Chopra: Prentice-Hall

PAGES JUL. 19, 2000 14:27

ICC Oregon (503) 221-9911
Page 447

11
Damping in Structures

PREVIEW

Several issues that arise in defining the damping properties of structures are discussed in
this chapter. It is impractical to determine the coefficients of the damping matrix directly
from the structural dimensions, structural member sizes, and the damping properties of the
structural materials used. Therefore, damping is generally specified by numerical values
for the modal damping ratios and these are sufficient for analysis of linear systems with
classical damping. The experimental data that provide a basis for estimating these damp-
ing ratios are discussed in Part A of this chapter, which ends with recommended values for
modal damping ratios. The damping matrix is needed, however, for analysis of linear sys-
tems with nonclassical damping and for analysis of nonlinear structures. Two procedures
for constructing the damping matrix for a structure from the modal damping ratios are pre-
sented in Part B; classically damped systems as well as nonclassically damped systems are
considered.

PART A: EXPERIMENTAL DATA AND RECOMMENDED

MODAL DAMPING RATIOS

11.1 VIBRATION PROPERTIES OF MILLIKAN LIBRARY BUILDING

Chosen as an example to discuss damping, the Robert A. Millikan Library building is

a nine-story reinforced-concrete building constructed in 1966–1967 on the campus of the
California Institute of Technology in Pasadena, California. Figure 11.1.1 is a photograph of

447
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 448

448 Damping in Structures Chap. 11

Figure 11.1.1 Millikan Library, California

Institute of Technology, Pasadena. (Courtesy
of G. W. Housner.)

the library building. It is 69 by 75 ft in plan and extends 144 ft above grade and 158 ft above
the basement level. This includes an enclosed roof that houses air-conditioning equipment.
Lateral forces in the north–south direction are resisted mainly by the 12-in. reinforced-
concrete shear walls located at the east and west ends of the building. In the east–west
direction the 12-in. reinforced-concrete walls of the central core, which houses the elevator
and the emergency stairway, provide most of the lateral resistance. Precast concrete win-
dow wall panels are bolted in place on the north and south walls. These were intended to
be architectural but provide stiffness in the east–west direction for low levels of vibration.
The vibration properties—natural periods, natural modes, and modal damping
ratios—of the Millikan Library have been determined by forced harmonic vibration tests
using the vibration generator shown in Fig. 3.3.1. Such a test leads to a frequency-response
curve that shows a resonant peak corresponding to each natural frequency of the structure
(e.g., the frequency-response curve near the fundamental natural frequency of vibration in
the east–west direction is shown in Fig. 11.1.2). From such data the natural frequency and
damping ratio for the fundamental vibration mode were determined by the methods of
Section 3.4.2, and the results are presented in Table 11.1.1. The natural period for this
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 449

Sec. 11.1 Vibration Properties of Millikan Library Building 449

10

Acceleration amplitude, 10-3 × g

8

0
1.3 1.4 1.5 1.6 1.7
Frequency, Hz

Figure 11.1.2 Frequency response curve for Millikan Library near its fundamental natu-
ral frequency of vibration in the east–west direction; acceleration measured is at the eighth
floor. [Adapted from Jennings and Kuroiwa (1968).]

TABLE 11.1.1 NATURAL VIBRATION PERIODS AND MODAL DAMPING RATIOS OF

MILLIKAN LIBRARY

Fundamental Mode Second Mode

Roof
Excitation Acceleration (g) Period (sec) Damping (%) Period (sec) Damping (%)
North–South Direction

Vibration 5 × 10−3 to 0.51–0.53 1.2–1.8 a a

generator 20 × 10−3
Lytle Creek 0.05 0.52 2.9 0.12 1.0
earthquake
San Fernando 0.312 0.62 6.4 0.13 4.7
earthquake

East–West Direction

Vibration 3 × 10−3 to 0.66–0.68 0.7–1.5 b b

generator 17 × 10−3
Lytle Creek 0.035 0.71 2.2 0.18 3.6
earthquake
San Fernando 0.348 0.98 7.0 0.20 5.9
earthquake
a Not measured.
b Data not reliable.
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 450

