, and MODAL damping—all enable construction of a damping matrix
that is consistent with the aforementioned experimental data. This is the main reason these
alternative models have been of interest in earthquake engineering.
However, the Discusser favors c = a1k for nonlinear RHA as a basis for ‘capped’ viscous damping.
This recommendation could be promising if it can be demonstrated that the inconsistency of this model
relative to the aforementioned experimental evidence has negligible influence on the results of
nonlinear RHA of buildings.

NOTES

1. +
Figures, equations, and references in the original paper are referred to by their original numbers. New figures
are numbered sequentially.

Copyright © 2016 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2016; 45:2235–2238
DOI: 10.1002/eqe

You might also like