EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS, VOL, LO, 381401 (1982)

OPTIMUM ABSORBER PARAMETERS FOR VARIOUS

COMBINATIONS OF RESPONSE AND EXCITATION
PARAMETERS

G . B. WARBURTON?
Department of Mechanical Engineering, University of Nottingham, Nottingham, U.K.

SUMMARY
In recent papers the author has shown that when determining optimum parameters for an absorber which minimizes the
vibration response of a complex system, the latter may be treated as an equivalent single degree-of-freedom system if its
natural frequencies are well separated. Emphasis was on minimizing the displacement response when the excitation was a
harmonic force. In the present paper simple expressions for optimum absorber parameters are derived for undamped one
degree-of-freedom main systems for harmonic and white noise random excitations with force and frame acceleration as
input and minimization of various response parameters. These expressions can be used to obtain optimum parameters
for absorbers attached to complex systems provided that optimization is with respect to an absolute, rather than a
relative, quantity. The requirement that the natural frequencies should be well separated is investigated numerically for
the different cases. The effect of damping in the main system on optimum absorber parameters is investigated also.

INTRODUCTION
In the classical problem a viscously damped single degree-of-freedom (1DOF)absorber system is attached to
an undamped main system. The optimum absorber parameters (i.e. tuning ratio, or ratio of absorber natural
frequency to that of the main system, and absorber damping ratio), which will minimize the displacement
response of the main mass, when the latter is subjected to a harmonic force, are simple functions of the
assumed mass ratio.' The author2*3has shown that these simple expressions for optimum absorber
parameters and for the optimized response of the main system can be used when the main system is an elastic
body, provided that an effective mass is used for the latter and that the natural frequencies of the body are
well separated. This has been demonstrated for absorbers attached to various beams and plates. However, for
cylindrical shells, where the frequency separation is smaller, optimum absorber parameters diverge from Den
Hartog's values with this divergence increasing as the effective mass ratio (i.e. ratio of absorber mass to
effective mass of the shell) increases.
In order to study the effect of natural frequency distribution, optimum parameters have been determined
for absorbers, which minimize the maximum response of a 2DOF main system: for which the natural
frequency ratio w/wl can be adjusted by varying the ratio of spring stiffnesses. Minimization is applied over
an excitation frequency range, which includes only the first and second resonances of the combined system
(narrow band optimization). As w2/w1increases, the optimum parameters for the 2DOF main system are
asymptotic to the relevant parameter from Den Hartog. For small and practical values of the effective mass
ratio peff(absorber mass/effective mass of 2DOF system), the equivalent l D O F system can be used to predict
optimized maximum response with acceptable accuracy if o2/w12 2; for large values of pelf this becomes
w 2 / o , 3 3 . This is in broad agreement with the results for absorbers attached to elastic bodies.
In the classical problem the main mass is subjected to a harmonic force and optimum absorber parameters
are determined to minimize the displacement response of that mass. Consequently, there has been emphasis
on this excitation and response in subsequent work. However, other types of excitation and response are of

t Professor of Applied Mechanics.

0098-8847/82/030381-21$02.10 Received 16 April 1981
@ 1982 by John Wiley & Sons, Ltd.
382 G . B. WARBURTON

practical importance. The other excitation-time history which is amenable to simple analytical treatment is
random with a white noise spectrum. For this the absorber parameters to minimize some mean square
response quantity are determined. An alternative input is the acceleration of the frame of the main system.
Response quantities of interest include the velocity and acceleration of the main mass. Some of these
alternative problems have been considered for l D O F main systems3* l4 and by Snowdon for beams6 and
'9

plates.'. In this paper the author obtains expressions for the optimum absorber parameters and minimized
response quantity, in terms of the mass ratio p, for various response and excitation parameters; these are of
similar form to the classical expressions.
Real systems have an infinite number of degrees of freedom and are usually modelled by approximate
multi DOF systems, for example, by the finite element method. Thus the collected expressions for absorber
parameters for l D O F main systems are of practical significance only if the concept of replacing real systems
by equivalent l D O F systems, which has been established for harmonic force excitation and displacement
response: applies to the other types of excitation and response. This can be demonstrated for all the cases
considered except one: the relative displacement response to an acceleration of the frame of the main system.
The use of a relative, rather than an absolute, quantity for response leads to the breakdown in the analogy;
this is unfortunate as this case has considerable importance in structural dynamics.
If a real system is represented by an equivalent l D O F system, contributions to response from higher
modes are neglected; the numerical effect of this approximation is investigated for different cases. For
harmonic excitation the equivalent l D O F plus absorber system must give acceptable values for two maxima
of the complex transfer function H(w)for the real system plus absorber. In narrowband optimization other
maxima of H(w) are neglected. For random excitation by white noise all maxima of H(w) contribute to the
mean square response. Thus predictions from the equivalent IDOF system will be reasonable only if
contributions to response from the higher maxima are negligible.
Numerical results suggest that for harmonic excitation the conclusions from earlier work, i.e. for small
values of the mass ratio peff the equivalent l D O F system predicts optimized maximum response with
reasonable accuracy if w 2 / 0 1 > 2, apply to the various combinations of excitation and response parameters.
However, for random excitation more stringent criteria are suggested. Strictly the equivalent l D O F system
replaces an undamped main system; i.e. the analogy breaks down when main system damping is included.
Nevertheless light damping in the main system, which must exist in practice and reduces appreciably the
higher resonances, is of importance when formulating the criteria for equivalence. The effect of damping in a
1D O F main system on optimum absorber parameters is investigated for random excitations; the results
supplement those in the literature for harmonic excitation.

OPTIMUM ABSORBER PARAMETERS FOR VARIOUS EXCITATION AND RESPONSE

QUANTITIES
For an absorber attached to a IDOF undamped main system (Figure 1) there are two forms of excitation
time-history for which simple expressions for optimum parameters can be derived, namely steady-state
harmonic and random with a white noise spectral density. There are several possible combinations of input

Figure 1. Single degree-of-freedom undamped main system (parameters k, and MM)

with attached absorber (parameters k,, M A and c,)
 OPTIMUM ABSORBER PARAMETERS 383

and output, i.e. where and how the excitation is applied and the response quantity of interest. For excitations
we consider a force applied to the mass MM (with the frame fixed) and an acceleration applied to the frame AB
in the X-direction; for harmonic excitation there are two types of force: constant amplitude P eiotand inertial
bo2 eiwt . Response quantities of interest are the displacement, velocity and acceleration of the main mass; the
force transmitted to the frame for force excitation; and the relative displacement of the main mass for frame
acceleration.
For each problem the response at excitation frequency o (or excitation frequency ratio r) can be written
non-dimensionally as H i r ) cia', where HXr) is the complex transfer function or receptance for case number j .
(The case number refers to a particular pair of excitation and response parameters and will be defined in
Table I.) The transfer function
A j + iBj
HAr) = v (1)
C+iD
where A,, B,, C and D are functions of:
the excitation frequency ratio r = w/oM, where w& = k d M M ,
the tuning ratio f = wA/wM, where = kA/MA,
the mass ratio p = MJMM
and the absorber damping ratio yA = c ~ / 2 ( MA)*.
k~
The non-dimensional amplitude of response can be expressed as

