Comparison of Methods for Computing Equivalent Viscous

Damping Ratios of Structures with Added Viscous Damping

Finley A. Charney, Ph.D., P.E.1; and Robert J. McNamara, S.E.2

Abstract: Three analysis methods are compared for determining the equivalent viscous damping ratios in a simple one-story, one-bay
system with three different added damping configurations. The analysis methods are modal strain energy using the undamped mode
shapes, free vibration log decrement, and complex eigenvalue-eigenvector analysis. The damping configurations include a diagonal
configuration, a horizontal configuration, and a toggle-braced system. For each of these systems, a variety of linkage stiffness and damper
capacities are considered. In all cases the structure, exclusive of the dampers, is assumed to remain elastic. A brief discussion is provided
for the behavior of systems with nonlinear viscous dampers. The results of the study show that significantly different values may be
obtained for the effective viscous damping in the same system when different analysis approaches are used. For the toggle-braced system
with relatively flexible linkages, it was shown that the modal strain energy approach consistently produced increasing effective damping
with increased damper capacity, while the other two methods indicated the opposite results. Subsequent analysis of the toggle-braced
structure using the complex eigenvalue-eigenvector approach showed that the effective system damping was indeed decreasing with an
increased device damping constant, and that the behavior could be attributed to the axial flexibility of the toggle brace linkage. The
analysis indicated that the greater the flexibility of the linkage relative to the damping constant of the device, the greater the phase
difference between the deformational velocity in the device and the relative horizontal velocity of the story containing the device. While
the phase shift is evident from response history analysis, the most efficient visualization tool is a plot of the damped mode shapes on a
complex plane.
DOI: 10.1061/共ASCE兲0733-9445共2008兲134:1共32兲
CE Database subject headings: Damping; Passive control; Seismic effects; Structural analysis; Structural dynamics; Computation.

Introduction lized dampers in a variety of configurations, one of which was a

toggle-braced configuration 共Constantinou et al. 2001兲. The
The use of viscous fluid damping devices has become an accepted reader is referred to McNamara 共2001兲 for a detailed description
approach to control building response to wind and earthquake of this structure. The other system analyzed was a long span
loads in building structures 共Constantinou et al. 1998; Hanson and 共approximately 200 m兲 roof truss, which was to be part of a re-
Soong 2001兲. Early in the design phase of such structures, it is tractable roof system for a major league baseball stadium. This
desirable to estimate the effectiveness of the added devices in truss utilized a single linear viscous damper in a linkage that, like
terms of an equivalent viscous damping ratio. One of the most a toggle system, was intended to amplify the deformational ve-
popular approaches for computing the effective damping ratio is locity in the device.
the modal strain energy 共MSE兲 method. This method, which uti- In each of these situations, the MSE approach was used to
lizes the undamped mode shapes, is provided in many commer- estimate the equivalent viscous damping ratio provided by the
cial and academic software packages. The method is also used in devices, and the results indicated that the system damping ratio
building code documents, such as the NEHRP Seismic Provisions increased in direct proportion to the increase in the device damp-
共FEMA 2004兲 and ASCE 7-05 共ASCE 2005兲. ing constant. Subsequent response history analysis 共RHA兲 of the
The authors’ experience with the MSE approach is that it is structures showed an opposite trend; as the device damping con-
generally accurate, but that it can significantly overestimate stant increased, the equivalent damping ratio for the system
damping ratios for certain systems. In fact, this paper was moti- decreased. Further analysis of the systems using complex valued
vated by the lead author’s experience with two building projects. eigenvalues for the damped system produced results that were
One of these projects was a 39-story office building, which uti- nearly identical to those obtained through RHA. Additionally, the
real and complex-valued components of the eigenvectors of the
1
Associate Professor, Dept. of Civil and Environmental Engineering, systems were plotted to illustrate how the geometry and proper-
Virginia Tech, Blacksburg, VA 24061. ties of the system affect the time domain phasing relationship
2
President, McNamara/Salvia Inc., Boston, MA 02111. between the deformational velocity in the device and the defor-
Note. Associate Editor: Michael D. Symans. Discussion open until mational velocity in the story that contains the device.
June 1, 2008. Separate discussions must be submitted for individual pa- The significance of miscalculating the effective damping ratios
pers. To extend the closing date by one month, a written request must be
depends on the application 共wind or seismic兲 and on the final
filed with the ASCE Managing Editor. The manuscript for this paper was
submitted for review and possible publication on February 22, 2005; method of analysis to be utilized. In wind applications, the struc-
approved on April 13, 2006. This paper is part of the Journal of Struc- ture is generally assumed to remain elastic, and if linear damping
tural Engineering, Vol. 134, No. 1, January 1, 2008. ©ASCE, ISSN devices are to be utilized, the MSE approach may be used to
0733-9445/2008/1-32–44/$25.00. determine damping ratios in each mode. These damping ratios

32 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 onal of the bay, and the system shown in Fig. 1共c兲 uses the device
in a horizontal position.
In the analysis, the damping constant of the device and the
axial stiffness of the linkage members were systematically varied,
and the responses of the systems as computed by the various
analysis approaches were compared. As discussed in some detail
later, the equivalent viscous damping ratios computed for the di-
agonal and chevron systems were not nearly as sensitive to sys-
tem parameters and analysis method as was the toggle-braced
system. For this reason, the analysis of the toggle system is the
primary focus of this paper. The response of the toggle system is
presented in the context of the other systems because the com-
parison provides insight that is essential to the understanding of
the behavior of the toggle system.
All of the analysis was performed on structures that remain
linear throughout the response, and for which the damping de-
vices were also linear. However, there is some concern that
toggle-braced systems that utilize nonlinear “softening” devices
may be vulnerable to losses in damper efficiency when the de-
vices are exercised at low deformational velocities. For this rea-
son, the implications of the use of nonlinear devices are briefly
discussed at the end of the paper.

