Abstract
The principle of the tuned mass damper is described - first the clas-
sic case of a single mass damper attached to a rigid structural mass,
and then the generalization of this theory to one or more tuned mass
dampers on a flexible structure. A brief discussion is given of design
considerations based on the desired damping ratio. Exercises are used
to illustrate the application to a small pedestrian bridge.
Lightly damped structures may develop large amplitude vibrations for loads
acting near the resonance frequency. These vibrations may be reduced by
attaching a secondary mass through a suitably selected spring and damper.
The ‘tuning’ of the spring and damper to produce optimal reduction is an
important feature, and the device is therefore called a tuned mass damper.
1
Figure 2: Tuned mass damper from Langelinie Footbridge 2006.
The tuned mass damper consists of a damper mass 𝑚 attached to the struc-
ture by a spring of stiffness 𝑘 and a viscous damper with parameter 𝑐 as
shown in the figure. These parameters must be chosen to
The mass often consists of smaller steel blocks that can be used to adjust
the resonance frequency of the damper. This is illustrated in Fig. 2, showing
a tuned mass damper fitted inside the closed bridge girder of the Langelinie
Footbridge in Copenhagen. For lighter bridges tuned mass dampers can be
fitted under the bridge deck as shown in Fig. 3
2
2 Basic theory of the tuned mass damper
The energy balance equation is a statement that the rate of change of the
mechanical energy is equal to the rate of dissipation,
𝑑( )
𝐸𝑘𝑖𝑛 + 𝐸𝑝𝑜𝑡 = −𝐷 (1)
𝑑𝑡
In the present contact it is convenient to describe the motion of the combined
system by the motion of the structure 𝑥0 and the relative motion of the
damper 𝑥𝑑 . These variables give simple equations, and they are also the
two important design variables: reduced motion of the structure 𝑥0 (𝑡) for
comfort and use, and limited relative motion of the damper 𝑥𝑑 (𝑡) due to
design constraints on the damper (limited space, length of springs etc.).
With these variables to describe the motion the energy terms take the form,
1
𝐸𝑘𝑖𝑛 = ˙ 20
2 𝑚0 𝑥 + 12 𝑚(𝑥˙ 0 + 𝑥˙ 𝑑 )2
1
𝐸𝑝𝑜𝑡 = 2
2 𝑘0 𝑥0 + 12 𝑘 𝑥2𝑑 (2)
𝐷 = 𝑐 𝑥˙ 2𝑑
𝑚0 𝑥
¨0 + 𝑚(¨
𝑥0 + 𝑥
¨𝑑 ) + 𝑘0 𝑥0 = 𝐹 (𝑡)
(3)
𝑚 (¨
𝑥0 + 𝑥
¨𝑑 ) + 𝑐 𝑥˙ 𝑑 + 𝑘 𝑥𝑑 = 0
The structural mass 𝑚0 and stiffness 𝑘0 are assumed known. The task is to
select the parameters 𝑚, 𝑘 and 𝑐 of the damper to provide optimal vibration
reduction. This problem will be considered in the frequency domain.