The transfer functions for the combined system are easily determined and listed for reference purposes for
the different cases in Table X (Appendix I). (The expressions for A , B j etc. in Table X depends also on yM,
the main system damping factor; numerical results for yM > 0 will be given later, but in this section the main
system is undamped, yM = 0.)
For harmonic excitation the procedure for each case follows the classical method of determining
optimum parameters.' There are two values of r at which the response R j is independent of damping ratio
yA-the invariant points. In general, values of R , at the two invariant points differ. When the tuning ratio
has its optimum value, f,,,,the responses at the invariant points are equal. The two values of yA, y1 and
y2, which make the gradient of the relevant response curve zero at the invariant points, are determined.
Then the optimum damping value is given by

YLopt = M+Y3 (3)

The optimized maximum response R,,, is taken as the common value of the responses at the invariant
points. In Table I expressions for the optimum absorber parameters,f,,, and yA,,,,, and optimized response
of the main mass, Ropl in non-dimensional form, are given in terms of the mass ratio p for nine
combinations of excitation and response parameters. In non-dimensional terms there are only five distinct
cases, as can be seen by inspecting the transfer functions in Table X . For completeness the classical
problem (case l), that for velocity response (case 2), for which Neubert obtainedf,,,, and that for excitation
by an inertial force, for which Sauer and Garlandt4 obtained optimum parameters, are included. For very
small values of p, fop, -+ 1, yA,,,, -+ (3p/8)* and R,,, 4(2/p)*. As p increases, fop, decreases; for a specified
p,foptfor case 3 is largest, followed by cases 2, 1, 5 and 7 in descending order. Values of R,,, for specified p
are in the reverse order; yA,,,, for case 2 is smallest, followed by cases 1, 3, 5 and 7 in ascending order.
Closed form expressions for R,,, and for case 7 have not been found and series expressions of
reasonable accuracy have been given.
The procedure adopted here is approximate, as it assumes that the maximum response occurs at an
invariant point and uses the mean square value of damping, equation (3), as optimum. This was discussed
for the classical problem (case 1) by the author,' who showed that the difference between the true value of
R,,, and that assumed from invariant point theory was insignificant, even for large values of p. It has not
 L.
P
I

5
0
G
P
vi M

G
I
0
0
B m
3
I

0
E I

E
k gm
e
vi
c
.-
vi
e
2
w
G .-
-F
-P
L

9
2
n- P
0
5
-0
w
'8
.-
-+3 - -+3
- 8
2
3
.d
a
I

0" 3
c
C
0
m
c-
9
3
p h Bm
-
m
-+ +3
h-
n
t-I3
P W
0
!2
e
-0
p:
'5
.-

dl :*
I

0" NI
3 O

d v1
 OPTIMUM ABSORBER PARAMETERS 385

been investigated for the other cases of Table I, but is expected to be of little significance, because of the
similar form of the expressions for optima for the different cases.
The determination of optimum absorber parameters for random excitation has been described by the
author3 and applied to cases 1, 5 and 6 (definitions of types of excitation and response are given in Table I).
For white noise excitation of spectral density So the variance of a response quantity x, 02, is given by

as the variance equals the mean square value for a random quantity with zero mean. The complex transfer
function H,(w) is the dimensional form of Hj(r)of equation (1) and Table X.Closed form expressions for G:
from equation (4) are obtained by using integrals from Crandall and Mark.' The optimizing conditions are

Applying these conditions, simple expressions forf,,,, yA,opt and No,, are obtained. ( N is a non-dimensional
form of 0:). These are given in Table I1 for cases 1, 2 and 5. Results for cases 4 and 6 can be inferred from
those for case 1. For the other cases either the transfer function H(r) -+ 1 as r -, co or H(0) -, co and an
infinite response is predicted; this is a consequence of assuming white noise excitation. For p sufficiently
small,fopl 1, yA,opt-,$p* and No,, + p - * . For specified p,foptfor case 2 is largest, followed by cases 1 and
--f

5 respectively. Values of No,, are again in reverse order; yA,opl for case 1 is smallest, followed by cases 5 and
2 respectively. For a particular case and specified p,foptfor random excitation is greater than or equal to fop,
for harmonic excitation (equality exists only for case 5); for small, practical values of p, yA,opl for random
excitation is less than yA,opt for harmonic excitation.
When the main system is damped, optimizing conditions (5) lead to two non-linear equations in f and yA,
which have been investigated numerically. (Results are given in a later section.) These equations are quoted
in Appendix 11.
The numerical significance of the different expressions for optimized parameters in Tables I and I1 will
be illustrated by two examples. Firstly, it is required to optimize the displacement response for a harmonic

Table 11. Optimum parameters for absorbers attached t o undamped lDOF main systems. Random excitation with white
noise spectral density, So. Various excitation and response parameters

Optimized absorber
Excitation Optimized response parameters
~~

Applied Parameter Definition

Case Type to optimized of N NOPI fo,, Y:,opt

u: ki (1 + P m * P(1+3P/4)
1 Force Main XM
mass 2nS0 WM l+P W+P)(l+P/2)

u,?k i 1 P
2 Force Main i M
-
mass 2ns0 a; 4

Notes: u;, ui and u: are the variances of the absolute displacement of the main mass, of its velocity and of its relative displacement
respectively.
Results for cases 4 and 6 (defined in Table I) are as case 1, except that the definitions of N are u5/2nS0 wM and o:/2nS0wM
respectively, where u: and ui are the variances of the force transmitted to the frame and the acceleration of the main mass
respectively.
For cases 3,8 and 9 the transfer function H ( r ) + 1 as r -+ co.For case 7 H(0) + co.Thus for these cases an infinite mean square
response is predicted for white noise excitation.
The results for cases 1 and 5 are quoted from Reference 3.
386 G . B. WARBURTON

frame acceleration (case 7); if the absorber is designed from the optimum parameters for the classical
problem (case l), instead of from the true parameters, the actual maximum response is 43 per cent greater
than that predicted from Rop, for case 1, if p = 01. However, as noted, Rapt for case 7>R,,, for case 1, and
the maximum response with the incorrect absorber parameters exceeds the true optimum by 19 per cent.
Second, the mean square displacement response for a random force excitation has to be minimized, but
optimum absorber parameters from the corresponding harmonic problem (case 1 of Table I) are used
instead of those for case 1 of Table 11. For p = 0.1 the mean square response with incorrect parameters
exceeds the optimum value by 19 per cent; these differences increase relatively rapidly as p increases, and
for the latter case are 4 and 37 per cent for p = 0.01 and 0-2 respectively.