Basic System Descriptions

The toggle-brace system uses dampers in a special linkage

arrangement that amplifies the deformational velocity in the
device, relative to the deformational velocity of the story that
contains the device 共Constantinou et al. 2001兲. Toggle systems are
useful in situations where low interstory drifts occur, but where
these drifts, acting at some frequency, induce uncomfortable vi-
brations in the occupants of the structure. The systems are simi-
larly useful in seismic applications where the primary lateral force
resisting system is very stiff, thereby limiting the travel in the
Fig. 1. Simplified structures used for analysis pistons of the devices if they are placed in the more traditional
configuration.
The amplification provided by the linkage geometry allows for
could then be used directly in a modal response history analysis sufficient stroke in the piston to make the device useful in con-
of the structure. Overestimated damping will lead to underesti- troling vibrations. The amplification factor, which is equal to the
mates of the gust component of wind pressures, and to underes- deformation 共or deformational velocity兲 in the device divided by
timates of story accelerations, both of which depend on damping the concurrent horizontal deformation 共or deformational velocity兲
共ASCE 2005兲. In seismic applications, the effectiveness of the of the story that contains the device, is 3.28 for the structure of
devices might also be evaluated using MSE, and then a subse- Fig. 1共a兲 if the linkage is assumed to be axially rigid. Amplifiers
quent nonlinear response history analysis could be used to com- computed on the basis of axially rigid linkage members are re-
pute building response. Although the response will be accurately ferred to as “ideal” in this paper. As will be explained later, the
computed in this scenario, the devices, thought to be effective, damping ratio obtained in an ideal toggle-braced system, as well
could in fact be quite ineffective. as the damping force in the device, increases by the square of the
amplification factor. Because of the high damper forces, which
are directly transferred to the linkages, the actual amplification
Objectives and Scope factor for toggle systems is highly dependent on the axial stiffness
of the linkages, and may be significantly less than ideal.
The principal objective of the study described in this paper was to In the diagonal configuration 关Fig. 1共b兲兴 the deformation in the
understand why differences in predicted response occur when sys- device is less than the interstory drift, and hence, a deamplifica-
tems are analyzed with the three different approaches 共MSE, tion occurs. For the frame configured as shown, the amplification
RHA, and complex eigenvalue/eigenvector analysis兲, and on the factor is 0.92 if the linkage is assumed to be axially rigid. This
basis of this understanding, provide guidance to engineers that ideal amplification is simply the cosine of the angle ␪ shown in
routinely perform such analysis in practice. To this end, a simple Fig. 1共b兲.
one-story, one-bay system with three different damper configura- The device configuration shown in Fig. 1共c兲 is the most com-
tions was analyzed. The system shown in Fig. 1共a兲 is a toggle- monly used in practice. Here, the device is attached between the
brace system, very similar to that used in the 39-story building. apex of a chevron brace and the frame. If the linkage is axially
The system shown in Fig. 1共b兲 places the device along the diag- rigid, the deformation in the device is identical to the interstory

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008 / 33

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
drift, and hence, there is no amplification or deamplification of Table 1. Actual Linkage Areas as Related to Linkage Area Multiplier
displacement or velocity 共ideal amplification= 1.0兲. Linkage area
multiplier  Area 共cm2兲 Area 共in.2兲
1 6.45 1.0
Detailed System Descriptions
5 32.3 5.0
10 64.5 10.0
All analysis was performed on the structures shown in Fig. 1. The
20 129.0 20.0
pertinent degrees of freedom, shown in Fig. 1, are U1, U2, and
U3. Each system has a first mode undamped period of vibration 40 258.0 40.0
of 4.7 sec. This relatively long period is similar to that of the first 60 387.0 60.0
mode period of vibration of the 39-story building that was men- 80 516.0 80.0
tioned earlier. Aside from the device, no other damping is present Note: The 6.45 cm2 and 32.3 cm2 cross-sectional areas are not practical
in the systems. in a true design. These are included only for the purpose of providing a
In general, the force-velocity relationship for a viscous fluid lower bound on this parameter.
damping device is
and the actual damped frequencies and mode shapes were com-
puted using MATHCAD 共Mathsoft, Inc. 2002兲.
FD = c sgn共v̇兲兩v̇兩␣ 共1兲 For the response history analysis, the first 24 sec of the 1940
El Centro earthquake was used, followed by 30 sec of free vibra-
where FD⫽force in the device; c⫽damping constant of the de- tion. Equivalent viscous damping ratios were obtained by “exact”
vice; v̇⫽deformational velocity in the device; and ␣⫽exponent logarithmic decrement analysis 共Clough and Penzien 2003兲 of the
that controls the shape of the force-velocity relationship. last 30 sec of the response history traces. The earthquake loading
In Eq. 共1兲, sgn is the signum function, which has the value of was used because other aspects of the behavior of the system, not
the sign of the velocity. Note also that the units of the damping reported in this paper, required analysis with a realistic ground
constant c depend on the exponent ␣. For ␣ = 1, the units are in motion. For the purposes of this paper, any loading that induced a
terms of force time per unit length. The value of the damping free vibration response in the system could have been used.
constant c and the exponent ␣ are usually determined from re-
gression analysis of damper force output from laboratory tests run
at a constant frequency, but at variable displacement amplitude.
Results Generated by the Modal Strain Energy
The damping constant will be different if the test is run at a Approach
different frequency. Dampers are typically designed, however,
such that the dependence of c and ␣ on frequency is negligible at The modal strain energy approach is often used to estimate
frequencies close to the structure’s first mode natural frequency. equivalent viscous damping ratios for structures with added linear
Damper properties are also a function of ambient temperature and viscous fluid devices. For each mode k, the equivalent viscous
temperature increase during loading 共Constantinou and Symans damping ratio is computed as follows:
1992兲. For the purpose of this paper, it is assumed that the damper
properties are independent of loading frequency and temperature. ␾Tk CA␾k
For seismic applications, dampers are usually designed such ␰k = 共2兲
2␻k␾Tk M␾k
that the velocity exponent ␣ is in the range of 0.4 to 0.6. This
effectively causes the dampers to soften or “yield” at high veloci- where ␾k⫽undamped mode shape; ␻k⫽undamped circular fre-
ties, thus limiting the forces that are transferred into the super- quency; CA⫽added damping matrix; and M⫽mass matrix. The
structure 共that part of the structure that does not include the mass matrix includes the mass of the story plus some contribution
devices or the linkages兲. When ␣ = 1.0, the device behavior is from the device and linkage. For this analysis, however, M was
linear. All methods of analysis described in this paper require not modified to represent changes in the linkage area or changes
linearity in the devices and in the structure containing the devices, in the capacity 共size兲 of the device.
and hence, such behavior was assumed in the analysis presented For each system shown in Fig. 1, Eq. 共2兲 predicts a modal
below. Nonlinear behavior in the devices is briefly considered in a damping ratio that increases linearly with the device damping
later part of the paper. constant c. The computed damping ratio is independent of the
A variety of linkage cross-sectional areas and device damping cross-sectional area of the linkage because the linkage, while rep-
constants were assumed in the analysis. The basic linkage area A
was taken as 6.45 cm2 共1.0 in.2兲, and the basic damping constant
c was taken as 1.75 kN s / cm 共1.0 k sec/ in.兲. Separate analyses Table 2. Actual Damping Constants as Related to Damping Constant
Multiplier
were performed with linkage area multipliers  of 1, 5, 10, 20,
40, 60, and 80 and with damping constant multipliers ĉ of 10, 20, Damping constant Damping constant Damping constant
30, 40, 50, 60, and 80. Actual linkage areas and device damping multiplier ĉ 共kN s / cm兲 共k sec/ in.兲
constants are shown in Tables 1 and 2, respectively. The basic 10 17.5 10.0
geometry of the structures 共story height, bay width, device orien- 20 35.0 20.0
tation兲 remained constant for each structure type. 30 52.5 30.0
In all cases it was assumed that the structure was responding in 40 70.0 40.0
a linear elastic manner. The modal strain energy analysis and the
50 87.5 50.0
response history analysis was performed using the well-known
60 105.0 60.0
program DRAIN-2Dx 共Prakash et al. 1993兲. For the complex ei-
80 140.0 80.0
genvalue analysis, the system matrices were developed by hand,