3
2.1 Frequency equations
The load and response are assumed harmonic with angular frequency 𝜔,
(5)
−𝜔 2 𝑚 𝑥0 + 𝑘 + 𝑖𝜔 𝑐 − 𝜔 2 𝑚 𝑥𝑑 = 0
[ ]
These equations are now solved for the amplification factor 𝑥0 𝑘0 /𝐹 for the
structural motion and 𝑥𝑑 𝑘0 /𝐹 for the relative motion. Solution of these
equations gives the complex displacement amplitudes
𝑥0 𝑘 − 𝜔 2 𝑚 + 𝑖𝜔 𝑐
= [ ][ ]
𝐹 𝑘0 − 𝜔 2 (𝑚0 + 𝑚) 𝑘 − 𝜔 2 𝑚 + 𝑖𝜔 𝑐 − (𝜔 2 𝑚)2
(6)
𝑥𝑑 𝜔2 𝑚
= [ ][ ]
𝐹 𝑘0 − 𝜔 2 (𝑚0 + 𝑚) 𝑘 − 𝜔 2 𝑚 + 𝑖𝜔 𝑐 − (𝜔 2 𝑚)2
In practice the structure is characterized by its mass 𝑚0 and its natural fre-
quency 𝜔0 , while the damper is characterized by its mass 𝑚𝑑 , frequency and
damping. It is therefore convenient to introduce the following parameters,
𝑘0 𝑚
𝜔02 = 𝜇 =
𝑚0 𝑚0
(7)
𝑘 𝑐
𝜔𝑑2 = 𝜁𝑑 = √
𝑚 2 𝑘𝑚
This gives the amplification factors
𝜔02 𝜔𝑑2 − 𝜔 2 + 2𝑖𝜁𝑑 𝜔𝑑 𝜔
[ ]
𝑥0
= ] (8)
𝐹/𝑘0
[
𝜔 4 − [𝜔 2 + (1 + 𝜇)𝜔 2 ] 𝜔 2 + 𝜔 2 𝜔 2 + 2𝑖𝜁𝑑 𝜔𝑑 𝜔 𝜔 2 − (1+𝜇)𝜔 2
0 𝑑 0 𝑑 0
and
𝑥𝑑 𝜔02 𝜔 2
= ] (9)
𝐹/𝑘0
[
𝜔 4 − [𝜔02 + (1 + 𝜇)𝜔𝑑2 ] 𝜔 2 + 𝜔02 𝜔𝑑2 + 2𝑖𝜁𝑑 𝜔𝑑 𝜔 𝜔02 − (1+𝜇)𝜔 2
where the denominator 𝐹/𝑘0 is the equivalent static displacement of the
structure. The mass ratio 𝜇 is selected, and the damper system is then
characterized by the frequency ratio 𝜔𝑑 /𝜔0 and the damping ratio 𝜁𝑑 . Opti-
mal damper characteristics can now be identified from a frequency analysis
of the two amplification factors.
4
2.2 Frequency analysis
If damping is too low, undesirable dynamic amplification will occur near the
two undamped resonance frequencies, because the damper does not absorb
enough energy. On the other hand, if damping is too high, the damper
restricts the relative motion, and thereby creates the effect of a single mass
𝑚0 + 𝑚 oscillating with frequency
√
𝑘0 𝜔0
𝜔∞ = = √ (10)
𝑚0 + 𝑚 1+𝜇
The transition from two resonance peaks to one with increased damping is
illustrated in Fig. 5.
15
x0 / xstatic
10
5 A
B
0
0.6 0.8 1 1.2
ω/ω
0
Figure 5: Structure response 𝑥0 for 𝜇 = 0.05, 𝜔𝑑 = 𝜔0 and 𝜁: ⋅ ⋅ ⋅ 0, — 0.1,
− ⋅ − 0.3, −− ∞.
It is a remarkable fact that there are two frequencies 𝜔𝐴 and 𝜔𝐵 around the
natural frequency 𝜔0 for which the magnitude of the response is indepen-
dent of the damping parameter 𝑐. These points can be used to determine
a suitable tuning frequency 𝜔𝑑 of the damper. The idea is to select the
damper frequency 𝜔𝑑 such that the dynamic amplification is equal at the
two frequencies 𝜔𝐴 and 𝜔𝐵 , Den Hartog (1956). The optimal damping 𝜁𝑑
can then be determined and the resulting amplification factors for the struc-
ture response 𝑥0 and the relative motion of the damper 𝑥𝑑 can be calculated.
A detailed derivation is presented in the following. The key results are pre-
sented as framed formulae to permit their use without following the details
of their derivation.
5
of a real part that is independent of the damping ratio 𝜁𝑑 and an imaginary
term that has the factor 2𝜁𝑑 . This is expressed in the form
𝑥0 𝐴 + 2𝑖𝜁𝑑 𝐵
= (11)
𝐹/𝑘0 𝐶 + 2𝑖𝜁𝑑 𝐷
The corresponding dynamic amplification is the absolute value, determined
from
𝑥0 2 2 2 2
= 𝐴 + (2𝜁𝑑 ) 𝐵 (12)
𝐹/𝑘0 𝐶 2 + (2𝜁𝑑 )2 𝐷 2
For the magnitude to be independent of 𝜁𝑑 the limits for 𝜁𝑑 = 0 and 𝜁𝑑 → ∞
must be equal, i.e.