450 Damping in Structures Chap. 11

mode of vibration in the east–west direction was 0.66 sec. This value increased roughly
3% over the resonant amplitude range of testing: acceleration of 3 × 10−3 g to 17 × 10−3 g
at the roof. The mode shape corresponding to this mode was found from measurements
taken at various floors of the structure but is not presented here. In the vibration test the
damping ratio in the fundamental east–west mode varied between 0.7 and 1.5%, increas-
ing with the amplitude of response. In the north–south direction, the natural period of
the fundamental mode was 0.51 sec, increasing roughly 4% over the resonant amplitude
range of testing: acceleration of 5 × 10−3 g to 20 × 10−3 g at the roof. The damping ra-
tio in this mode varied between 1.2 and 1.8%, again increasing with the amplitude of
response.
The Millikan Library is located approximately 19 miles from the center of the Mag-
nitude 6.4 San Fernando earthquake of February 9, 1971. The strong motion accelero-
graphs installed in the basement and the roof of the building recorded three components
(two horizontal and one vertical) of accelerations. The recorded accelerations in the north–
south direction given in Fig. 11.1.3 show that the peak acceleration of 0.202g at the base-
ment amplified to 0.312g at the roof. Figure 11.1.4 shows that in the east–west direc-
tion the peak acceleration at the basement and roof were 0.185g and 0.348g, respectively.
The accelerations at the roof represent the total motion of the building, which is com-
posed of the motions of the building relative to the ground plus the motion of the ground.
The total displacement at the roof of the building and the displacement of the basement
were obtained by twice-integrating the recorded accelerations. The north–south and east–
west components of the relative displacement of the roof, determined by subtracting the
ground (basement) displacement from the total displacement at the roof, are presented in
Fig. 11.1.5.

0.4
Roof

0
Acceleration, g

0.312g
-0.4
0.4
Basement

0.202g
-0.4
0 10 20 30 40

Time, sec
Figure 11.1.3 Accelerations in the north–south direction recorded at Millikan Library
during the 1971 San Fernando earthquake.
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 451

Sec. 11.1 Vibration Properties of Millikan Library Building 451

0.4

Roof 0
Acceleration, g

-0.4 0.348g
0.4
Basement

0.185g
-0.4
0 10 20 30 40

Time, sec
Figure 11.1.4 Accelerations in the east–west direction recorded at Millikan Library dur-
ing the 1971 San Fernando earthquake.

5 (a) North-south
1.06 in.
0
Roof displacement, in.

-5
5
2.71 in. (b) East-west

-5
0 10 20 30 40

Time, sec
Figure 11.1.5 Relative displacement of the roof in (a) north–south direction; (b) east–
west direction. [Adapted from Foutch, Housner, and Jennings (1975).]

It can be seen that the horizontal accelerations of the roof of the building are larger
and their time variation is different from the ground (basement) accelerations. These dif-
ferences arise because the building is flexible, not rigid. It is seen in the displacement plots
that the displacement amplitude of the roof relative to the basement was 1.06 in. in the
north–south direction and 2.71 in. in the east–west direction. The building vibrated in the
north–south direction with a fundamental-mode period of approximately six-tenths of a
 Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 452

452 Damping in Structures Chap. 11

second, while in the east–west direction this period was about 1 sec. These period values
were estimated as the duration of a vibration cycle in Fig. 11.1.5. More accurate values
for the first few natural periods and modal damping ratios can be determined from the
recorded accelerations at the basement and the roof by system identification procedures
(not presented in this book). The results for the first two modes in the north–south and
east–west directions are presented in Table 11.1.1 for the Millikan Library building.
Acceleration records were also obtained in the basement and on the roof of this
building from the Lytle Creek earthquake of September 12, 1970. The Magnitude 5.4 Lytle
Creek earthquake, centered 40 miles from Millikan Library, produced a peak ground accel-
eration of approximately 0.02g and a roof acceleration of 0.05g in the building, fairly low
levels for measured earthquake motion. System identification analysis of these records led
to values for natural periods and damping ratios shown in Table 11.1.1. For the
low-level vibrations due to the Lytle Creek earthquake, the fundamental periods of 0.52
and 0.71 sec in the north–south and east–west directions, respectively, were similar to—
only slightly longer than—those determined in vibration generator tests. Similarly, the
damping ratios were slightly increased relative to the vibration generator tests.
For the larger motions of the building during the San Fernando earthquake, the nat-
ural periods and damping ratios were increased significantly relative to the values from
vibration generator tests. The fundamental period in the north–south direction increased
from 0.51 to 0.62 sec and the damping ratio increased substantially, to 6.4%. In the east–
west direction the building vibrated with a fundamental period of 0.98 sec, which is 50%
longer than the period of 0.66 sec during vibration generator tests; the damping increased
substantially, to 7.0%.
The lengthening of natural periods at the larger amplitudes of motion experienced
by the building during the San Fernando earthquake implies a reduction in the stiffness
of the structure. The stiffness in the east–west direction is reduced substantially, although
except for the collapse of bookshelves and minor plaster cracking, the building suffered
no observable damage. The apparent damage of the structure due to the earthquake is
also the cause of the substantial increase in damping. Following the earthquake there is
apparent recovery of the structural stiffness, as suggested by measured natural periods
(not presented here) that are shorter than during the earthquake. Whether this recovery
is complete or only partial appears to depend on how strongly the structure was excited
by the earthquake. These are all indications of the complexity of the behavior of actual
structures during earthquakes. We return to this issue in Chapter 13 (Section 13.6) after we
have presented analytical procedures to compute the response of linearly elastic structures
to specified ground motion.