34 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 lowed by a period of free vibration. For the first analysis, the
cross sectional area for the toggle linkages and the diagonal brace
was taken as 645 cm2 共100 in.2兲. This large area was used to
remove the effect of linkage flexibility from the analysis and to
provide near-ideal device amplification factors for the systems.
The damping device was modeled explicitly using stiffness pro-
portional damping in DRAIN’s Type-1 truss element. For a device
damping constant of 17.5 kN s / cm 共10 k sec/ in.兲, the following
values were recorded from the response history traces:

For the toggle brace system:

Maximum extensional velocity of device: 23.5 cm/ s 共9.25 in./ sec兲
Maximum lateral velocity of roof beam: 7.16 cm/ s 共2.82 in./ sec兲
共Ideal兲 amplification factor ␳: =23.5/ 7.16= 3.28

For the diagonal brace system:

Fig. 2. First mode damping ratios obtained from the modal strain Maximum extensional velocity of device: 29.7 cm/ s 共11.7 in./ sec兲
energy method Maximum lateral velocity of roof beam: 32.3 cm/ s 共12.7 in./ sec兲
共Ideal兲 amplification factor ␳: =29.7/ 32.3= 0.920

resented in the stiffness matrix, does not contribute to the lateral Put another way, for a unit horizontal velocity at the roof, the
stiffness of the system. More specifically, a lateral load applied at extensional velocity in the toggle device is 3.28, and the
U1 will not produce forces in the linkage because the device, extensional velocity in the diagonally configured device is 0.920.
which connects the linkage to the superstructure, has no static Thus, when the roof is at a unit horizontal velocity, the
axial stiffness. Eq. 共2兲 produces different damping ratios for the deformational velocity in the toggle device is 3.28/ 0.92= 3.56
individual structures of Fig. 1 共given the same linkage area and times the velocity in the diagonal device. Under the same
damping constant兲 because the linkage geometry, as represented maximum steady state harmonic displacement, each structure has
in the stiffness matrix does influence the mode shapes. the same strain energy, but the energy dissipated by the toggle
The damping ratios obtained from the MSE analysis of each device is 3.56⫻ 3.56= 12.7 times that of the diagonal device. This
structure are plotted in Fig. 2. All reported damping ratios are for agrees closely with the finding from the MSE approach that the
the first mode. Note that the cross-sectional area of the linkage is ideal toggle system 共with a very stiff linkage兲, has 12.5 times the
not represented in the figure because this parameter has no effect equivalent viscous damping of the diagonal system.
on modal damping. As seen from the figure, the damping ratio for Response history analysis was also used to estimate the
each system increases linearly with damping constant, but the damping ratio in the first mode of the structure. This was done by
damping ratio for each system, given a particular damping con- the use of the exact logarithmic decrement approach. The results
stant in the device, is different. In fact, the damping ratio for the of the log decrement analysis are shown in Figs. 3共a–c兲, for the
toggle system is 12.5 times that of the diagonal brace system, and structures shown in Figs. 1共a–c兲, respectively. 关Due to the very
10.7 times that of the chevron brace system. As expected, the small amplitudes of free vibration response, it was not possible to
diagonal brace system has less effective damping than the chev- obtain damping ratios by logarithmic decrement for systems with
ron brace. linkage area multipliers of 40, 60, and 80 and with damping
Based on Fig. 2, it appears that the toggle system is the obvi- constant multipliers of 50 and 60. This is the reason that the
ous choice among the three systems because a much higher ef- curves for area multipliers of 40, 60, and 80 appear truncated in
fective damping ratio is obtained for this system for a given Fig. 3共a兲兴. Note that the results from the MSE analysis are
damping constant c. As will be seen from the response history represented by dashed lines in each figure. Note also that the
analysis, however, the true damping ratios for a toggle system vertical scale of Fig. 3共a兲 is ten times that of Figs. 3共b and c兲.
共with a particular device damping constant兲 may be significantly For the toggle system with lower cross-sectional areas in the
less than that indicated in Fig. 2, with the actual value being linkage 关Fig. 3共a兲兴, the equivalent viscous damping ratio
highly dependent on the cross-sectional area of the linkage. The consistently decreases with increased damping constant. For the
response history analysis also shows that the MSE damping ratios larger linkage areas, the damping ratio initially rises with device