ANALOGIES
In earlier papers2-4 the author has established l D O F analogies for multi D O F systems and simple elastic
bodies, when absorbers are attached to the main system. This is based on the assumption that the response of
the complete system can be formulated in terms of a single mode of the main system. Then Den Hartog’s

Table 111. Use of lDOF analog to replace general multi DOF systems (and elastic bodies). Relations between optimized
parameter for the multi DOF system and optimized response parameters, Rj.optand Nj,optr which are as given in Tables I
and 11, except that perfreplaces p. Expressions for optimum absorber parameters are also as in Tables I and 11, except that
peffreplaces p. Main system damping y M = 0

Relation
Case Harmonic excitation relation Random excitation

9
~~

Notes: perf= MJM,,,, where Me,, = effective mass of main system; kerf = effective stiffness of main system. (See equations (22),(30) and
(31) for definitions.) 0ptimization.is with respect to a response parameter of mass i (cases 1, 2, 3, 6, 7 and 8) or the force
transmitted to the frame (cases 4 and 9) and is with respect to mode r for harmonic excitation. Excitation is applied as a force at
mass F (cases 1,2,3,4,8 and 9) or as a frame acceleration (cases 6 and 7). The frame is directly connected to masses MI,M2,..., M s
through springs of stiffness k,, k,, ..., k, respectively.
For elastic bodies we consider only cases 1, 2, 3 and 8, for which the response and excitation parameters do not include frame
motion or force. If the amplitude of the body at position (x,y) is W(x,y), the-rth mode is given by W , ( x , y )= 4 d x ) $ , ( y ) and
optimization is with respect to a response parameter at position (x, y). W(x, y), W(w,y) w(x, y), and W(x, y) replace Xi, 1, andxi
X, in cases 1, 2, 3 and 8 respectively and I &(x) $Ay) &(xF) $,(yF) I replaces I,zi ,zF I.
 OPTIMUM ABSORBER PARAMETERS 387

simple expressions for optimum absorber parameters and optimized response can be used for the
minimization of response of these multi D O F systems and elastic bodies. However, in this earlier work the
excitation was a harmonic force applied to any mass of the multi D O F system (or to any point in the elastic
body) and optimization was with respect to the displacement response of a specified mass (or point on the
elastic body). Here analogies for other forms of excitation and response parameters are considered, so that
the results of Tables I and I1 can be applied to the minimization of response for multi D O F systems and
elastic bodies. If the dynamic transfer function (or receptance) determined approximately for the multi DOF
system is equivalent to that for the l D O F system, the analogy is established and can be used for harmonic
and random excitations, as the response to both depends on this transfer function, although in different ways.
It requires only a simple extension of the analysis in Reference 4 to demonstrate the existence of the
analogy for cases in which the excitation is a force applied to mass MF. Details will not be given, but the
relations between the optimized response for the multi D O F system and that for the equivalent lDOF
system are listed in Table 111 for harmonic excitation and, when appropriate (see previous section), random
excitation. The minor changes in these expressions if an elastic body replaces the multi D O F system are noted
in the table.
The analysis for a general multi D O F system with attached absorber, when the frame is subjected to an
acceleration X , e'"', is given in Appendix 111. When the absolute displacement of mass Mi, x i , is the response
parameter (case 7), an analogy exists and the relationship is included in Table 111. However, when the relative
displacement of mass Mi,yi (= x i - x , ) , is the response parameter (case 5), there is no simple analogy. The
concept of effective mass of the multi D O F system is essential in the derivation of any analogy; its definition
depends upon equivalence of kinetic energy, which is defined, of course, in terms of absolute velocities. The
use of relative displacement in the analysis of case 5 causes this breakdown in the analogy. This is unfortunate
as the combination of frame acceleration and relative displacement response is of considerable importance in
structural dynamics.
When the excitation is the frame acceleration (cases 6 and 7) or the response is the force transmitted to the
frame (cases 4 and 9), expressions for elastic bodies are not given in Table 111, as these will depend upon the
actual boundary conditions for the problem considered.
For harmonic excitation the relations are in terms of mode r, for which the natural frequency of the main
system is orand the modal vector is rz;i.e. the purpose of the absorber is to minimize the resonant response in
the vicinity of a,.(Most numerical results relate to minimization of the fundamental resonance, but
optimization with respect to higher modes has been considered.2-') For random excitation the relations are
defined in terms of the fundamental mode (defined by natural frequency w1 and modal vector ]z). As will be
shown later, the effectiveness of these relations depends upon the transfer function, which occurs in the
integrand for mean square response, being reduced by the addition of the absorber for frequencies in the
vicinity of w1 and being relatively small in the vicinity of higher natural frequencies. An absorber, which is
effective in the vicinity of a higher resonance, would not decrease the large contribution to mean square
response from frequencies near to wl.

NUMERICAL RESULTS
In previous sections it has been shown that equivalent l D O F systems can be established for all but one of the
nine combinations of excitation and response, which are defined in Table I; thus the simple expressions for
optimum absorber parameters and response (Tables I and 11) can be used, provided that the natural
frequency separation is sufficient to justify the one-mode approximation to response. In previous ~ o r k , ~ - ~
where minimization of the fundamental resonance was required, it was concluded from results for elastic
bodies and for the symmetrical 2DOF system of Figure 2 that for small and practical absorbers w2/w1> 2
was a reasonable criterion; for larger values of the mass ratio peffthis should be amended to a 2 / w 1> 3, where
o1and o2are the fundamental and second natural frequencies of the main system. This criterion was based
on numerical results for case 1 only. It is necessary to investigate how the numerical effect of the
approximation varies for different cases. This will be achieved by considering the symmetrical 2DOF system
388 G. B. WARBURTON

I
+X

(a)
Figure 2. (a) Two degree-of-freedom main system. (b) Absorber system, attached to upper mass of (a)
of Figure 2(a) with the absorber of Figure 2(b) attached to the upper mass. For the 2DOF system the ratio of
natural frequencies is given by
(wz/w1)2 = 1 +26
where 6 is the ratio of the stiffneks of the middle spring to that of the upper (or lower) spring. In the
fundamental mode the amplitude ratio XJX, = 1. Excitations considered are a force applied to the upper
mass and acting in the X-direction and an acceleration in the X-direction of the complete frame AB and CD.
Response quantities of interest are the displacement, velocity and acceleration of the upper mass, the total
force transmitted to the frame, F( = kx, + kx,), and, for the frame acceleration as input, the relative
displacement of the upper mass y , ( = x1 -xo, where xo is the frame displacement).
For harmonic excitation, when an equivalent system is used, absorber tuning and damping ratios,fe, and
yA,eq are given by the simple expressions of Table I; the corresponding maximum response, Re,, is based on
the value at the invariant points. When an absorber defined by feq and is attached to the 2DOF main
system of Figure 2, the maximum response of this system R , is somewhat greater than Re,. Optimum
absorber parameters fop, and yA,opt, which minimize this maximum response Rapt, can be found; in general,
R,> Rapt> Re,. If the difference between R , and Re, is small, the use of the equivalent l D O F system is
justified and the relatively laborious determination of optimum parameters is unnecessary. For force
excitation and displacement response (case 1)of the 2DOF system the percentage difference between R , and
Re, increased as perfincreased and decreased as o,/o, increased; specificallyfor 0 2 / w 1 = 2 and perf= 0.1 the
difference was 8.4 per cent (Table I1 of Reference 4).As this was the lower limit for 0 2 / w 1 and a reasonably
large value of perf,the other cases were tested with the same data. As shown in Table IV, the percentage
differences for the other cases are slightiy smaller than that for case 1. Thus it was concluded that for all the
cases when the excitation is harmonic and small practical values of peffare used, the equivalent lDOF system
predicts optimized response with acceptable accuracy provided that oz/ol> 2.
As shown in Table IV, the equivalent system yields accurate results for case 5, but this is not a typical
result. As noted in Appendix 111, the analogy does not exist in general for case 5, because of the formulation in
terms of relative displacement, but the symmetric 2DOF system is an exception (i.e. conditions (35) and (36)
are both satisfied). For comparison a more general 2DOF system, for which wz/wl = 2 and the stiffnesses in
Figure 2(a) from top to bottom are k, 2.1k and 2k respectively, is considered; the two masses are equal; for the
absorber perf= 0.1. For this 2DOF system the analogy does not hold as it is possible to satisfy only condition
(36). As perfis unchanged, the equivalent system parameters are as given in Table IV. The maximum response
of the 2DOF system exceeds that predicted from the equivalent system by 4.9,4.4and 13.6 per cent for cases
1,4 and 5 respectively. The increased error for case 5 indicates that errors are likely to occur if the concept of
the equivalent l D O F system is misused.
 OPTIMUM ABSORBER PARAMETERS 389