11.2 ESTIMATING MODAL DAMPING RATIOS

It is usually not feasible to determine the damping properties or natural vibration periods
of a structure to be analyzed in the way they were determined for the Millikan Library.
If the seismic safety of an existing structure is to be evaluated, ideally we would like to
determine experimentally the important properties of the structure, including its damping,
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 453

Sec. 11.2 Estimating Modal Damping Ratios 453

but this is rarely done, for lack of budget and time. For a new building being designed,
obviously its damping or other properties cannot be measured.
The modal damping ratios for a structure should therefore be estimated using mea-
sured data from similar structures. Although researchers have accumulated a substantial
body of valuable data, it should be used with discretion because some of it is not directly
applicable to earthquake analysis and design. It is clear from the Millikan Library data that
the damping ratios determined from the low-amplitude forced vibration tests should not be
used directly for the analysis of response to earthquakes that cause much larger motions of
the structure, say, up to yielding of the structural materials. Modal damping ratios for such
analysis should be based on data from recorded earthquake motions.
The data that are most useful but hard to come by are from structures shaken strongly
but not deformed into the inelastic range. The damping ratios determined from structural
motions that are small are not representative of the larger damping expected at higher
amplitudes of structural motions. On the other hand, recorded motions of structures that
have experienced significant yielding during an earthquake would provide damping ra-
tios that also include the energy dissipation due to yielding of structural materials. These
damping ratios would not be useful in dynamic analysis because the energy dissipation in
yielding is accounted for separately through nonlinear force–deformation relationships (see
Section 5.7).
Useful data on damping are slow to accumulate because relatively few structures
are installed with permanent accelerographs ready to record motions when an earthquake
occurs and strong earthquakes are infrequent. The bulk of records of earthquake-induced
structural motions in the United States are from multistory buildings in California: more
than 50 buildings in the greater Los Angeles area during the 1971 San Fernando earth-
quake; over 40 buildings in the Monterey Bay and San Francisco Bay areas during the
1989 Loma Prieta earthquake; and over 100 buildings in the greater Los Angeles area dur-
ing the 1994 Northridge earthquake. Furthermore, the recorded motions of only some of
these buildings have been analyzed to determine their natural periods and modal damping
ratios.
Ideally, we would like to have data on damping determined from recorded earthquake
motions of many structures of various types—buildings, bridges, dams, etc.—using differ-
ent materials—steel, reinforced concrete, prestressed concrete, masonry, wood, etc. Such
data would provide the basis for estimating the damping ratios for an existing structure to
be evaluated for its seismic safety or for a new structure to be designed. Until we accumu-
late a sufficiently large database, selection of damping ratios is based on whatever data are
available and expert opinion. Recommended damping values are given in Table 11.2.1 for
two levels of motion: working stress levels or stress levels no more than one-half the yield
point, and stresses at or just below the yield point. For each stress level, a range of damping
values is given; the higher values of damping are to be used for ordinary structures, and
the lower values are for special structures to be designed more conservatively. In addition
to Table 11.2.1, recommended damping values are 3% for unreinforced masonry struc-
tures and 7% for reinforced masonry construction. Most building codes do not recognize
the variation in damping with structural materials; and typically a 5% damping ratio is
implicit in the code-specified earthquake forces and design spectrum.
 Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 454

454 Damping in Structures Chap. 11

TABLE 11.2.1 RECOMMENDED DAMPING VALUES

Type and Condition Damping Ratio

Stress Level of Structure (%)

Working stress, Welded steel, prestressed 2–3

no more than about concrete, well-reinforced
1
2 yield point concrete (only slight cracking)
Reinforced concrete with 3–5
considerable cracking
Bolted and/or riveted steel, 5–7
wood structures with nailed or
bolted joints

At or just below Welded steel, prestressed 5–7

yield point concrete (without complete
loss in prestress)
Prestressed concrete with no 7–10
prestress left
Reinforced concrete 7–10
Bolted and/or riveted steel, 10–15
wood structures with
bolted joints
Wood structures with nailed 15–20
joints

Source: N. M. Newmark, and W. J. Hall, Earthquake Spectra and Design, Earth-

quake Engineering Research Institute, Berkeley, Calif., 1982.

The recommended damping ratios can be used directly for the linearly elastic anal-
ysis of structures with classical damping. For such systems the equations of motion when
transformed to natural vibration modes of the undamped system become uncoupled, and
the estimated modal damping ratios are used directly in each modal equation. This concept
was introduced in Section 10.9.2 and will be developed further in Chapters 12 and 13.