for the diagonal and chevron configurations are of acceptable ac- damping constant c, but then reduces sharply. Clearly, the MSE
curacy for all practical situations. approach is a poor predictor for this system. For the systems with
relatively large damping constants and low cross-sectional areas
in the linkages, the true damping ratios are only a fraction of those
Results Generated by Response History Method predicted by MSE.
For the diagonal and chevron system 关Figs. 3共b and c兲兴, the
The response history method of analysis is used to verify the fact equivalent viscous damping always increases with damping
that the toggle system is generally more efficient when compared constant c, and is well predicted by the MSE approach. The only
to the diagonal system. It is also used to show that the MSE exception is the system with the unrealistic linkage area multiplier
approach can significantly overpredict system damping ratios of 1.0 共giving an actual linkage area of 6.25 cm2 or 1.0 in.2兲, and
when the damping constant of the device increases relative to the even this reduces only very slightly.
cross-sectional area of the linkage. It is clear from all of the systems that the lower the linkage area
To demonstrate the efficiency of the toggle system, the sys- relative to the device damping constant, the less efficient the
tems in Figs. 1共a and b兲 were subjected to a dynamic load, fol- system in terms of utilizing the full capacity of the device. For the

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008 / 35

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 Fig. 4. Forces in linkages of each system

diagonal and chevron system, it appears that any practical area for
the linkage will lead to highly efficient utilization of the device.
For the toggle system, however, the lower cross-sectional areas
produce responses that are certainly less than optimum when
compared to MSE. It must be noted, however, that the toggle
systems with linkage area multipliers greater than 10 共linkage
areas greater than 10 in.2 or 64.5 cm2兲 are much more efficient in
terms of the damping ratio achieved than are the other systems.
The primary reason for the poor efficiency of the toggle
system, compared to the other systems, is that the forces and
corresponding axial deformations developed in the linkage of the
toggle system are more than ten times larger than those in the
linkages of the diagonal and chevron systems. The large forces in
the linkage elements of the toggle system are simply a function of
the geometry of the system 共Constantinou et al. 2001兲. The more
collinear the two main elements of the linkage, the larger the
resisting force in the elements for a given damper force. The

Fig. 3. First mode damping ratios computed from logarithmic Fig. 5. Velocity history traces for diagonal and chevron systems
decrement analysis 共linkage area multiplier= 20, damping constant multiplier= 20兲

36 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 Fig. 6. Velocity history traces for toggle-braced system with linkage area multiplier= 20 and various damping constants

forces in the linkages for the various systems are shown in Fig. 4, velocity in the device is slightly lower than that at U1, and that the
which is a plot of the maximum linkage force versus velocities are completely in phase. In Fig. 5共b兲, the velocity traces
cross-sectional area for each system when the damping constant for U1 and the device cannot be distinguished from each other
multiplier of 40 共c = 7,000 kN s / m or 40 k sec/ in.兲. Note that the because there is no amplification or deamplification of
forces in Fig. 4 have been scaled to produce a maximum linkage deformation in this configuration.
force in the toggle system of 100 kN 共22.5 kips兲. Figs. 6共a–c兲 show similar response history traces for the
Figs. 5共a and b兲 show the response history trace for the lateral toggle-braced system. The linkage area multiplier was 20 in each
velocity at U1 and the deformational velocity in the device for the case, but the damping constant multiplier increases from 20 in
diagonal and chevron braced systems, respectively. In each case, Fig. 6共a兲, to 60 in Fig. 6共c兲. Two observations may be made from
the linkage area multiplier is 20, and the damping constant the figures. First, the amplification in horizontal velocity decreases
multiplier is 20. In Fig. 5共a兲, it may be seen that the deformational with increasing device damping constant. These values are

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008 / 37

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 Table 3. Damped First Mode Shape for System with Linkage Area
Multiplier= 20 and Damping Constant Multiplier= 40
Complex Phase Phase
DOF coordinates Magnitude 共rad兲 共deg兲
U1 1.0 1.000 0.00 0.00
U2 −0.847+ 1.24i 1.503 2.170 124.3
U3 3.20− 2.81i 4.259 −0.722 −41.4