Table IV. Effect of using absorber parameters for equivalent lDOF system on the maximum response of the 2 D O F
system of Figure 2. Harmonic excitation. Narrowband optimization. For 2DOF system 0 2 / w 1 = 2; peff= 0.1

Maximum
Excitation response
~
Response Non-dimen- From equivalent system of 2DOF
Applied quantity system
Case TYPe to optimized fcq 7A.q Rcq Rm
- -

Force Upper Displace-

peiol 0.9091 0.1846 4383 4.967 8.4
mass ment
XI

Force Upper Velocity 0.9315 0.1846 4.369 4.670 6.9

p eiol mass *I

Force Upper Accelera-

p eiot 0.9535 0.1890 4.264 4.520 6.0
mass tion
21

Force Upper Frame

p eior 0.9091 0.1846 4.583 4.762 3.9
mass force F

Accel- Relative
Frame displace- 0.8861 0.1894 4.919 5.093 3.5
Xeration
0e
io1 ment
Yl

Accel- Absolute
Frame displace- 0.8567 0.2060 5.481 5.606 2.3
eration ment
X, eio'
XI
Force Upper Frame
bw2 eiwr 0.9535 0.1890 4.264 4518 6.0
mass force F

Notes: kerf = 2k; w: = k/M (Figure 2).

f e q , yA.eq and Re, are obtained by putting perf= 0.1 in appropriate expressions in Table I.

R , is the non-dimensional maximum response of the upper mass, when an absorber with parameters defined by f e , and is
attached to this mass.
Numerical results for cases 6 and 8 are identical to those for cases 4 and 3 respectively.

For harmonic excitation the equivalent l D O F plus absorber system must give acceptable values for two
maxima of the complex transfer function H(w)for the real system plus absorber. In narrowband optimization
other maxima of H ( o ) are neglected. For random excitation by white noise all maxima of H(o)contribute to
the mean square response. Thus predictions from the equivalent l D O F system will be reasonable only if
contributions to response from the higher maxima are negligible. Numerical results3 for the optimization of
displacement response of beams and cylindrical shells which are subjected to force excitation suggest that
w2/w1> 2 is the required criterion for results which are based on the equivalent system to be accurate.
However, those results were for elastic bodies with small hysteretic damping. Damping, which must exist in
practice, reduces significantly the contributions from higher resonances to the mean square response. As no
comparative results for undamped main systems were included in Reference 3, this point was not emphasised
sufficiently.
For random problems N , and N , , are the non-dimensional mean square responses, using f,,and yA,,, from
the equivalent system to define the absorber parameters and H(w) for the actual and equivalent systems
respectiveiy in the integrand for mean square response. Tables V to VIII give the percentage differences
between N , and Neqfor the 2DOF main system of Figure 2 for cases 1, 2, 4 and 5 respectively for various
values of perfand main system damping ratio yM. The optimum parameters,f,,, and yA,opI, for the lDOF
system become the equivalent parameters, f,, and yA,,,, when applied to a 2DOF system and are tabulated
390 G. B. WARBURTON

also. (The transfer functions for the l D O F and 2DOF systems are given in Table X and XI respectively;
closed-form expressions for the integrals, equation (4),which yield Neqand N,, are obtained from formulae in
References 9 and 10 respectively; the former integrals, which are of greater interest, are quoted in Appendix
11.) The level of agreement between N , and Neqdepends upon the case studied; for force excitation good
agreement occurs if (a) wz/wl > 3, the main system is lightly damped and displacement response is optimized
(case 1); (b) 02/01 > 9, the main system is lightly damped and velocity response is optimized (case 2); and (c)
w2/w1 > 2 and the force transmitted to the frame is optimized (case 4). When excitation is caused by a frame
acceleration and relative displacement of the upper mass is optimized, good agreement occurs for w2/w1> 2.
The presence of light damping reduces appreciably the third resonance in H ( o ) for the system. When yM = 0,
the percentage difference between N , and Neqis large for cases 1 and 2, particularly the latter. However, this
difference is small for cases 4 and 5, but this is a consequence of the particular 2DOF system investigated. The
transfer functions HI@) for the different cases are listed in Table XI. The higher resonance, which tends to
cause differences in N , and Neq,occurs in the vicinity of the second natural frequency w2, or expressed non-
dimensionally when r2 2: 1 + 26. For this value of r and yM = 0 the numerator of the transfer function is zero
for case 4 and the numerator is small when peffis small for case 5. Thus for case 4 the integrand, which
contributes to N , has a zero at r2 = 1 + 26, while the integrand corresponding to N , , is a smooth curve in this
frequency range; this leads to the negative percentage differences which appear in Table VII.
Table V. Optimum parameters for absorbers attached to damped lDOF
systems; also variance of response for 2DOF system of Figure 2 when an
absorber with parameters defined by the equivalent lDOF system is attached.
Case 1. Random force with white noise spectral density So applied to main
mass. Response: variance of displacement of main (or upper mass, ur. For
I
IDOF system N = u,’ k32nS0 w,; for 2DOF system N = u, kerr2/2nS,w 1

Main system
2DOF
02/w1 = 2 3 9
Main
system lDOF
damping ($-l)xlOO%
Peff YM fom YA,OD~ No,,

0.0 1 0 09926 0.0498 1 9.988 733 447 61.2

0.0 1 0.9921 0.0498 1 7.632 20.8 4.17 0.07
0.02 0.99 16 0.0498 1 6,118 13.5 2.74 0.05
0.05 09901 0.0498 1 3.748 10.9 2.41 0.08
0.1 09877 0.0498 1 2.222 15.0 3.72 0.21
0.03 0 09781 008565 5.752 258 155 209
0.01 09773 0.08565 4-893 30.9 6.54 0.11
0.02 09765 0.08565 4.239 19.5 4.07 0.08
0.05 0.9741 0.08565 2.985 13.8 3.02 0.11
0.1 0.9700 0.08565 1.959 16.7 4.13 0.24
0.1 0 0.9315 0.1525 3.126 96.0 53.4 6.93
0.0 1 09302 0.1525 2.856 40.8 10.5 0.24
0.02 09289 0.1525 2.624 28.6 6.73 0.19
0.05 09250 0.1525 2.098 19.8 4.60 019
0-1 0 9 186 0.1525 1554 20.7 5-24 0.30
0.2 0 08740 0.2087 2.1 89 62.5 32.2 4.01
0.01 08724 0.2087 2.053 41.3 12.6 0.38
0.02 08708 0.2087 1.932 32.7 8.84 0.30
0.05 08662 0.2087 1.635 24.5 6.20 0.28
0.1 08586 0.2087 1.291 24.4 6.47 0.39
Note:&, yA,op, and Nopt.areoptimized absorber parameters and system response for a
damped lDOF main system. Iffop, and yA.optare used to define an absorber which is
attached to the damped 2DOF system, the variance of response, expressed non-
for the lDOF system) is the variance predicted
dimensionally, is N,; Neq( = NOPI
from the equivalent system
 OPTIMUM ABSORBER PARAMETERS 39 1

Table VI. Optimum parameters for absorbers attached to damped l D O F

systems; also variance of response for 2DOF system of Figure 2 when an
absorber with parameters defined by the equivalent l D O F system is attached.
Case 2. Random force with white noise spectral density So applied to main
mass. Response: variance of velocity of main (or upper), mass, 02. For l D O F
system N = a:k32nSO a$;for 2DOF system N = O: kZrr/2nSow:
Main system
2DOF
w2/a, = 2 3 9
Main

Perf
system
damping
YM fopt
1D O F
opt No,,
(2 - 1) x 100%

0.0 1 0 0.9950 0.05 9.950 2896 4020 4936

001 0.9956 005008 7.604 82.0 36.6 4.06
0.02 09964 0.05018 6,096 52.3 23.1 2.56
0.05 1~oooo 005056 3.736 37.7 16.1 1.76
01 1.0102 0.05153 2.215 41.2 16.8 1.78
0.03 0 0.9853 008660 5.689 1024 1383 1677
0.01 09863 0.08684 4.840 122 56.4 6.38
002 0.9874 008710 4194 748 33.3 3.71
0.05 0.9921 0.08805 2.954 47.6 20.3 2.21
01 1.0040 0.09022 1.940 46.7 19.0 2.01
01 0 0.9535 0.1581 3.015 366 461 539
0.01 09550 0.1588 2.755 157 87.3 11.1
0.02 09568 0.1596 2.532 107 52.4 6.12
0.05 0.9633 01623 2.026 61.3 29.4 3.23
0.1 0.9783 0.1680 1.501 59.3 24.4 2.51
0.2 0 0.9 1 29 0.2236 2.041 224 262 295
0.01 0,9148 0.2250 1.915 149 100 15.5
0.02 0.9171 0.2265 1.802 116 65.5 8.54
005 0.9250 0.2313 1.526 80.3 37-8 4.32
0.1 0.9424 0.2412 1.206 69.6 301 3.23
Note: See note on Table V.

Tables V to VIII show also the optimized absorber parameters and optimum value of mean square
response for a damped lDOF main system with attached absorber. The values for yM = 0 are obtained from
expressions in Table 11. For yM >0 the optimizing conditions ( 5 ) lead to a pair of non-linear equations infand
yA. (These equations, together with the closed-form expressions for the integrals Nj, are listed in Appendix 11.)
Numerical solutions for optimum values have been obtained iteratively. (The results, which show the effect of
yM on optimized parameters, supplement those for harmonic excitation for case 1.2*11- I 3 * l 5 )&,, decreases as yM
increases, except for case 2, but the dependence on yM is greatest for case 5. is insensitive to changes in
yM, except for case 2 where it increases as yM increases. R,,, decreases as yM increases; this is expected, as the
mean square response with no absorber decreases rapidly as yM is increased, being 25.0, 125,5*0and 2.5 for
yM = 001,002, 0.05 and 0.1 respectively for cases 1, 2 and 5. (Values for case 4 are slightly higher.) Thus for
peffsmall and yM relatively large the addition of the optimun absorber produces only a small decrease in the
mean square response.
Tables V to VIII do not include optimized parameters for absorbers attached to the 2DOF main system.
It may be thought that if these optimum values are determined, the corresponding optimized mean square
response, Nopt,will be significantly smaller than the mean square response, N,, which occurs if the equivalent
system is used to determine the absorber parameters. In Table IX the optimum absorber parameters,&,, and
and the optimized mean square response No,, are given for case 1 for some values of perf,yM and w2/01.
The optimum value of the tuning ratio, Table IX, does not differ greatly from the value based on the
equivalent system, Table V; the differences are largest for high perfand yM and 0 2 / w 1 = 2. Large differences
392 G . B. WARBURTON

Table VII. Optimum parameters for absorbers attached to damped lDOF

systems; also variance of response for 2DOF system of Figure 2 when an
absorber with parameters defined by the equivalent lDOF system is attached.
Case 4. Random force with white noise spectral density So aqplied to main
mass. Response: variance of force transmitted to frame, up. For lDOF
system N = ag/2/2nS00,;for 2DOF system N = ag/2/2nSow 1
Main system
2DOF
w,/w, = 2 3 9

Perf
system
damping
YM fop,
lDOF

YA.opt
(2 - 1) x 100%

0.01 0 0.9926 00498 1 9.988 - 056 0.12 0

001 0,9921 00498 1 7.635 -0.02 0.02 0
002 0.9916 00498 1 6.128 -0.12 -0.01 0
005 0.9902 004982 3786 -0.28 -0.07 0
01 0.9885 004990 2.310 -0.41 -0.12 -0.01
0.03 0 0.9781 008565 5.752 0.29 0.13 0-01
0 01 0.9773 008565 4.895 0.09 0.09 0.01
002 0.9765 008565 4.246 -0.09 0.04 0.01
005 0.9742 008568 3.015 - 0.50 -0.09 -0.01
01 0.9713 008587 2.038 -0.84 -0.22 -0.02
0.1 0 0,9315 01525 3.126 1.06 0.47 0.06
001 0.9302 01525 2.857 057 035 0.04
002 09289 01525 2.628 0.20 025 0.04
005 09253 0 1527 2.1 19 - 0.64 -0.01 0.02
01 0.9206 01532 1.615 - 1.46 -0.33 -0.02
0.2 0 08740 02087 2.189 1.82 0.81 0.09
0 01 0.8724 02087 2.054 1.17 0.64 0.08
002 0.8709 0.2087 1.935 0.64 0.50 0.07
005 0.8667 0,2090 1.651 -0.57 0.13 0.04
01 0.861 1 02101 1.341 - 1.81 -0.34 -0.01
Note: See note on Table V.

between Y ~and ,Y ~~ occur ~ ~ ~yM = 0. (Numerical problems prevented the determination of optimum
,~ when
values for yM = 0, peff= 001 and oz/wl < 3.) The percentage difference between the mean square response
values, N , and Nap,, is given in Table IX. For comparison the percentage difference between N , and Neqfrom
Table V is included also. It was noted earlier that the equivalent system approach was inaccurate when the
latter difference was large. That could imply that significantly smaller mean square responses would be
obtained if the absorber parameters were optimum values rather than based on equivalent values. The
tabulated differences between N,,, and No,, show that this is not true. (The differences are less than 1 per cent,
except for the idealized case of yM = 0.) These results suggest that there is no significant advantage in
determining optimum absorber parameters, and use of parameters based on the equivalent system is
satisfactory, but for low values of wz/ol the true mean square response of the 2DOF system, N,, should be
determined as its estimate from the equivalent system Neqmay be significantly low. The determination of
optimum absorber parameters for the other cases was not considered worthwhile.

CONCLUSIONS
Simple expressions for optimum absorber parameters and system response have been given for an absorber
which is attached to an undamped l D O F main system. The two simple excitation time histories of steady-
state harmonic and random with a white noise spectral density are considered. (Although the derived
expressions relate to white noise excitation, they could be used for excitations with a constant spectral density
up to a cut-off frequency, provided that the latter is large compared to the natural frequency of the main
 OPTIMUM ABSORBER PARAMETERS 393

Table VIII. Optimum parameters for absorbers attached to damped lDOF

systems; also variance of response for 2DOF system of Figure 2 when an
absorber with parameters defined by the equivalent lDOF system is attached.
Case 5. Random acceleration with white noise spectral density So applied to
frame. Response: variance of relative displacement of main for upper) mass, 0;.
For l D O F system N = cr," u ~ ~ K S for
, ; 2DOF system N = G: w:/2nS0
Main system
2DOF
WJW1 =2 3 9
Main
system lDOF
damping $(+"%
Perf YIll hot yA.0~1

0.0 1 0 09876 004981 10138 0.84 007 0.16

0.01 0.9850 0.04981 7.743 0.40 016 002
0.02 0.9819 0.04981 6.205 0.27 0.15 0.03
0.05 0.9704 0.04982 3.798 0.21 0.15 0.05
0.1 0.9436 0.04982 2.249 0.20 0.18 0.11
0.03 0 0.9636 0.08566 6.058 0.52 -0.23 -0.67
0.01 0.9592 0.08566 5.110 1.16 0.51 008
0.02 0.9545 0.08566 4.424 1.08 051 011
0.05 09380 0.08567 3.109 0.95 0.56 021
0.1 0.9032 0.08569 2.036 1.00 0.73 043
01 0 08861 01527 3.602 3.87 1.56 016
0.01 0.8789 01527 3.285 3.84 1.69 0.31
0.02 0.8714 0.1528 3.014 3.85 1.81 0.46
0.05 08468 01529 2.399 4.05 2.27 0.98
01 07991 0.1531 1.765 4.72 3.17 1.93
0.2 0 07906 0.2097 2.865 6.81 2.76 0.29
0.01 07815 0.2098 2.680 7.09 3.15 0.67
0.02 0.7721 02099 2.516 7.42 3.56 1.08
0.05 0.7421 0.2103 2.113 8.53 4.86 2.35
01 06862 0,2112 1.649 10.6 7.17 463
Note: See note on Table V.

system.) Appropriate combinations of a force applied to the main mass or an acceleration imposed on the
frame for excitation parameter with an optimized response parameter, which can be the displacement,
velocity or acceleration of the main mass or the force transmitted to the frame, are studied. These results
complement those for the classical problem, where force excitation and displacement response are the
parameters. Although the pattern of behaviour for the various cases is similar, the expressions differ.
Provided that the response of multi D O F system (or elastic body) plus absorber can be represented with
acceptable accuracy by the contribution from a single mode, the multi DOF system (or elastic body) can be
replaced by an equivalent l D O F system for the purpose of determining optimum absorber parameters. This
was established p r e v i o u ~ l yfor
~ * ~force excitation and displacement response. Provided that the response
parameter is an absolute quantity (i.e. excluding case 5 where relative displacement is considered), an
equivalent system can be derived for any combination of excitation and response parameters. In principle the
equivalence exists for harmonic and random excitations, as the response to both excitations depends upon the
same complex transfer function. In practice, for harmonic excitation the approximate transfer function has to
model the true function only in the vicinity of the relevant resonance, while for random excitation the
approximate model has to apply to the full frequency range. For minimization of the fundamental resonance
by a practical absorber of small mass ratio satisfaction of the natural frequency condition wz/wl > 2 ensures
that the equivalent system yields optimum absorber parameters and minimized response of acceptable
accuracy for all cases for which the analogy holds. Strictly the analogy exists only when the main system is
undamped. The effect of small damping in a main system with l D O F on optimum absorber parameters is very
small; this is shown for random excitation with different response and excitation parameters and was shown
394 G . B. WARBURTON

Table IX. Optimum parameters for absorbers attached to damped 2DOF system (Fig. 2). Comparison
of optimum response Noptwith response when absorber parameters are obtained from equivalent
lDOF system, N , (Table V). Excitation :random force with white noise yectral density So applied to
upper mass. Response: variance of displacement of upper mass, ux,N = u: k&/2nS0 o1

0.01 0 9 0.99 1 0.0768 14.85 8.4 61.2

0.01 0.01 2 0.989 0.0488 9.21 0.10 20.8
3 0.990 0,0498 7.95 0.0 1 4.17
9 0.992 0.0498 7.64 0 0.07
0.01 0.05 2 0.978 0.0478 4.15 0.15 10.9
3 0.986 0.0488 3.84 0.08 2.41
9 0490 0.0498 3.75 0 0-08
0.936 0.243 5.52 11.1 96.0
0.931 0.214 4.53 5% 53.4
9 0.931 0.162 3.24 0.20 6.93
0.1 0.01 2 0.916 0.173 3.99 0-70 40.8
3 0.923 0.154 3.15 0.08 10.5
9 0.930 0.152 2.86 0 0.24
0.1 0.05 2 0.897 0.146 2.49 0.8 1 19.8
3 0.9 13 0.149 2.19 0.15 4.60
9 0.923 0.152 2.10 0.0 1 0.19

earlier2 for harmonic force excitation and displacement response. Thus for real systems with light damping,
which satisfy the above frequency condition, it is reasonable to use optimum absorber parameters from
Table I to minimize harmonic response.
For random excitation the mean square response may be seriously underestimated if the equivalent system
is used for a hypothetical undamped main system. This discrepancy is caused by contributions to mean
square response from the higher resonances. For real systems with light damping the mean square response is
predicted with reasonable accuracy from the equivalent system, if (a) 02/01 > 3 for displacement response
and force excitation and (b) wz/wl > 9 for velocity response and force excitation. In these circumstances the
damping causes the contribution to response associated with higher resonances to be insignificant. No
general conclusion is drawn from the results for other cases (Tables VII and VIII), as they are affected
significantly by the symmetry of the 2DOF model.
From results in this and previous paper^^-^ a design procedure can be outlined. It is required to minimize
the response in the vicinity of the fundamental resonance of a multi D O F system, which has a fundamental
natural frequency o1and associated normalized modal vector, lz.The location of the absorber (selection of
integer a) and size of the absorber mass MA are chosen from practical considerations. (In general, optimized
response decreases as perfincreases, so the largest allowable value of MA should be used.) Using 1z and
equations (22) and (30), the effective mass Me,, is determined. Knowing perf(= M A / M e f f )the , optimized
absorber parameters,f,,, and yA,opt,and response parameter Roptare obtained from expressions in Table I for
the relevant combination of response and excitation. The actual optimized response, for example
displacement of mass Mi, Xi,or frame force F , is then found from the appropriate relation in Table 111; if
required, the effective stiffness kerf is given by equation (31). From f = wA/w1, wi = k A / M A and
y A = c A / 2 M A w A ,absorber stiffness and damping, k , and cA, are obtained. If w2/01> 2 and perf<0-1 or
02/w1> 3 for perflarge, this procedure predicts optimized absorber constants and minimized system response
of adequate accuracy. If the inequalities are not satisfied, the method supplies useful starting values for an
iterative determination of the optimum parameters. It has been assumed that absorber parameters, which are
derived for an undamped main system, can be used satisfactorily for a real system with light damping.
(Application of absorbers to systems with moderate to heavy damping is of infrequent occurrence.) For
random excitation the procedure is similar, except that expressions in Table I1 are used for optimum
 OPTIMUM ABSORBER PARAMETERS 395