PART B: CONSTRUCTION OF DAMPING MATRIX

11.3 DAMPING MATRIX

When is the damping matrix needed? The damping matrix must be defined completely if
classical modal analysis is not applicable. Such is the case for structures with nonclassical
damping (see Section 11.5 for examples), even if our interest is confined to their linearly
elastic response. Classical modal analysis is also not applicable to the analysis of nonlinear
systems even if the damping is of classical form. One of the most important nonlinear
problems of interest to us is calculating the response of structures beyond their linearly
elastic range during earthquakes.
 Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 455

Sec. 11.4 Classical Damping Matrix 455

The damping matrix for practical structures should not be calculated from the struc-
tural dimensions, structural member sizes, and the damping of the structural materials used.
One might think that it should be possible to determine the damping matrix for the struc-
ture from the damping properties of individual structural elements, just as the structural
stiffness matrix is determined. However, it is impractical to determine the damping matrix
in this manner because unlike the elastic modulus, which enters into the computation of
stiffness, the damping properties of materials are not well established. Even if these prop-
erties were known, the resulting damping matrix would not account for a significant part of
the energy dissipated in friction at steel connections, opening and closing of microcracks in
concrete, stressing of nonstructural elements—partition walls, mechanical equipment, fire-
proofing, etc.—friction between the structure itself and nonstructural elements, and other
similar mechanisms, some of which are even hard to identify.
Thus the damping matrix for a structure should be determined from its modal damp-
ing ratios, which account for all energy-dissipating mechanisms. As discussed in Sec-
tion 11.2, the modal damping ratios should be estimated from available data on similar
structures shaken strongly during past earthquakes but not deformed into the inelastic
range; lacking such data the values of Table 11.2.1 are recommended.

11.4 CLASSICAL DAMPING MATRIX

Classical damping is an appropriate idealization if similar damping mechanisms are dis-

tributed throughout the structure (e.g., a multistory building with a similar structural system
and structural materials over its height). In this section we develop two procedures for con-
structing a classical damping matrix for a structure from modal damping ratios which have
been estimated as described in Section 11.2. These two procedures are presented in the
following two subsections.

11.4.1 Rayleigh Damping

Consider first mass-proportional damping and stiffness-proportional damping:

c = a0 m and c = a1 k (11.4.1)
−1
where the constants a0 and a1 have units of sec and sec, respectively. For both of these
damping matrices the matrix C of Eq. (10.9.4) is diagonal by virtue of the modal orthog-
onality properties of Eq. (10.4.1); therefore, these are classical damping matrices. Phys-
ically, they represent the damping models shown in Fig. 11.4.1 for a multistory building.
Stiffness-proportional damping appeals to intuition because it can be interpreted to model
the energy dissipation arising from story deformations. In contrast, mass-proportional
damping is difficult to justify physically because the air damping it can be interpreted
to model is negligibly small for most structures. Later we shall see that, by themselves,
neither of the two damping models are appropriate for practical application.
We now relate the modal damping ratios for a system with mass-proportional damp-
ing to the coefficient a0 . The generalized damping for the nth mode, Eq. (10.9.10), is
C n = a0 M n (11.4.2)
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 456

456 Damping in Structures Chap. 11

a0m3 m3
u3
a1k3
k3
a0m2 m2
u2
a1k2
k2
a0m1 m1
u1
a1k1
k1

(a) (b)

Figure 11.4.1 (a) Mass-proportional damping; (b) stiffness-proportional damping.

and the modal damping ratio, Eq. (10.9.11), is

a0 1
ζn = (11.4.3)
2 ωn
The damping ratio is inversely proportional to the natural frequency (Fig. 11.4.2a). The
coefficient a0 can be selected to obtain a specified value of damping ratio in any one mode,
say ζi for the ith mode. Equation (11.4.3) then gives
a0 = 2ζi ωi (11.4.4)
With a0 determined, the damping matrix c is known from Eq. (11.4.1a), and the damping
ratio in any other mode, say the nth mode, is given by Eq. (11.4.3).
Similarly, the modal damping ratios for a system with stiffness-proportional damping
can be related to the coefficient a1 . In this case
a1
Cn = a1 ωn2 Mn and ζn = ωn (11.4.5)
2
wherein Eq. (10.2.4) is used. The damping ratio increases linearly with the natural fre-
quency (Fig. 11.4.2a). The coefficient a1 can be selected to obtain a specified value of the
damping ratio in any one mode, say ζ j for the jth mode. Equation (11.4.5b) then gives
2ζ j
a1 = (11.4.6)
ωj
With a1 determined, the damping matrix c is known from Eq. (11.4.1b), and the damping
ratio in any other mode is given by Eq. (11.4.5b). Neither of the damping matrices defined
by Eq. (11.4.1) are appropriate for practical analysis of MDF systems. The variations
of modal damping ratios with natural frequencies they represent (Fig. 11.4.2a) are not
consistent with experimental data that indicate roughly the same damping ratios for several
vibration modes of a structure.
As a first step toward constructing a classical damping matrix somewhat consistent
with experimental data, we consider Rayleigh damping:
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 457