The procedure to obtain the damped system properties is pre-

sented in detail by Liang and Lee 共1991兲, and is summarized
below. It is noted that this important method of analysis is unfa-
miliar to most practicing engineers, is rarely covered in graduate
共civil engineering兲 courses in structural dynamics, and is only
lightly covered in modern texts 共Clough and Penzien 2003;
Chopra 2001兲. Some older texts 共Humar 1990; Hurty and Ruben-
Fig. 7. Damping ratios from complex eigenvalue analysis
stein 1964兲 provide a more complete treatment, as does the recent
monograph by Hanson and Soong 共2001兲. Goel 共2001兲 provides a
thorough background of the approach and an excellent example of
approximately 2.7, 1.7, and 1.2 when the damping constant its use in the analysis of systems with dampers positioned to
multiplier is 20, 40, and 60, respectively. Recall that the reduce torsional response in buildings. Unfortunately, the ap-
theoretical amplification is 3.28. The next observation is that the proach is virtually absent in all commonly used structural analysis
larger the device damping constant relative to the linkage area, the software. For example, the commercial programs ETABS
larger the phase difference between the velocity at U1 and in the 共CSI 2002兲 and SAP2000 共CSI 2004兲 do not have complex eigen-
device. As can be seen from Fig. 6共c兲, the phase difference is so value capabilities, nor do the academic 共research兲 programs
extreme that for some pulses it appears that the deformational DRAIN-2Dx 共Prakash et al. 1993兲 and OpenSEES 共http://
velocity in the device is nearly zero when the lateral velocity 共and opensees.berkeley.edu/兲.
kinetic energy兲 at U1 is at its maximum. The damped mode shapes and frequencies for the damped sys-
tem are obtained by creating the system matrix H, which is ob-
tained by transforming the equations of motion into state space
Results of Analysis Using Complex Eigenvalues format. The system matrix is as follows:
and Eigenvectors

While comparing the velocity history traces of various points in a

H= 冋 − M−1CA
I
− M−1K
Z
册 共3兲

system is a useful method to evaluate the efficiency of a system’s For the systems shown in Fig. 1, H is of size 6 ⫻ 6, and each
damper deployment, a simpler method may be to compute the submatrix 共M, CA, and K兲 of Eq. 共3兲 is 3 ⫻ 3. K⫽stiffness matrix
damped modal properties of the system, and compare these to the of the structure; I⫽3 ⫻ 3 identity matrix; and Z⫽3 ⫻ 3 null 共zero兲
undamped properties. Because damper deployments in real struc- matrix. The eigenvalues and eigenvectors of the state matrix sat-
tures are never truly proportional to mass and/or stiffness of the isfy the relationship
superstructure, the systems are nonclassically damped and the
H = P⌳P−1 共4兲
frequencies and mode shapes will be in terms of complex
numbers. where the diagonal eigenvalue matrix is

⌳= 冋 ⌳1
⌳*1
册 共5兲

and the matrix of eigenvectors is

P= 冋 P 1⌳ 1
P1
P*1 ⌳*1
P*1
册 共6兲

The asterisk superscript in the terms of Eqs. 共5兲 and 共6兲 represent
the complex conjugates of the nonasterisk terms. The complex
mode shapes are stored in P1. The complex-valued eigenvalues

Table 4. Undamped First Mode Shape for System with Linkage Area
Multiplier= 20 and Damping Constant Multiplier= 40
Complex Phase Phase
DOF coordinates Magnitude 共rad兲 共degrees兲
U1 1.0 1.000 0 0
U2 −1.579 1.579 3.14 180
Fig. 8. Circular frequencies from complex eigenvalue analysis U3 4.86 4.86 0 0

38 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 Fig. 11. Harmonic modal amplitude traces for  = 20 and ĉ = 40

ratio for the overdamped mode is greater than 1.0, and must be
found with a different expression 共Inman 2001兲.
It is important to note that this paper does not follow the
nomenclature of some texts 共Humar 1990兲 that refer to the fre-
quencies ␻ obtained from Eq. 共9兲 as “undamped” frequencies.
These frequencies would, in fact, be “undamped” only if CA in
Fig. 9. Argand diagram for  = 20 and ĉ = 40 Eq. 共3兲 is identically zero 共null兲. The confusion arises from the
fact that Eqs. 共3兲–共9兲 will produce identical frequencies for a sys-
tem with no added damping 共CA = null兲 and for a system with
occur in conjugate pairs 共⌳1 , ⌳*1 兲. For a given mode, the pair k added damping when the added damping is proportional to mass
can be written as and or stiffness.
In the case of proportionally damped systems, Eqs. 共3兲–共9兲 are
␭C,k = − ␰C,k␻C,k ± 冑1 − ␰2C,k␻C,k 共7兲
not required because the standard eigenvalue solution will pro-
where the subscript “C” is used to indicate that the quantities are duce exactly the same frequencies and mode shapes that are pro-
derived from the complex eigenvalue analysis. duced by Eqs. 共3兲–共9兲. Additionally, the modal damping ratios for
If Ak and Bk represent the real and imaginary parts of the proportionally damped systems may be determined from the MSE
known eigenvalues approach 关Eq. 共2兲兴 in lieu of Eqs. 共3兲–共10兲. As noted before, how-
ever, it is virtually impossible to have a real structure that is
␭C,k = Ak + iBk 共8兲 proportionally damped. For the systems under consideration, the
the frequency and damping 共for each mode兲 may be found as flexibility in the linkages, taken alone, renders the systems non-
proportionally damped.
␻C,k = 冑A2k + B2k = 兩␭C,k兩 共9兲 To avoid the confusion in nomenclature, the frequencies ob-
tained from Eq. 共9兲 are referred to as the “complex frequencies”
and the modal damping ratios obtained from Eq. 共10兲 are called
− Ak − Ak the “complex damping ratios.” Further, the mode shapes P of Eq.
␰C,k = 共10兲
冑A2k + B2k = 兩␭C,k兩 共6兲 are referred to as the “complex mode shapes.” The damped
frequencies for any system, whether proportionally or nonpropor-
Note that Eqs. 共9兲 and 共10兲 are only applicable for underdamped
modes, for which the eigenvalue is a complex number. For over- tionally damped, are given by the expression ␻D = ␻冑1 − ␰2, where
damped modes, the eigenvalue is real, and Eq. 共10兲 will produce the appropriate values are used for ␻ and ␰. The reader is referred
a damping ratio of exactly 1.0 for the mode. The true damping to Lallement and Inman 共1995兲 for a more thorough discussion on
the interpretation of the complex eigenvalues obtained in the
analysis of nonclasically damped systems.
Eqs. 共3兲–共10兲 were used to analyze the structure of Fig. 1共a兲
with a variety of toggle axial stiffness values and device damping
constant values. The toggle link area multiplier ranged from 10 to
80, and the device damping constant multiplier varied from 10 to

Fig. 10. Argand diagram for  = 20 and ĉ = 0 Fig. 12. Harmonic modal amplitude traces for  = 20 and ĉ = 0

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008 / 39

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 Fig. 13. Effect of increased damping on first mode shape coordinates. Â = 20 and ĉ varies.