𝐴2 𝐵2 𝐴 𝐵
= ⇒ = ± (13)
𝐶2 𝐷2 𝐶 𝐷
This corresponds to the equations
𝐴𝐷 = ±𝐵𝐶 (14)
after cancelling the common factor 𝜔/𝜔𝑑 . Use of the plus sign leads to
the root 𝜔 = 0. This is the static solution where there is no motion, and
therefore no damping force. Use of the minus sign leads to
[ ( )2 ][ ( )2 ] ( )2
𝜔 𝜔 𝜔
1− 2−(2 + 𝜇) = 𝜇 (16)
𝜔𝑑 𝜔0 𝜔0
The roots of this quadratic equation are 𝜔𝐴 2 and 𝜔 2 . The roots 𝜔 2 and 𝜔 2
𝐵 𝐴 𝐵
are not needed explicitly, but only in the form of their sum. The sum of the
roots of a quadratic equation is equal to minus the coefficient of the linear
term, divided by the coefficient of the quadratic term. In this case
( )2
𝜔𝐵 2
( ) [ ( )2 ]
𝜔𝐴 2 𝜔𝑑
+ = 1 + (1 + 𝜇) (18)
𝜔0 𝜔0 2+𝜇 𝜔0
This provides one relation of the frequencies 𝜔𝐴 and 𝜔𝐵 .
The other relation is determined by specifying equal magnitude of the dy-
namic amplification at the frequencies 𝜔𝐴 and 𝜔𝐵 . At these frequencies
6
the response magnitude is independent of 𝜁𝑑 , and the relevant response can
therefore be determined from (8) for 𝜁𝑑 → ∞, where
𝑥0 1
= ( )2 (19)
𝐹/𝑘0 𝜔
1 − (1+𝜇)
𝜔0
This response is that of the combined mass 𝑚0 + 𝑚 moving as a unit.
At the frequency 𝜔𝐴 the response is in phase with the load, while at 𝜔𝐵
it is in opposite phase. Thus, equal response magnitude at 𝜔𝐴 and 𝜔𝐵
corresponds to
1 1
( )2 = − (20)
𝜔𝐵 2
( )
𝜔𝐴
1 − (1+𝜇) 1 − (1+𝜇)
𝜔0 𝜔0
Multiplication with the denominators gives the equation
( )2
𝜔𝐵 2
( )
𝜔𝐴 2
+ = (21)
𝜔0 𝜔0 1+𝜇
This is the condition that the the dynamic amplification is identical at 𝜔𝐴
and 𝜔𝐵 .
The optimal mass ratio 𝜇 is now determined by elimination of the sum of
squares of the frequencies 𝜔𝐴 and 𝜔𝐵 between the equations (18) and (21),
( )2
𝜔𝑑 2+𝜇
1 + (1 + 𝜇) = (22)
𝜔0 1+𝜇
This equation determines the optimal damper frequency as
𝜔𝑑 1
= (23)
𝜔0 1+𝜇
Thus, it is seen that optimal frequency tuning of the secondary system
corresponds to 𝜔𝑑 < 𝜔0 .
The frequency tuning (23) is optimal in a more strict sense than appears
from this classic derivation using damping-independent points. It has been
shown by Krenk (2005) that for this frequency tuning the two vibration
modes of the combined structure-damper system are equal.
7
The roots of this equation are
𝜔𝐴,𝐵 2
( )
𝜇
√
(1 + 𝜇) = 1± (25)
𝜔0 2+𝜇
Substitution of these frequencies back into the structural response (19) for
infinite damping gives
√
𝑥0 1 2+𝜇
= )2 = ± (26)
𝐹/𝑘0 𝜇
(
𝜔𝐴,𝐵
1 − (1+𝜇)
𝜔0
As the dynamic amplification at these two frequencies is independent of
damping the general result is
𝑥 √
0 2+𝜇
= (27)
𝐹/𝑘0 𝐴,𝐵 𝜇
In practice the mass of the damper system is most often considerably less
than that of the original structure, and thus the mass ratio 𝜇 is considerably
less than one. It is seen that this limits the reduction of dynamic amplifi-
cation that can be attained at the frequencies 𝜔𝐴 and 𝜔𝐵 . Thus, a mass
ratio of 𝜇 = 0.02 leads to a dynamic amplification of the structural response
of 10.
8
At this frequency the structural response is
√
𝑥0 𝜇−2𝑖𝜁𝑑 1+𝜇
= (29)
𝐹/𝑘0 𝜇
𝑥0 2 𝜇2 + (2𝜁𝑑 )2 (1+𝜇)
= (30)
𝐹/𝑘0
𝜔∞ 𝜇2
2 1 𝜇
𝜁𝑜𝑝𝑡 = (31)
2 1+𝜇
This is larger than the classic value of Brock (1946), given by Den Hartog
(1956),
2 3 𝜇
𝜁classic = (32)
81+𝜇
The difference in the dynamic amplification of the structural response is
illustrated in Fig. 6a. It is seen that the classic value leads to a dip between
the frequencies 𝜔𝐴 and 𝜔𝐵 , where the amplification obviously √ is the same.
The third graph in the figure refers to a damping ratio 𝜁∗ = 𝜇/1+𝜇, which
is the largest damping that can be applied without changing the nature of
the damping of the combined system.
There are two reasons, why the value of 𝜁𝑑 given by (31) can be considered
preferable relative to the classic value. It has recently been demonstrated
12
10 20
𝑥𝑑 /𝑥𝑠𝑡𝑎𝑡𝑖𝑐
𝑥0 /𝑥𝑠𝑡𝑎𝑡𝑖𝑐
8
15
6
10
4
5
2
0 0
0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
𝜔/𝜔0 𝜔/𝜔0
9
that when the damper frequency is given by (23) the two modes of the
combined structure damper system have the same damping ratio 𝜁, and
that this damping ratio is nearly exactly half that of the damper constant 𝜁𝑑 ,
Krenk (2005),
1
𝜁 ≃ 2 𝜁𝑑 (33)
The implication is that 𝜁𝑜𝑝𝑡 determined by (31) leads to about 12 pct. higher
damping ratio than the classic value.
The second reason that 𝜁𝑑 given by (31) may be termed optimal is related
to the relative motion of the damper. It can be seen from the formula (9)
that the amplification factor 𝑥𝑑 𝑘0 /𝐹 of the relative motion is independent
of damping for 𝜔 = 𝜔∞ ,
𝑥𝑑 1+𝜇
𝐹/𝑘0 = (34)
𝜔∞ 𝜇
For design reasons it is desirable not to exceed this value. It can be shown
that the amplification factor of the relative motion is flat in the sense that
its first three derivatives with respect to frequency vanish at this particular
frequency, when the damping ratio is 𝜁𝑑 as given by (31). This is illustrated
in Fig. 6. Any value of 𝜁𝑑 smaller that 𝜁𝑜𝑝𝑡 will lead to peaks exceeding the
amplification at 𝜔∞ . In contrast, any value of 𝜁𝑑 larger than 𝜁𝑜𝑝𝑡 will lead
to a central peak in the amplification of the structural response.
When the flexible structure is discretized, the equation of motion has the
form
M ẍ + C ẋ + K x = F(𝑡) (35)
where M, C and K are the mass, the damping and the stiffness matrix,
respectively. Typically, in connection with tuned mass dampers individual
modes are considered in the design phase. In mode No. 𝑗 the undamped
angular frequency is 𝜔𝑗 and the motion is described in terms of the mode-
shape vector u𝑗 as
x(𝑡) = u𝑗 𝑟𝑗 (𝑡) (36)
10
𝑟𝑗 (𝑡) is the modal coordinate of mode 𝑗. It is important to realize that for a
given displacement history x(𝑡) its representation in the form (36) depends
on the normalization of the mode-shape vector u𝑗 . Clearly, a consistent
generalization of the theory of tuned mass dampers must be independent of
the particular normalization of the mode-shape vector.
When the modal representation (36) into the equation of motion the result-
ing modal equation of motion becomes
𝑓𝑗 (𝑡)
𝑟¨𝑗 (𝑡) + 2𝜁𝑗 𝜔𝑗 𝑟˙𝑗 (𝑡) + 𝜔𝑗2 𝑟𝑗 (𝑡) = , 𝑗 = 1, 2, . . . , 𝑛 (37)
𝑚𝑗
where the modal mass 𝑚𝑗 is defined by
𝑚𝑗 = u𝑇𝑗 M u𝑗 (38)
The role of the normalization is illustrated by the simple 3-span bridge and
the first mode-shape shown in Fig. 7. In the figure the mode-shape vector
is normalized such that the maximum displacement – at the center – is
1. This implies that a concentrated load acting at the center will directly
represent the modal load 𝑓1 (𝑡), and the modal mass represents the bridge
deck mass activated by a motion of magnitude 1 at the center of the bridge.
Although this particular normalization is not necessary, it may help the
intuitive understanding of the meaning of the modal load and the modal
mass. Often the lower mode-shapes can be considered as modifications of
0.5
−0.5
−1
0 20 40 60 80 100
𝑥
11
a regular sine-function. For that case the modal mass would be half of the
bridge deck mass, and the modal mass of the modified mode-shapes, can
often be estimated directly in the interval 0–0.5 of the total deck mass.
𝑚𝑑,𝑗 u𝑇𝑗 M𝑑 u𝑗
𝜇𝑗 = = 𝑇 (43)
𝑚𝑗 u𝑗 M u𝑗
12
It is seen, that this definition of the modal mass ratio is indeed independent
of the normalization of the mode-shape vector u𝑗 .
It is seen from the modal damper mass and the corresponding modal mass
ratio that dampers are most effective, when mounted at the points of ex-
treme displacement of the mode-shape vector. Thus, for example a damper
mounted in the middle of one of the side spans of the bridge of Fig. 7 will
only experience a modal displacement of around 0.3 times the displacement
at the center of the bridge, and thus it will contribute only about 0.09 of
the modal mass relative to a position at the center of the mid-span. Thus,
it is very important for the efficiency of tuned mass dampers that they are
mounted at points of considerable modal amplitude.
𝑓𝑗
𝜔𝑗2 𝑟𝑗 = (44)
2𝜁𝑗 𝑚𝑗
Consider a simple design example for a very light single span bridge with
simple supports. The total mass of the (homogeneous) bridge deck is 25 ⋅
103 kg. The single span and simple support conditions imply that the modal
13
mass 𝑚𝑗 is half of total mass, i.e. 𝑚𝑗 = 12.5 ⋅ 103 kg. Let the lowest natural
frequency be 2.0 Hz, corresponding to 𝜔1 = 2𝜋 ⋅ 2.0 = 12.6 rad/s.
The Danish Road Directorate (2002) has a comfort criterion for the max-
imum acceleration of smaller pedestrian bridges in normal use, expressed
as ( 𝜔 )0.78
𝑗
max ẍ ≤ 0.25 [ m/s2 ]
2𝜋
For the present natural frequency of 2.0 Hz this gives a maximum accelera-
tion of 𝑎𝑚𝑎𝑥 = 0.429 m/s2 . The load for two persons walking is given by the
Danish Road Directorate (2002) as a point force of harmonic time variation,
However, data collected by Willford (2002) indicate that this may severely
underestimate the pedestrian load that typically would be 2-3 times larger.
In the present example the load amplitude 𝐹0 = 800 N is used. The re-
sponse builds up gradually, but assuming for simplicity that it has reached
an approximately stationary value, when passing the center, the effect of
the load can be calculated by (45), assuming the load to be at the center of
the bridge. With a unit value of the mode-shape vector at the center, the
formula (45) gives the following requirement on the damping ratio,
𝐹0 800
𝜁1 ≥ = = 0.075
2 𝑚1 𝑎𝑚𝑎𝑥 2 ⋅ 12.5 ⋅ 103 ⋅ 0.429
The bridge is assumed to have an intrinsic structural damping of 𝜁𝑠 = 0.01.
It has been shown by Krenk & Høgsberg (2008) that the effective damping of
the to modes created by the introduction of the damper can be approximated
by
𝜁1 ≃ 34 𝜁𝑠 + 12 𝜁𝑑
Thus, the additional damping to be provided by the tuned mass damper is
2𝜁𝑑2 2 ⋅ 0.1352
𝜇1 = = = 0.038
1 − 2𝜁𝑑2 1 − 2 ⋅ 0.1352
14
4 Exercise
The present exercise illustrates the use of tuned mass dampers on a pedes-
trian bridge. The exercise illustrates an increasingly common situation with
slender pedestrian bridges requiring additional damping to prevent excessive
vibrations in the modes with frequencies less than around 5 Hz. The design
problem has been discussed e.g. by Krenk et al. (2005).
𝜌 𝐴 𝐸 𝐼 𝑙1 𝑙2 𝑙3
[kg/m3 ] [m2 ] [N/m2 ] [m4 ] [m] [m] [m]
The bridge considered here is simply supported and has three spans as illus-
trated in Fig. 7. The bridge is modelled as a beam with parameters given in
Table 1. The model has been implemented as a finite element model using
beam elements in a Matlab script-file bridge.m. The finite element model
is contained in the Matlab function bridgemodel.m, called by the script-
file. The script-file bridge.m performs a modal analysis of the bridge, giving
modal frequencies and plots of selected mode-shape vectors. This serves as
a tool for the following design of a tuned mass damper for the bridge.
The following questions should be addressed in a written report, that presents
the problem, its solution, and a brief discussion of the results.
b) Determine the modal mass 𝑚𝑗 of the first five modes using mode-
shape vectors that are normalized to maximum displacement equal to
1. Compare the modal mass to the mass of the bridge deck. For a
sine-shaped mode the modal mass would be exactly half of the bridge
deck mass. How are the results here, and why.
15
d) Consider a design scenario with 25 people passing the bridge simulta-
neously in uncorrelated walk. When people act uncorrelated the√ load
only increases ny the square root of the number of people 𝐹𝑁 ≃ 𝑁 𝐹1 .
In the present case the the amplitude of the total force is set to
𝐹 = 3000 N. i) Calculate a reduced amplitude accounting for the
fact that the people and thereby the load are spread uniformly over
the bridge deck, ii) Find the maximum displacement and acceleration
of the bridge deck, iii) Find the maximum relative displacement 𝑥𝑑 of
the damper.
e) Another design scenario consists of a group of 5 people deliberately
trying to excite vibrations by jumping. They are assumed to act in a
fully correlated way, and therefore the force amplitude is proportional
to the number of people 𝐹𝑁 ≃ 𝑁 𝐹1 . The load amplitude for each
person is assumed to be the double of that of walk, and assuming full
correlation in this case 𝐹 = 6000 N. In this case the load acts locally
at the point of maximum modal displacement. i) Calculate the modal
load for mode 1, ii) Find the maximum displacement and acceleration
of the bridge deck, iii) Find the maximum relative displacement 𝑥𝑑 of
the damper.
5 References
J.E. Brock, A note on the damped vibration absorber, Journal of Applied
Mechanics, Vol. 13, A284, 1946.
Danish Road Directorate, Broteknik, vej- og stibroer. Delastnings og bereg-
ningsregler, Danish Road Directorate, Copenhagen, November 2002.
J.P. Den Hartog, Mechanical Vibrations (4th edn.), McGraw-Hill, New York,
1956. (Reprinted by Dover, New York, 1985).
S. Krenk, Frequency analysis of the tuned mass damper, Journal of Applied
Mechanics, Vol. 72, 936–942, 2005.
S. Krenk, A. Brønden and A. Kristensen, Placement and tuning of resonance
dampers on footbridges, FOOTBRIDGE 2005, 2nd International Confe-
rence, Venice, Italy, December 6-8, 2005. (CD-Rom)
S. Krenk and J.R. Høgsberg, Modal analysis, Structural Dynamics Note 2.
Department of Mechanical Engineering, Technical University of Denmark,
February 2007.
S. Krenk and J.R. Høgsberg, Tuned mass absorbers on damped structures
under random load, Probabilistic Engineering Mechanics, Vol. 23, 408–415,
2008.
M. Willford, Dynamic actions and reactions of pedestrians, FOOTBRIDGE,
International Conference, 10 pp. Paris, 2002.
16