Sec. 11.4 Classical Damping Matrix 457

c = a0 m + a1 k (11.4.7)
The damping ratio for the nth mode of such a system is
a0 1 a1
ζn = + ωn (11.4.8)
2 ωn 2
The coefficients a0 and a1 can be determined from specified damping ratios ζi and ζ j for
the ith and jth modes, respectively. Expressing Eq. (11.4.8) for these two modes in matrix
form leads to
    
1 1/ωi ωi a0 ζi
= (11.4.9)
2 1/ω j ω j a 1 ζ j

These two algebraic equations can be solved to determine the coefficients a0 and a1 . If
both modes are assumed to have the same damping ratio ζ , which is reasonable based on
experimental data, then
2ωi ω j 2
a0 = ζ a1 = ζ (11.4.10)
ωi + ω j ωi + ω j
The damping matrix is then known from Eq. (11.4.7) and the damping ratio for any other
mode, given by Eq. (11.4.8), varies with natural frequency as shown in Fig. 11.4.2b.
In applying this procedure to a practical problem, the modes i and j with specified
damping ratios should be chosen to ensure reasonable values for the damping ratios in
all the modes contributing significantly to the response. Consider, for example, that five
modes are to be included in the response analysis and roughly the same damping ratio ζ
is desired for all modes. This ζ should be specified for the first mode and possibly for the
fourth mode. Then Fig. 11.4.2b suggests that the damping ratio for the second and third

(a) (b)

ζn ζn
Rayleigh damping
c = a0 m a aω
c = a 1k ζn = 0 + 1 n
ζn = a0/2ωn 2ωn 2
ζn = a1ωn/2
ζ

ω1 ω2 ω3 ω4 ωi ωj
Natural frequencies ωn Natural frequencies ωn

Figure 11.4.2 Variation of modal damping ratios with natural frequency: (a) mass-proportional
damping and stiffness-proportional damping; (b) Rayleigh damping.
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 458

458 Damping in Structures Chap. 11

modes will be somewhat smaller than ζ and for the fifth mode it will be somewhat larger
than ζ . The damping ratio for modes higher than the fifth will increase monotonically with
frequency and the corresponding modal responses will be essentially eliminated because
of their high damping.
Example 11.1
The properties of a three-story shear building are given in Fig. E11.1. These include the floor
weights, story stiffnesses, natural frequencies, and modes. Derive a Rayleigh damping matrix
such that the damping ratio is 5% for the first and second modes. Compute the damping ratio
for the third mode.

200 kips
kstory, kips/in.
ωn = 12.57, 34.33, 46.89 rad/sec
610
400
0.401
   0.803  0.401
φ1 = 0.695 , φ2 =  0 , φ3 = -0.695
610 0.803 -0.803  0.803
400

610

Figure E11.1

Solution
1. Set up the mass and stiffness matrices.
   
400 2 −1 0
1
m= 400 k = 610 −1 2 −1 (a)
386
200 0 −1 1
2. Determine a0 and a1 from Eq. (11.4.9).



1/12.57 12.57 a0 0.05
=2 (b)
1/34.33 34.33 a1 0.05
These algebraic equations have the following solution:

a0 = 0.9198 a1 = 0.0021 (c)

3. Evaluate the damping matrix.

 
3.55 −1.30 0
c = a0 m + a1 k = 3.55 −1.30 (d)
(sym) 1.78
4. Compute ζ3 from Eq. (11.4.8).
0.9198 0.0021(46.89)
ζ3 = + = 0.0593 (e)
2(46.89) 2
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 459

Sec. 11.4 Classical Damping Matrix 459

11.4.2 Caughey Damping

If we wish to specify values for damping ratios in more than two modes, we need to con-
sider the general form for a classical damping matrix (see Derivation 11.1), known as
Caughey damping:

N −1
c=m al [m−1 k]l (11.4.11)
l=0

where N is the number of degrees of freedom in the system and al are constants. The first
three terms of the series are
a0 m(m−1 k)0 = a0 m a1 m(m−1 k)1 = a1 k a2 m(m−1 k)2 = a2 km−1 k (11.4.12)
Thus Eq. (11.4.11) with only the first two terms is the same as Rayleigh damping. Suppose
that we wish to specify the damping ratios for J modes of an N -DOF system. Then J
terms need to be included in the Caughey series and these could be any J of the N terms
in Eq. (11.4.11). Typically, the first J terms are included:

J −1
c=m al [m−1 k]l (11.4.13)
l=0

and the modal damping ratio ζn is given by (see Derivation 11.2)


J −1
ζn = 1
2
al ωn2l−1 (11.4.14)
l=0

The coefficients al can be determined from the damping ratios specified in any J modes,
say the first J modes, by solving the J algebraic equations (11.4.14) for the unknowns
al , l = 0 to J − 1. With al determined, the damping matrix c is known from Eq. (11.4.13),
and the damping ratios for modes n = J + 1, J + 2, . . . , N are given by Eq. (11.4.14).
It is recommended that these damping ratios be computed to ensure that their values are
reasonable; in particular, they should not take on negative values, which may happen de-
pending on the number of terms included in the Caughey series. Negative values of modal
damping ratios are obviously unrealistic because they imply free vibration response that
grows with time instead of decaying with time.
While the general classical damping matrix given by Eq. (11.4.13) makes it possible
to specify the damping ratios in any number of modes, there are two problems associated
with its use. First, the algebraic equations (11.4.14) are numerically ill conditioned because
the coefficients ωn−1 , ωn , ωn3 , ωn5 , . . . can differ by orders of magnitude. Second, if more
than two terms are included in the Caughey series, c is a full matrix, although k is a banded
matrix, and for a lumped-mass system, m is a diagonal matrix. Since the computing cost
for analyzing large systems is increased by a very significant amount if the damping matrix
is not banded, Rayleigh damping is assumed in most practical analyses.
Example 11.2
For the system of Fig. E11.1, evaluate the classical damping matrix if the damping ratio is 5%
for all three modes.
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 460

460 Damping in Structures Chap. 11

Solution
1. Caughey series for a 3-DOF system:
c = a0 m + a1 k + a2 km−1 k (a)
2. Determine a0 , a1 , and a2 from Eq. (11.4.14):
a0 1 a1 a2
ζn = + ωn + ωn3 n = 1, 2, 3 (b)
2 ωn 2 2
or
   
1/12.57 12.57 (12.57)3 a0 0.05
1/34.33 34.33 (34.33)3 a1 =2 0.05 (c)
1/46.89 46.89 (46.89)3 a2 0.05
These algebraic equations have the following solution:
a0 = 0.8377 a1 = 0.0027 a2 = −4.416 × 10−7 (d)
3. Evaluate c. Substituting a0 , a1 , and a2 from Eq. (d) in Eq. (a) gives
 
3.40 −1.03 −0.159
c= 3.08 −1.03 (e)
(sym) 1.62

Derivation 11.1
The natural frequencies ωr and modes φr satisfy
kφr = ωr2 mφr (a)
Premultiplying both sides by φnT km−1 gives
 
φnT km−1 k φr = ωr2 φnT kφr = 0 n = r (b)
wherein the second equality comes from the orthogonality equation (10.4.1a). Premultiplying
both sides of Eq. (a) by φnT (km−1 )2 gives
   
φnT (km−1 )2 k φr = ωr2 φnT km−1 km−1 m φr
 
= ωr2 φnT km−1 k φr = 0 n = r (c)
wherein the second equality comes from Eq. (b). By repeated application of this procedure,
a family of orthogonality relations can be obtained which can all be expressed in a compact
form:
φnT cl φr = 0 n = r (d)
where
 l
cl = km−1 k l = 0, 1, 2, 3, . . . , ∞ (e)
The matrices cl can be written in an alternative form by premultiplying Eq. (e) by the identity
matrix, I = mm−1 :
cl = mm−1 km−1 km−1 · · · km−1 k
 l
= m m−1 k l = 0, 1, 2, 3, . . . , ∞ (f)
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 461

Sec. 11.4 Classical Damping Matrix 461

Premultiplying Eq. (a) by φnT mk−1 and following the procedure above it can be shown that
Eq. (d) is satisfied by another infinite sequence of matrices:
 l
cl = m m−1 k l = −1, −2, −3, . . . , −∞ (g)
Combining Eqs. (f) and (g) gives

∞
 l
c=m al m−1 k (h)
l=−∞

It can be shown that only N terms in this infinite series are independent, leading to Eq.
(11.4.11) as the general form of classical damping matrices.

Derivation 11.2
For the nth mode the generalized damping is

N −1
Cn = φnT cφn = φnT cl φn (a)
l=0

where cl is given by Eq. (f) of Derivation 11.1; and the various terms in this series are

l = 0: φnT co φn = φnT (a0 m)φn = ao Mn

l = 1: φnT c1 φn = φnT (a1 k)φn = a1 ωn2 Mn

l = 2: φnT c2 φn = φnT (a2 km−1 k)φn = a2 ωn2 φnT kφn = a2 ωn4 Mn

wherein Eq. (10.2.4) is used. Thus Eq. (a) becomes

N −1
Cn = al ωn2l Mn (b)
l=0

The damping ratio for the nth mode, Eq. (10.9.11), is given by

N −1
ζn = 1
2 al ωn2l−1 (c)
l=0

which is similar to Eq. (11.4.14).