80. For example, a system with a toggle linkage area multiplier of of the squares of the real and imaginary terms of each coordinate兲
20 and a device damping constant multiplier of 40 has the fol- and the phase angles at each DOF. The phase angle is measured
lowing first-mode eigenvalue: ␭C,1 = −0.562± 1.811i. Using Eqs. counterclockwise from the real axis. The coordinates for the mode
共9兲 and 共10兲, the damped frequency ␻C,1 is 1.89 rad/ sec and the are plotted on the complex plane in Fig. 9. Clearly evident from
damping ratio ␰C,1 is 0.296. this figure is the fact that the coordinates of U2 and U3 are not
Complex damping ratios for all of the systems analyzed are collinear with U1.
plotted in Fig. 7. As may be seen from this figure, the trend first For contrast, the undamped mode shape for the same system,
noticed from the response history analysis 关Fig. 3共a兲兴 is repeated. when normalized to have U1 = 1, will have U2 and U3 as real
The MSE approach serves as a 共near兲 upper bound to the pre- numbers, and will have all coordinates of the mode lying along
dicted equivalent viscous damping. The reduction in the effective the real axis. The various modal properties for the undamped
damping ratio is more gradual for the lower cross-sectional areas, system are listed in Table 4, and the coordinates are plotted in
and is rather sharp for the higher cross-sectional areas. Fig. 10.
The complex frequencies for the various systems are plotted in Figs. 9 and 10 may be interpreted as Argand diagrams, where
Fig. 8. For systems with lower linkage areas, the frequencies the vectors are rotating counterclockwise at the appropriate fre-
gradually increase as the damping constant increases. For systems quency 共1.89 rad/ sec in Fig. 9, and 1.32 rad per sec in Fig. 10兲. It
with higher linkage areas, there is a sudden increase in complex is illustrative to represent the vibration of the system as three
frequency 共at a damping constant multiplier of between 40 and sine waves, one each for U1, U2, and U3. This is done for the
50兲, and then the frequency remains relatively constant. For the damped system in Fig. 11, and for the undamped system in Fig.
system with linkage area and damping constant multipliers of 80, 12. As may be seen, U2 and U3 are not in phase with U1 when
the complex frequency is 3.4 rad/ sec, which is about 2-1 / 2 times the system is damped, and are in phase when the system is
the undamped frequency of 1.32 rad/ sec. undamped.
The first complex mode shape for the system with a linkage As the damping increases 共for a given linkage area兲, the phase
area multiplier of 20 and damping constant multiplier of 40 are differences in U1, U2, and U3 become larger, and the magnitude
given in the second column of Table 3. Note that the mode shape of U2 and U3 with respect to U1 becomes smaller. This is shown
has been normalized to give a value of 1.0 at U1. Given also in in a series of Argand diagrams in Fig. 13. Each of the individual
the table are the magnitudes 共equal to the square root of the sum diagrams in the figure are for the toggle system with a linkage

40 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 Fig. 14. Harmonic responses for U1 and device using damped mode shapes

area multiplier of 20, and with damping constant multipliers in- Systems Utilizing Nonlinear Viscous Dampers
creasing from 0 to 80 in increments of 10. For the last system,
In the previous analysis, all systems were linear elastic and the
with a damping constant multiplier of 80, U1 and U2 are almost
force-velocity relationship for the devices was assumed to be lin-
90 deg out of phase, and U2, which is greater than U1 in the ear. As mentioned earlier, however, linear devices are rarely used
undamped system, is less than U1 in the highly damped system. in earthquake engineering practice. Unfortunately, some analysts
It is even more instructive to show the relationship between erroneously attempt to predict effective damping ratios for non-
U1 and the deformation acting along the axis of the device. This linearly damped systems by log decrement. This approach will
is done in Fig. 14 for each of the damped systems represented in not work because the equivalent viscous damping is amplitude
Fig. 13. At the left of each subfigure in Fig. 14, several properties dependent.
for each system are summarized. Here, ␻⫽frequency; The force-velocity relationship for a nonlinear viscous device
␰⫽damping ratio; ␪⫽phase lag between U1 and the device defor- is given in Eq. 共1兲 of this paper. Hanson and Soong 共2001兲 show
mation; and ␳⫽effective amplification factor, which is equal to that the energy dissipated per cycle for a harmonically deformed
nonlinear viscous device is
the maximum value of device deformation divided by the maxi-
mum 共nonconcurrent兲 deformation at U1. This value decreases ED,NL = ␩cNLv1+␣
max␻¯␣ 共11兲
from 3.37 for the undamped system to 1.61 for the system with a where vmax⫽maximum amplitude of the harmonic displacement;
damping constant multiplier of 80. ␻
¯ ⫽frequency of loading 关v共t兲 = vmax sin共␻
¯ t兲兴; cNL⫽device

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008 / 41

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 Fig. 14. 共Continued兲.