Table X. Complex transfer functions HAr) for various excitation and response parameters for damped
lDOF system with attached absorber
A.+iBj
HAr) = 2
-
C+iD
and
Rj = I ffjr)I
C = ( f -?)(I - r ’ ) - p f r’ -4yA yMfr’
D = 2yArf( 1 -r’ -pi-’) + 2yMr ( f 2- r 2 )

Case Response
.i Excitation parameter Ri Ai Bi

peiwt
XM f2-r’ 2YA rf

peiot
*M -2YA r2 f r(f - r2)

peiot
3, ’
- r’( f - r’) -2yAr3f

peiwt
Frame
force F f -r2 -47, yMfr2 27, rf+ 27, r ( f ’ - r 2 )

XI3 eior YM f 2(1 + p ) - r 2 2YA rf(l +

x0 eiwr
XM As case 4

x 0e iwr
XM
- A4 - B4
r2 r’
Force
b w 2 eiwr XM -A3 -B3

Force
bW2 eiwr
F r’ A, r2 8,

parameters, the frequency inequalities are more restrictive and consequently the iterative procedure is more
likely to be invoked.
APPENDIX I.
Transfer functions
Table X gives the non-dimensional complex transfer functions HAr) for the various combinations of
response and excitation parameters for the lDOF main system with attached absorber of Figure 1. However,
in the tabulated functions viscous damping in the main system is included, i.e. there is a dashpot of constant
cMin parallel with the spring of stiffness k,; the non-dimensional damping ratio y, = 4 2 ( k , M,)*.
Table XI gives the non-dimensional complex transfer functions HAr) for the 2DOF main system with
attached absorber of Figure 2. Proportional viscous damping in the main system has been added; i.e.
dashpots with constants c, 6c and c are assumed to be in parallel with the springs of stiffness k, 6k and k
respectively. Then 7, = c/2(kM)* and equals the modal damping for the fundamental mode of the 2DOF
system without the absorber.
APPENDIX I1
Optimizing conditions for random response
For an absorber attached to a damped lDOF main system, which is subject to white noise excitation by
396 G. B. WARBURTON

Table XI. Complex transfer function Hjr) for various excitation and response parameters for damped 2DOF system with
attached absorber (Figure 2). Case number j defined in Table I
a j iPj
H,{r) = 7
+
K+lL

K + 26) -r2[1 + 26 +f2(2 +p)(1+6) +4yi f Z ( l+ 26)+ 8yA yMf(1+ 241

= f2( 1
+ r4[f2(i + p ) + 2(1 +6)+4yi( 1+ 26)+4yAyMf(1+ 6)(2 +p)] -r6
L = 2yArf[1 +26-r2(2+p)(1 +6)+r4(1 +p)]+2yMr[2(1 +26)f2-r2{f2(2+p)(1 +6)
+ 2(1+ 26)) + 2(1+6)r4] - 8 y 7;~ fr3(1+ 26)

+
(1 6 -r2) (f2 -r2) -47, yMf(1 + 6) r2 + + +
2yAfr( 1 6 -r2) 2yMr( 1 6) ( f 2 - r2)

2 kXJPw, -r81 r.a 1

3 kX,/Pw: -r2 a1 -r2 81

4 FIP (1 + 26 -r2)(f2-r2)-4yA yM fr2 2yAfil +26-r2)+2yMr(f2-~2)

x [2(1 +26)-r2] x [2(1 +26)-r2]-8)’~y$fr3(1 +26)
-4y$r2(1 +26)(f2-r2)

5 a: YlIXO -[(I +26 -r2)(f2 -rZ)-4yA yM f r 2 + +

- 2yAfr[l+ 26 - r2 p( 1 6 - r2)]
x {(1+26)+p(l+6)}+pf2(1+6-r2)] -2yMr[(1 +26)(fZ-r2)+(1 + 6 ) p f 2 ]

m4

-a&’

--a3( = r2 al)

r2 u4

Notes: For yM = 0: (i) the expressions are compatible with those for a general undamped multi DOF system, noting that for this 2DOF
system kerf = 2k, Me,, = 2M, perf= p / 2 and l z I = Iz2 = 1. (ii) Also, as 6 + 00, [(w2/w1)’ = 1+ 263, the expressions reduce to
those appropriate to an undamped lDOF system (Table X). (This applies also to case 5 for this special 2DOF system.)

either a force applied to the main mass or an acceleration applied to the frame, the non-dimensional variance
of the response Nj can be expressed as

wherej is the case number as defined in Tables I and 11; the relation between Nj and the corresponding
variance o2 is also given in Table 11, and HJ{r) is the non-dimensional complex transfer function, defined in
Table X. Using integrals given by Crandall and
OPTIMUM ABSORBER PARAMETERS 397
398 G . B. WARBURTON

APPENDIX 111
Existence of analogies
For the main system we consider an undamped general multi DOF system; its free vibrations can be
represented by the matrix equation
MX+Kx = 0 (20)
where K and M are the symmetric stiffness and mass matrices, and x is the vector of displacements.
The analogy between the steady-state response of this system, when a harmonic force is applied to one
mass of the system, and that of a single DOF system, when both systems have an attached absorber and
response is approximated by the single relevant mode of the system without the absorber, has been
established? In this appendix we follow similar analysis in order to establish the analogy when the excitation
is caused by motion of the frame of the system.
We assume that the frame has an imposed displacement XOeiwrand that it is connected to masses
MI, M 2 ,... ,M,, S < n, which have displacements xl, x2,... ,x, respectively, through springs of stiffness
k,,k 2 , ...,k, respectively. All displacements and the lines of action of the spring forces are parallel to the X -
axis. The equations of motion for forced vibration of the multi DOF system are
MX + Kx = pXo eiwt (21)
where
ps = k,, s < S
0, s>s
If ,z is the vector of normalized amplitudes for the rth mode, the normalizing condition is
,zTM,z = m, (22)
where m, is the rth modal mass. Also
,zTK ,z = m, o,‘ (23)
where oris the natural frequency of the rth mode. We assume that the response can be represented by the
single contribution associated with the rth mode, i.e.
X =rzqr (24)
 OPTIMUM ABSORBER PARAMETERS 399

Substituting equation (24) in equation (21), pre-multiplying by ,zT and using equations (22) and (23),
S
rn, q, + rn, of qr = ,zT pXo eiwr= ( J ,k,) X , eior
s= 1

If the absorber system, which is defined in Figure 2(b), is attached to mass Ma with the absorber
displacement xA being in the same direction as x,, the equation of motion for the absorber is
M A XA + CA k~ + k A XA -CA k, - k A X, = 0 (26)
The addition of the absorber leads to the additional terms on the left-hand side of the ath equation of the set
(21):
CA k, + k A x, -CA i~- k A XA (27)
Putting k, = iox,, fa= - o2x, etc. and using equation (26) to eliminate xA from terms (27), the modified
form of equation (25) is

If we wish to minimize the displacement response of mass Mi,Xi, and the input is specified as the frame
acceleration Xo eiot,i.e. we are considering case 7 as defined in Table 111, then from equations (24) and (28)

where
F(O) = [mr(Of -0')f,Z,Z(k, + iOCA)] (kA- MA O2+iwCA)-,z,Z(k~ +iWC.J2
The effective mass of the main system Me,, is defined to give the same kinetic energy, when placed at the
coordinate to which the absorber is attached, as that of the main system, when both are vibrating in the rth
mode, i.e.
+Me,,1: = +XT MX
Using equations (22) and (24)
Me,frz,2 2 mr (30)
The effective stiffness kerf is defined so that the strain energy of the main system in mode r is +kerfqf. From
equations (23) and (24)
kefr = mr of (31)
Using the non-dimensional factors peff= MAIMeff,r = o/wr,f = wA/or,where
mi= kA/MA, 1/A = cA/2(kA MA)f

and equations (30) and (31), equation (29) can be written

S

o2X . -
I rZi s =
-
1 ("' ks) ( A , + iB,) eiwt

x, keff C+iD
where
1 .
A 7 -- - - ( f 2 - - r Z ) , B 7 = -2y~f/r
rz
c = (f2-?)(1 -'2)-pefffZr2

D = 2yA rf (1 - r 2 - peffr 2 )
400 G . B. WARBURTON

Inspection shows that A,, B,, C and D agree with expressions for case 7 for the l D O F system in Table X,
noting that p in the table is replaced by perfand for an undamped main system yM = 0. This establishes the
analogy. Optimum absorber parameters and response for case 7 are given in Table I for the equivalent lDOF
S
system and can be used with the multiplying factor, i.e. (*zSkS),zi/kefr,(which is listed in Table HI), to
s= 1
determine the optimized response of the multi D O F system.
Considering case 5, the relative displacement response of mass Mi, yi = x i - xo, has to be minimized for the
same frame acceleration xo eiW'.For the multi D O F system without the absorber equation (2) is replaced by
My + Ky
- rnxoeiwf
= (33)
where mT = [ M , M2--- M,]. Following analysis similar to that for case 7, equation (29) is replaced by

The definition of effective mass Me,, depends on equivalence of kinetic energies, which are defined in terms
of absolute velocities ii,rather than the relative velocities jiof this analysis. If this problem is neglected,
acceptance of definition (30) for Meffensures that the non-dimensional equivalent of F ( o ) in equation (34) is
(C + iD), as defined in equation (32). For the analogy to hold it is necessary that the non-dimensional form of
the square bracket in equation (34) should be proportional to ( A , +iB,), where A , and B , are defined in
Table X. This occurs only if
rZaMeff = rzs~s (35)
s= 1

However, Me,, has been defined by equation (30), which can be written for comparison

Equations (35) and (36) are not compatible definitions for general systems. (The 2DOF system of Figure 2
is a special case, for which both equations can be satisfied.) Thus for case 5, unlike all the others considered in
Table I, it is impossible to establish an analogy; it is noted that only in case 5 is the response formulated in
terms of a relative quantity.

APPENDIX IV
Notation
functions off, r, yA, y~ and p
constant (amplitude of inertial force = b o 2 )
viscous damping coefficient
tuning ratio
force transmitted to frame
complex transfer function
stiffness
effective stiffness of main system
stiffness matrix
rth modal mass
mass
mass matrix
number of degrees of freedom in general system
non-dimensional form of 0'
vector of excitation forces
amplitude of excitation force
rth generalized coordinate
ratio of excitation frequency to natural frequency
 OPTIMUM ABSORBER PARAMETERS 40 1

non-dimensional response amplitude

constant spectral density of excitation
time
displacement
velocity
acceleration
vector of displacements
amplitude of displacement
displacement relative to frame
normalized amplitude at D O F s in rth mode
vector of normalized amplitudes for rth mode
damping ratio
ratio of stiffnesses (Figure 2)
ratio of absorber mass to main mass
ratio of absorber mass to effective mass of main system
variance of response quantity x
excitation frequency
natural frequency of absorber system
natural frequency of main system
natural frequency of rth mode

degree of freedom to which absorber is attached

absorber system parameter
value using equivalent system
degree of freedom to which force is applied
degree of freedom at which response is evaluated
case number
maximum value
main system parameter
frame parameter
optimized value
mode number
integer
REFERENCES
1. J. P. Den Hartog, Mechanical Vibrations, 4th edn, McGraw-Hill, New York, 1956.
2. G. B. Warburton and E. 0. Ayorinde, ‘Optimum absorber parameters for simple systems’, Earthquake eng. struct. dyn. 8, 197-217
(1980).
3. E. 0. Ayorinde and G . B. Warburton, ‘Minimizing structural vibrations with absorbers’, Earthquake eng. struct. dyn. 8, 219-236
( 1980).
4. G. B. Warburton, ‘Optimum absorber parameters for minimizing vibration response’, Earthquake eng. struct. dyn. 9,251-262 (1981).
5. V. H. Neubert, ‘Dynamic absorbers applied to a bar that has solid damping’, J. acoust. soc. Am. 37, 673-680 (1964).
6. J. C. Snowdon, Vibration and Shock in Damped Mechanical Systems, Wiley, New York, 1968.
7. J. C. Snowdon, ‘Vibration of simply supported rectangular and square plates to which lumped masses and dynamic vibration
absorbers are attached, J. acoust. SOC. Am. 57, 646-654 (1975).
8. J. C. Snowdon, ‘Mechanical four-pole parameters and their application’, J . sound uib. 15, 307-323 (1971).
9. S. H. Crandall and W. D. Mark, Random Vibration in Mechanical Systems, Academic Press, New York, 1963.
10. H. M. James, N. B. Nichols and R. S. Phillips (Editors), Theory ofSeroornechanisrns, McGraw-Hill, New York, 1947.
11. T. Ioi and K. Ikeda, ‘On the dynamic damped absorber of the vibration system’, Bull, J S M E 21, 6 4 7 1 (1978).
12. S. E. Randall, D. M. Halsted and D. L. Taylor, ‘Optimum vibration absorbers for linear damped systems’, J. rnech. design A S M E 103,
908-913 (1981).
13. V. A. BaDat and H. V. Kumaraswamv, ‘Effect of primary damping on the tuning- conditions of a dynamic vibration absorber’, Proc.
5th world cong. theory of machines and mechanisms, 1 , 329-332 (1979).
14. F. M. Sauer and C. F. Garland, ‘Performance of the viscously damped vibration absorber applied to systems having frequency
squared excitation’, J. appl. mech. A S M E 16, 109-1 17 (1949).
15. A. G. Thompson, ‘Optimum tuning and damping of a dynamic vibration absorber applied to a force excited and damped primary
system,’ J . sound uib. 77, 403415 (1981).