11.4.3 Superposition of Modal Damping Matrices

An alternative procedure to determine a classical damping matrix from modal damping

ratios can be derived starting with Eq. (10.9.4):
ΦT cΦ = C (11.4.15)
where C is a diagonal matrix with the nth diagonal element equal to the generalized modal
damping:
Cn = ζn (2Mn ωn ) (11.4.16)
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 462

462 Damping in Structures Chap. 11

With ζn estimated as described in Section 11.2, C is known from Eq. (11.4.16) and
Eq. (11.4.15) can be rewritten as
 −1
c = ΦT CΦ−1 (11.4.17)
Using this equation to compute c may appear to be an inefficient procedure because it
seems to require the inversion of two matrices of order N , the number of DOFs. However,
the inverse of the modal matrix Φ and of ΦT can be determined with little computation
because of the orthogonality property of modes.
Starting with the orthogonality relationship of Eq. (10.4.5b),
ΦT mΦ = M (11.4.18)
it can be shown that
 −1
Φ−1 = M−1 ΦT m ΦT = mΦM−1 (11.4.19)
Because M is a diagonal matrix of generalized modal masses Mn , M−1 is known imme-
diately as a diagonal matrix with elements = 1/Mn . Thus Φ and Φ−1 can be computed
efficiently from Eq. (11.4.19).
Substituting Eq. (11.4.19) in Eq. (11.4.17) leads to
c = (mΦM−1 )C(M−1 ΦT m) (11.4.20)
Since M and C are diagonal matrices, defined by Eqs. (11.4.18) and (11.4.15), respectively,
Eq. (11.4.20) can be expressed as
 
N
2ζn ωn
c=m T
φn φn m (11.4.21)
n=1
Mn

The nth term in this summation is the contribution of the nth mode with its damping ratio ζn
to the damping matrix c; if this term is not included, the resulting c implies a zero damping
ratio in the nth mode. It is reasonable to include only the first J modes in Eq. (11.4.21)
that are expected to contribute significantly to the response. The lack of damping in modes
J + 1 to N does not create numerical problems if an unconditionally stable time-stepping
procedure is used to integrate the equations of motion; see Chapter 15.
Example 11.3
Determine a damping matrix for the system of Fig. E11.1 by superposing the damping matrices
for the first two modes, each with ζn = 5%.

Solution
1. Determine the individual terms in Eq. (11.4.21).
2(0.05)(12.57) 2(0.05)(34.33)
c1 = mφ1 φ1T m c2 = mφ2 φ2T m
1.0 1.0
   
0.217 0.376 0.217 2.37 0 −1.19
= 0.651 0.376 = 0 0
(sym) 0.217 (sym) 0.593
 Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 463

Sec. 11.5 Nonclassical Damping Matrix 463

2. Determine c.  
2.59 0.376 −0.969
c = c1 + c2 = 0.651 0.376
(sym) 0.810
Recall that this c implies a zero damping ratio for the third mode.

Example 11.4
Determine the damping matrix for the system of Fig. E11.1 by superposing the damping ma-
trices for the three modes, each with ζn = 5%.

Solution
1. Determine the individual terms in Eq. (11.4.21). The first two terms, c1 and c2 , are
already computed in Example 11.3, and
 
0.809 −1.40 0.810
2(0.05)(46.89)
c3 = mφ3 φ3T m = 2.43 −1.40
1.0
(sym) 0.811
2. Determine c.  

3 3.40 −1.03 −0.159
c= cn = 3.08 −1.03
n=1 (sym) 1.62
Note that this c is the same as in Example 11.2 because ζn = 5% for all three modes in both
examples.

11.5 NONCLASSICAL DAMPING MATRIX

The assumption of classical damping is not appropriate if the system to be analyzed con-
sists of two or more parts with significantly different levels of damping. One such example
is a structure–soil system. While the underlying soil can be assumed as rigid in the anal-
ysis of many structures, soil–structure interaction should be considered in the analysis of
structures with very short natural periods, such as the nuclear containment structure of
Fig. 1.10.1. The modal damping ratio for the soil system would typically be much dif-
ferent than the structure, say 15 to 20% for the soil region compared to 3 to 5% for the
structure. Therefore, the assumption of classical damping would not be appropriate for
the combined structure–soil system, although it may be reasonable for the structure and
soil regions separately. Another example is a concrete dam with water impounded behind
the dam (Fig. 1.10.2). The damping of the water is negligible relative to damping for the
dam, and classical damping is not an appropriate model for the dam–water system. While
substructure methods (not developed in this book) are especially effective for the analysis
of structure–soil and structure–fluid systems, these systems are also analyzed by standard
methods, requiring the damping matrix for the complete system.
The damping matrix for the complete system is constructed by directly assembling
the damping matrices for the two subsystems—structure and soil in the first case, dam
and water in the second. As shown in Fig. 11.5.1, the stiffness and mass matrices of the
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 464

464 Damping in Structures Chap. 11

Axis of radial symmetry

Structure Structure Structure

(ζ = 0.05) k m
I I
Soil Soil
kf mf

Structure I denotes degrees of

Foundation soil c freedom at the interface
(ζf = 0.20) I
Soil
cf

Figure 11.5.1 Assembly of subsystem matrices.

combined structure–soil system are assembled from the corresponding matrices for the
two subsystems. The portion of these matrices associated with the common DOFs at the
interface (I ) between the two subsystems include contributions from both subsystems.
Thus all that remains to be described is the procedure to construct damping matrices for
the individual subsystems, assumed to be classically damped.
In principle, these subsystem damping matrices could be constructed by any of the
procedures developed in Section 11.4, but Rayleigh damping is perhaps most convenient
for practical analyses. Thus the damping matrices for the structure and the foundation soil
(denoted by subscript f ) are
c = a0 m + a1 k c f = a0 f m f + a1 f k f
The coefficients a0 and a1 are given by Eq. (11.4.10) using an appropriate damping ratio
for the structure, say ζ = 0.05, where ωi and ω j are selected as the frequencies of the ith
and jth natural vibration modes of the combined system without damping. The coefficients
a0 f and a1 f are determined similarly and they would be four times larger if the damping
ratio for the foundation soil region is estimated as ζ f = 0.20. As mentioned earlier, ωi and
ω j may be taken as the first and fourth natural frequencies of the combined system if five
modes are to be included in linear analysis by the first procedure of Section 12.14.
The assumption of classical damping may not be appropriate also for structures with
special energy-dissipating devices (Section 6.8) or on a base isolation system (Chapter 20),
even if the structure itself has classical damping. The nonclassical damping matrix for the
system is constructed by first evaluating the classical damping matrix for the structure
alone (without the special devices or isolators) from the damping ratios appropriate for the
structure, using the procedures of Section 11.4. The damping contributions of the energy-
dissipating devices or the isolators are then included to obtain the damping matrix of the
complete system.
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 465

Chap. 11 Problems 465

FURTHER READING

Caughey, T. K., “Classical Normal Modes in Damped Linear Dynamic Systems,” Journal of Applied
Mechanics, ASME, 27, 1960, pp 269–271.
Caughey, T. K., and O’Kelly, M. E. J., “Classical Normal Modes in Damped Linear Dynamic
Systems,” Journal of Applied Mechanics, ASME, 32, 1965, pp. 583–588.
Foutch, D. A., Housner, G. W., and Jennings, P. C., “Dynamic Responses of Six Multistory Buildings
during the San Fernando Earthquake,” Report No. EERL 75-02, California Institute of Technology,
Pasadena, Calif., October 1975.
Hart, G. C., and Vasudevan, R., “Earthquake Design of Buildings: Damping,” Journal of the Struc-
tural Division, ASCE, 101, 1975, pp. 11–30.
Hashimoto, P. S., Steele, L. K., Johnson, J. J., and Mensing, R. W., “Review of Structure Damping
Values for Elastic Seismic Analysis of Nuclear Power Plants,” Report No. NUREG/CR-6011, U.S.
Nuclear Regulatory Commission, Washington, D.C., March 1993.
Jennings, P. C., and Kuroiwa, J. H., “Vibration and Soil–Structure Interaction Tests of a Nine-Story
Reinforced Concrete Building,” Bulletin of the Seismological Society of America, 58, 1968, pp.
891–916.
McVerry, G. H., “Frequency Domain Identification of Structural Models from Earthquake Records,”
Report No. EERL 79-02, California Institute of Technology, Pasadena, Calif., October 1979.
Newmark, N. M., and Hall, W. J., Earthquake Spectra and Design, Earthquake Engineering Research
Institute, Berkeley, Calif., 1982, pp. 53–54.
Rayleigh, Lord, Theory of Sound, Vol. 1, Dover Publications, New York, 1945.
Wilson, E. L., and Penzien, J., “Evaluation of Orthogonal Damping Matrices,” International Journal
for Numerical Methods in Engineering, 4, 1972, pp. 5–10.

PROBLEMS

11.1 The properties of a three-story shear building are given in Fig. P11.1. These include the
floor weights, story stiffnesses, natural vibration frequencies, and modes. Derive a Rayleigh
damping matrix such that the damping ratio is 5% for the first and third modes. Compute the
damping ratio for the second mode.

w/2 Rigid beams w = 100 kips k = 168 kips/in.

ωn = 12.01, 25.47, 38.90 rad/sec

3k/9
w
0.6375
   0.9827   1.5778 
φ1 = 1.2750, φ2 =  0.9829 , φ3 = -1.1270 
7k/9 1.9125 -1.9642   0.4508 
w

Figure P11.1
Chopra: Prentice-Hall
PAGES JUL. 19, 2000 14:27
ICC Oregon (503) 221-9911
Page 466

466 Damping in Structures Chap. 11

11.2 For the system of Fig. P11.1, using a Caughey series determine the classical damping matrix
if the damping ratio is 5% for all three modes.
11.3 Determine a damping matrix for the system of Fig. P11.1 by superimposing the damping
matrices for the first and third modes, each with ζn = 5%. Verify that the resulting damping
matrix gives no damping in the second mode.
11.4 Determine the classical damping matrix for the system of Fig. P11.1 by superimposing the
damping matrices for the three modes, each with ζn = 5%.