damping constant for a nonlinear viscous device; ␣⫽velocity An example of the use of Eq. 共13兲 is shown in Fig. 15. For this
exponent; and ␩ = 4 ⫻ 2␣关⌫2共1 + 0.5␣兲兴 / 关⌫共2 + ␣兲兴, where example, ␣ was taken as 0.5, and using CNL = 55.8 kN 共s / cm兲0.5
⌫⫽gamma function. 关20.0 k 共s / in.兲0.5兴. The system was excited at a constant frequency
For a linear viscous damper under the same harmonic loading, of 2.0 rad per sec. The plot in Fig. 15 shows how the effective
the energy dissipated per cycle is damping constant varies with loading amplitude. As may be seen,
ED,L = ␲cLv2max␻
¯ 共12兲 at very low amplitudes of displacement, the equivalent linear
damping constant is quite large. At much higher displacement
where cL⫽device damping constant for the linear viscous device. amplitudes, the equivalent linear damping constant is much lower.
By assuming equal energy dissipation in the linear and The effect shown in Fig. 15 is actually desired in seismic
nonlinear devices, Eqs. 共11兲 and 共12兲 may be used to find an
design situations because the designer wants to limit the damping
equivalent damping constant for a linear device loaded at the same
force at very high velocities. When the devices are deployed in a
frequency and amplitude as a nonlinear device
toggle configuration and are subjected to small deformational
␩ velocities, however, it is possible that less than desired damping
cL = ¯ 兲␣−1
cNL共vmax␻ 共13兲 will be obtained for the system because the effective damping
␲
constant is large compared to the area of the linkage. It has already
It is noted that a similar equation was derived by Filiatrault et al. been shown that significant loss of damper efficiency may occur
共2001兲. Lin and Chopra 共2002兲 also provide interesting insight
in such situations.
into behavior of systems with nonlinear viscous fluid dampers.

42 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 such devices should be assessed using models that allow yielding
in the superstructure when such yielding is expected. For struc-
tures designed for seismic loading, yielding in the superstructure
will likely occur, even when the system is very highly damped.
For systems designed for wind, assessment of performance using
a linear superstructure is always adequate.

Recommendations

The principal recommendation is that effective modal damping

ratios should not, in general, be computed by the modal strain
energy method. This approach is only applicable for proportion-
ally 共or nearly proportional兲 damped systems. Real structures with
added viscous fluid dampers are not proportionally damped, and
toggle systems, in particular, are highly nonproportionally
damped.
Fig. 15. Equivalent linear damping constant for system with ␣ = 0.5 The response history approach, using log decrement, is appli-
and CNL = 55.81 kN 共s / cm兲0.5 关20.0 k共s / in.兲0.5兴. Cyclically loaded at cable to all systems with linear devices or with nonlinear devices,
2.0 rad/ sec. but it is important to note that the damping ratios obtained for
systems with nonlinear devices are amplitude dependent. If the
damping exponent ␣ is less than 1.0, increasingly higher effective
Summary and Conclusions damping ratios will be obtained as one progresses further into the
free vibration history.
The analysis described herein has shown that a viscous fluid The effectiveness of devices in providing damping may also
damper placed in a toggle configuration is almost always more be determined by the complex eigenvalue approach. If the devices
efficient than using the same damper in a diagonal or chevron have linear force-velocity relationships, the complex eigenvalue
brace configuration. This efficiency is manifested in theoretically approach may be used directly. If the devices are nonlinear, the
higher damping ratios, but also in higher axial forces in the link- equivalent viscous damping constant, as obtained from Eq. 共13兲,
age system. Due to the large axial forces in the linkage, there may should be used at several target displacements. In particular, if the
be very significant losses in efficiency in the toggle system, par- devices are used to control discomforting lateral accelerations, an
ticularly when the cross-sectional area of the mechanical linkages effective damping constant should be determined at the appropri-
is low relative to the device damping constant. Although damping ate small displacements.
losses due to axial flexibility in the linkage also occur in the Unfortunately, most commercial software does not have the
diagonal and chevron systems, such losses are negligible when capability to perform complex eigenvalue analysis. It is recom-
reasonable linkage areas are used. mended that such features be added as soon as possible, together
The modal strain energy approach, which is based on the use with graphical visualization tools that are able to display phasing
of the undamped mode shapes to predict modal damping ratios, is relationships in the complex mode shapes.
not able to detect the potential losses in efficiency because the The damping ratios obtained from the complex eigenvalues
undamped mode shapes are not a function of 共or are insensitive should be used only for evaluating the general effectiveness of the
to兲 the axial flexibility of the linkages. Response history analysis devices, and should not be used in modal response history analy-
is able to detect losses in damper efficiency, but this approach sis. Final analysis must be performed using full response history
requires log decrement calculations of free vibration response his- analysis, with the devices explicitly modeled as discrete damping
tories, or requires direct comparison of deformational velocity elements. If the devices are nonlinear, or if it is expected that the
histories in the structure and the devices. structure will deform inelastically, nonlinear response history
For linear systems, computation of the complex eigenvalues analysis is essential. The “fast nonlinear analysis” approach de-
and eigenvectors provides accurate information on modal damp- veloped by Wilson 共2002兲 and utilized in SAP2000 and ETABS,
ing. Furthermore, the eigenvectors, when plotted in the complex is particularly efficient when only a few devices are arranged in
plane, allow for rapid assessment of system sensitivity to linkage an otherwise linear system.
flexibility: The larger the phase differences between the different
components of the modes, the larger the reduction in system
efficiency.
When devices are used with velocity exponents of less than 1, Notation
the loss in damper efficiency due to axial flexibility of the link-
ages will be most important when the deformational velocities in The following symbols are used in this paper:
the device are small. While this is not likely to be a problem for  ⫽ linkage area multiplier;
seismic resistant systems, it can present a significant problem in CA ⫽ added damping matrix;
structures for which the principal motivation for the use of the c ⫽ device damping constant;
devices is mitigation of wind induced vibrations. ĉ ⫽ device damping constant multiplier;
Finally, it is noted that all analysis described in this paper cL ⫽ damping constant for a linear device;
assumed that the superstructure remained linear elastic. While this cNL ⫽ damping constant for a nonlinear device;
assumption is adequate for determining basic device and configu- ED,L ⫽ energy dissipated by a linear viscous fluid
ration efficiency, the true performance of a system that utilizes damper;

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008 / 43

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
 ED,NL ⫽ energy dissipated by a nonlinear viscous fluid tional Center for Earthquake Engineering Research, State Univ. of
damper; New York, Buffalo, N.Y.
FD ⫽ viscous force in damping device; Constantinou, M. C., Tsopelas, P., Hammel, W., and Sigaher, A. N.
H ⫽ state space matrix; 共2001兲. “Toggle-brace-damper seismic energy dissipation systems.” J.
Struct. Eng., 127共2兲, 105–112.
I ⫽ identity matrix;
CSI. 共2002兲. ETABS version 8 user reference manual, Computers and
i ⫽ 冑−1; Structures, Berkeley, Calif.
K ⫽ stiffness matrix; CSI. 共2004兲. SAP2000 version 9 basic analysis reference manual, Com-
k ⫽ mode number; puters and Structures, Berkeley, Calif.
M ⫽ mass matrix; FEMA. 共2004兲. “NEHRP recommended provisions for seismic regula-
P ⫽ matrix of complex eigenvectors; tions for new buildings and other structures.” FEMA Rep. No. 450-1,
v ⫽ deformation in viscous fluid device; Federal Emergency Management Agency, Washington, D.C.
v̇ ⫽ deformational velocity in device; Filiatrault, A., Tremblay, R., and Wankitkorkul, A. 共2001兲. “Performance
vmax ⫽ maximum deformation in viscous fluid device; of passive damping systems for the seismic retrofit of steel moment
␣ ⫽ velocity exponent in damping device; resisting frames subjected to near field ground motions.” Earthquake
Spectra, 17共3兲, 427–456.
␩ ⫽ constant;
Goel, R. K. 共2001兲. “Simplified analysis of asymmetric structures with
⌳ ⫽ matrix of complex eigenvalues; supplemental damping.” Earthquake Eng. Struct. Dyn., 30共9兲, 1399–
␭C ⫽ complex eigenvalue; 1416.
␰ ⫽ damping ratio 共computed from undamped mode Hanson, R. D., and Soong, T. T. 共2001兲. Seismic design with supplemen-
shapes兲; tal energy dissipation devices, Earthquake Engineering Research In-
␰C ⫽ damping ratio 共computed using complex mode stitute, Oakland, Calif.
shapes兲; Humar, J. L. 共1990兲. Dynamics of structures, Prentice-Hall, Englewood
␳ ⫽ amplification in viscous fluid device; Cliffs, N.J.
␾ ⫽ mode shape; Hurty, W. C., and Rubenstein, M. F. 共1964兲. Dynamics of structures,
␻ ⫽ undamped circular frequency; Prentice-Hall, Englewood Cliffs, N.J.
␻C ⫽ complex circular frequency; Inman, D. J. 共2001兲. Engineering vibration, 2nd Ed., Prentice-Hall, Upper
␻D ⫽ damped circular frequency; and Saddle River, N.J.
Lallement, G., and Inman, D. 共1995兲. “A tutorial on complex eigenval-
␻
¯ ⫽ harmonic loading frequency.
ues.” Proc., 13th Int. Modal Analysis Conf., Nashville, Tenn.
Liang, Z., and Lee, G. C. 共1991兲. “Damping of structures: Part 1—Theory
of complex damping.” Rep. No. NCEER-91-0004, National Center for
References Earthquake Engineering Research, State Univ. of New York, Buffalo,
N.Y.
ASCE. 共2005兲. Minimum design loads for buildings and other structures, Lin, W. H., and Chopra, A. K. 共2002兲. “Earthquake response of elastic
ANSI/ASCE 7-05, American Society of Civil Engineers, Reston, Va. SDF systems with non-linear fluid viscous dampers.” Earthquake
Chopra, A. K. 共2001兲. Dynamics of structures, 2nd Ed., Prentice-Hall, Eng. Struct. Dyn., 31共9兲, 1623–1642.
Upper Saddle River, N.J. Mathsoft, Inc. 共2002兲. Mathcad 11 user’s guide, Cambridge, Mass.
Clough, R. W., and Penzien, J. 共2003兲. Dynamics of structures, 2nd Ed., McNamara, R. J. 共2001兲. “Practical solution to reduce the wind induced
共revised兲, Computers and Structures, Berkeley, Calif. response of tall buildings.” Proc., 6th World Congress on Tall Build-
Constantinou, M. C., Soong, T. T., and Dargush, G. F. 共1998兲. Passive ings, Melbourne, Australia.
energy dissipation systems for structural design and retrofit, Multidis- Prakash, V., Powell, G. P., and Campbell, S. 共1993兲. Drain-2DX base
ciplinary Center for Earthquake Engineering Research, Buffalo, N.Y. program description and user guide, Dept. of Civil Engineering,
Constantinou, M. C., and Symans, M. 共1992兲. “Experimental and analyti- Univ. of California, Berkeley, Calif.
cal investigation of seismic response of structures with supplemental Wilson, E. L. 共2002兲. Three dimensional static and dynamic analysis of
fluid viscous dampers.” Technical Rep. No. NCEER-92-0032, Na- structures, Computers and Structures, Inc., Berkeley, Calif.

44 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / JANUARY 2008

Downloaded 22 Feb 2010 to 128.82.252.58. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright