Você está na página 1de 7

Osamu Nishihara

Associate Professor,
Department of Systems Science,
Graduate School of Informatics,
Kyoto University,
Yoshida-Honmachi, Sakyo-ku,
Kyoto 606-8501, Japan
e-mail: nishihara@i.kyoto-u.ac.jp

Closed-Form Solutions to the


Exact Optimizations of Dynamic
Vibration Absorbers
(Minimizations of the Maximum
Amplitude Magnification Factors)

Toshihiko Asami
Associate Professor,
Department of Mechanical Engineering,
Himeji Institute of Technology,
Shosha, Himeji, Hyogo 671-2201, Japan
e-mail: asami@mech.eng.himeji-tech.ac.jp

A typical design problem for which the fixed-points method was originally developed is
that of minimizing the maximum amplitude magnification factor of a primary system by
using a dynamic vibration absorber. This is an example of usual cases for which their
exact solutions are not obtained by the well-known heuristic approach. In this paper, more
natural formulation of this problem is studied, and algebraic closed-form exact solutions
to both the optimum tuning ratio and the optimum damping coefficient for this classic
problem are derived under assumption of undamped primary system. It is also proven that
the minimum amplitude magnification factor, resonance and anti-resonance frequencies
are entirely algebraic. DOI: 10.1115/1.1500335

Introduction

Well-known design formulas for the tuning of dynamic vibration absorbers DVAs substantially originate from a note by J. E.
Brock, published in 1946 1. The design criterion considers the
minimization of the resonance amplitude magnification factor.
The concept was introduced in a paper by J. Ormondroyd and J. P.
Den Hartog 2, in which the existence of optimum damping was
numerically observed. Hahnkamm studied the optimal tuning condition, which minimizes the amplitude magnification factor at
fixed points. This is considered to be the origin of the fixed-points
method. J. E. Brock completed these works by calculating the
average of two optimum damping factors for the fixed points.
Though this is an empirical method, the formulas were presented
in Mechanical Vibrations written by J. P. Den Hartog 3. Thus,
the results have widely circulated in this field 4.
The present paper discusses a more natural formulation of the
design problem, which results in the exact solution to this classic
problem. This algebraic approach gives the closed-form solution,
which is obviously quite different from numerical optimizations
57. The fixed-points method tunes a DVA such that the magnification factors of two fixed points are equalized. If the damping
ratios that make the resonance curves horizontally pass through
the fixed points P and Q are identical, then the optimal solution is
produced. In this case, however, they are not equal, and the resonance points do not coincide simultaneously with the corresponding fixed points. The readjustment of the damping ratio yields two
resonance points with equal amplitude magnification factors, but
the amendment is almost meaningless as long as the tuning is
determined by the conventional method.
One of the authors of the present paper has found the new
approach for exact solution 8. To summarize, the approach is
based on an observation of the trade-off between two resonance
amplitude magnification factors. In light of this, the solution to the
original problem is reduced to a quadratic equation that is derived
in terms of the discriminant of a quartic equation. In the case of
undamped primary system, it turns out that the optimum parameters, minimum amplitude magnification factors, resonance and
anti-resonance frequencies, and the sensitivities of the amplitude
Contributed by the Technical Committee on Vibration and Sound for publication
in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received August 2000;
Revised April 2002. Associate Editor: R. L. Clark.

576 Vol. 124, OCTOBER 2002

magnification factors are entirely algebraic. The numerical extensions are also shown, enabling the formulation of efficient solutions for the damped primary system, resulting in more direct
applications.
We have tested the new formulation in several cases, for example, mobility and accelerance cases, which are simple variations of the compliance case that is discussed in the present paper
9,10. It turned out that the exact solutions are not more complex
than those of the fixed-points methods, and in some cases they are
rather simpler. Our formulation is also applicable to the hysteretically damped vibration absorbers, and in some cases, algebraic
exact solutions are obtained assuming hysteretically damped primary system. The present paper explains this new formulation by
taking the most typical case as an example, and refers to the
sensitivities and the optimality of the solution, which are not
clearly discussed in the previous reports.
The algebraic method is applicable to the linearized model of
the passive gyroscopic damper PGD, and the exact solution has
been obtained 11. The PGD means a dynamical model of gyroscopic stabilizer developed by one of the authors. It consists of a
single gimbal and a set of torsional springs and viscous dampers
attached to the gimbal axis. The gimbal is a housing of the rotor
driven by an electric motor to maintain constant revolution. The
basic dynamical characteristics are similar to that of the rotational
type dynamic vibration absorber. The PGD is essentially a semiactive system, which is more effective under ordinary conditions.
The assumption of an undamped primary system leads to the algebraic solution of optimum parameters.

Undamped System

2.1 Model for Analysis. The model for analysis is a system


with two-degree-of-freedom shown in Fig. 1, consisting of a primary system and a dynamic vibration absorber as an auxiliary
system. As external force acts on the primary system, the dynamic
vibration absorber is provided to reduce the resonant motion of
the primary system to within safe limits.
The equations of motion for this system are
m 1 x 1 c 1 c 2 x 1 k 1 k 2 x 1 c 2 x 2 k 2 x 2 w,
m 2 x 2 c 2 x 2 k 2 x 2 c 2 x 1 k 2 x 1 0,

Copyright 2002 by ASME

(1)

Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 11/18/2014 Terms of Use: http://asme.org/terms

tween G( j P ) and G( j Q ) can be postulated. On this assumption, it is guaranteed that the optimum design is derived using
equivalent resonance magnification factors, i.e.:
G s G j A G j B .

(5)

Therefore, the problem is reduced to determining p and z such that


the common resonance amplitude is minimized. This assumption
will be reviewed in detail after the solution of simplified problem.
In section 2.10, local optimality of the solutions is revalidated by
numerical evaluation of the sensitivities of amplitude magnification factors with respect to p and z.
The polynomials n( ) and d( ) are introduced to denote the
numerator and the denominator of G( j ) 2 :
n p 2 2 2 4z 2 2 ,

(6)

d p 2 2 1 p 2 2 4 2 4z 2 2 1 1 2 2 ,
(7)

Fig. 1 Dynamic vibration absorber attached to single-degreeof-freedom system

where m 1 and m 2 are the masses of the primary and auxiliary


systems respectively; x 1 and x 2 are the displacements of the primary and auxiliary systems respectively; k 1 and k 2 are the spring
constants of the primary and auxiliary system respectively; and c 1
and c 2 are the viscous damping coefficients of the primary and
auxiliary systems respectively.
The equations of motion 2 are entirely dimensionless.

x 1 2Z2 z x 1 1 p 2 x 1 2 zx
2 p 2x 2 w
,
.
x 2 2zx
2 p 2x 2 2zx
1 p 2x 1 0,
(2)

The dimensionless parameters are defined as m 2 /m 1 , z


c 2 /(2 m 1 k 1 ), Zc 1 /(2 m 1 k 1 ), p 2 k 2 /( k 1 ), and w

w/(k 1 x 0 ). The displacements x 1 and x 2 are transformed as x 1


x 1 /x 0 and x 2 x 2 /x 0 where x 0 is the arbitrary unit length. The
unit time for the nondimensionalization is determined according
to the undamped eigenperiod of the primary system t 0
m 1 /k 1 , and thus, the dimensionless time is defined as
t/t 0 . Note that a dot over the displacement in these equations
means the derivative with respect to the dimensionless time. For
1 /d . The transfer function from the excitation
example, x 1 dx
1 (s)/W
(s) is
force to the primary system displacement G(s)X
naturally dimensionless. The exact solution in this section assumes an undamped primary system (Z0). Hence, the transfer
function, or compliance, is written as:
s 2 2zsp 2
G s 2
.
2
s 1 s 2zsp 2 s 2 2zsp 2

(3)

2.2 Problem Formulation. Using the present notation, the


objective function is expressed by G(s) , i.e., H norm of
G(s); with decision variables p and z. It is reasonable to assume
that G(s) has two distinct resonance points. These are denoted A
and B, with frequencies A and B ( A B ). This leads to the
equation:
G s max G j A , G j B .

(4)

The new approach outlined in this paper originates from the observation of a trade-off relation between G( j A ) and G( j B ) .
It is well recognized that each fixed point is very close to the
corresponding resonance point, and that the trade-off relation beJournal of Vibration and Acoustics

where all terms are even-ordered in . For simplicity, 2 , n( ),


and d( ) are replaced by , N(), and D(), respectively.
These notations lead to the equation
N
h,
D

(8)

which holds at the resonance points A and B, and where h denotes


the square of the resonance amplitude magnification factor. Similarly, another equation holds:
N DND
0,
D2

(9)

where denotes /. Equation 9 holds because the tangents at


A and B are horizontal. Equations 8 and 9 are simplified to
NhD0 and N hD 0, respectively. Therefore, the equation F()0 has two double roots; A A2 and B B2 , where
F()DN/h.
2.3 Simplification. The function F() is a monic quartic
polynomial.
F 4 b 1 3 b 2 2 b 3 b 4

(10)

By definition, the coefficients are


b 1 2 1p 2 p 2 2z 2 4 z 2 2 2 z 2 ,

(11)

b 2 4p 2 2 p 2 p 4 2 p 4 2 p 4 r 2 8z 2 8 z 2 ,
(12)
b 3 2 p 4 p 4 p 2 r 2 2r 2 z 2 ,

(13)

b 4 p 4 r 2 ,

(14)

where h is replaced by 1/(1r 2 ). The new parameter, r


1h 1 (h1, 0r1) has been introduced for simplicity.
The existence of two double roots leads to the factorized form:
F A 2 B 2 .

(15)

Vietas theorem gives the alternative expressions for b i (i


1, . . . ,4):
b 1 2 A B ,

(16)

b 2 A2 4 A B B2 ,

(17)

b 3 2 A B A B ,

(18)

b 4 A2 B2 .

(19)

OCTOBER 2002, Vol. 124 577

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 11/18/2014 Terms of Use: http://asme.org/terms

By eliminating A and B , these four equations are reduced to


two equations:
b 1 b 4 b 3 0,
b 21
4

(20)

2 b 4 b 2 0.

f 1 2 1 2 p 2 rr 2 z 2 p 2 1r r 1 p 2 0.
(22)
Equation 22 is then solved with respect to z0:

8 43 3/2
r1

,
r2
6480 27 2
where r 2 is excluded because r 1 0r 2 .

2.6 Derivation of p. Substituting rr 1 into Eq. 24 results


in an eighth-order equation with even-ordered terms. The equation
can now be factorized as follows:
f 2

1r 2 1 6
p 2 2 p 2 2 p 2 2 p 2
r 2 p 2 1 2 r 2

(23)

Substitution of Eq. 23 into f 2 0 eliminates z in f 2 . Hence, f 2


0 is expressed as a polynomial in terms of , p, and r. For
simplicity, f 2 0 is now rewritten using xp 2 . The result is a
quartic equation in x:
f 2 x a 0 x a 1 x a 2 x a 3 xa 4 0,

(24)

a 0 1 6 1r 2 ,

(25)

a 1 4 1 4 r 1r 2 ,

(26)

where

a 2 2 1 2 3 r 2 1r 2 ,

(27)

a 3 2 1r r 22 2r2 r r ,

(28)

a 4 r 1r .

(29)

Given a positive solution x to f 2 (x)0 for the parameter r, the


substitution of p x into Eq. 23 determines the value of z that
satisfies f 1 0. Thus, the problem reduces to the minimization of
r under the positive root condition of Eq. 24.
2.4 Derivation of Discriminant. The number of real roots
to Eq. 24 depends on the parameter r. The coefficients a i (i
0, . . . ,4) are polynomials in r. The continuity of the root of
algebraic equation with respect to the coefficients, guarantees that
f 2 (x)0 yields the multiple root at the boundary between two
sections that correspond to the discrete numbers of real solution.
The discriminant D 4 is a factor of the resultant:
R f 2 , f 2 a 70 D 4 .

(30)

For simplicity, the equation R( f 2 , f 2 )0 is used instead of D 4


0. The resultant is defined as the determinant of a matrix, which
is convenient for use with a formula manipulation software such
as Mathematica 12.

a0

a1

a2

a3

a4

a0

a1

a2

a3

a4

a0

a1

a2

a3

a4

3a 1

2a 2

a3

4a 0

3a 1

2a 2

a3

4a 0

3a 1

2a 2

a3

4a 0

3a 1

2a 2

a3

R f 2 , f 2 4a 0
0

(34)

r 1 1 43
) 1

2r 1 43 2
) 1

(35)

(36)

where is the unique positive solution to Eq. 34. The fact that z
is positive can be confirmed algebraically by setting rr 1 and
p .
2.7 Optimum Parameters. It is proven algebraically that
the determination of p and z results in two double roots for Eq.
10. The details are omitted here to avoid prolixity. Equations
16 and 19 show that A and B satisfy the quadratic equation
2

b1
b 4 0.
2

(37)

Therefore, the optimum value of r is


r optr 1

8 43 3/2
.
6480 27 2

(38)

The characteristics of the compliance function in proximity to the


optimal solution are described as follows. If rr opt , then a
double root occurs to Eq. 24. For the case rr opt , it corresponds to two real roots. In this case, two equal height resonance
points are yielded, which are slightly higher than that of the optimum case. Note that two distinctive values of p exist for the
equivalent height. The remaining case rr opt , results in a pair of
conjugate complex solutions. This simply means that such height
equivalence is not achievable.
The optimum parameter values rr opt , pp opt , and zz opt
yield equal height resonance points at A and B . The optimum
parameter values are obtained by step-by-step substitutions as follows:
p opt

z opt

r opt 1 43
) 1

(39)

2
2
p opt
1r opt r opt 1 p opt

2
2r opt 1 2 p opt
r opt

(40)

The optimum resonance magnification factor is


(31)

2.5 Determination of r. The highest order term in Eq. 31


is the 26th power of r. The following quadratic equation is obtained by eliminating the common factor 16 6 (1 ) 18r 14(1
r) 4 (r 2 1) 3 . Here, r0 and r1 correspond to h1 and h
, respectively; then both solutions are apparently inadequate.
578 Vol. 124, OCTOBER 2002

(33)

2 p 2 2 2 ,

p 1r r 1 p
.
2r 1 2 p 2 r
2

(32)

The roots of Eq. 32 are

(21)

Substitution of definitions 1114 into Eqs. 20 and 21 yields


the simultaneous equations f 1 0, f 2 0 with respect to p and
z, where r is the given parameter. The first equation is rearranged
as:

6480 27 2 r 2 16 r64 1 0

G j A G j B

1
2
1r opt

(41)

Hence, maxG(jA),G(jB) is minimized.


Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 11/18/2014 Terms of Use: http://asme.org/terms

2.8 Resonance Frequencies. The application of the optimum parameters given by Eq. 3840 results in the real roots
A2 A and B2 B for the quadratic equation 37.

1
A
2
2
1 1 p opt
2 1 2 z opt

B
2

Since h/ 0 holds at resonance points A and B, the sensitivities now reduce to


dh h

,
dp p

(47)

dh h

.
dz z

(48)

or

1
2
2 2
2
2 1 2 z opt
r opt
1 1 p opt
p opt
4
(42)

2.9 Anti-resonance Frequency. The resonance and antiresonance frequencies correspond to the solutions of the equation
N DND 0, which is defined as the numerator of Eq. 9,
because the tangents at the resonance and anti-resonance points
are horizontal. The frequencies are the real roots of the monic
quintic equation:
5 14p 2 p 2 8z 2 4 z 2 2 2 z 2 4 4p 2 6p 4

In addition, the square amplitude magnification factor h is a


rational polynomial defined by Eqs. 6 and 7. The derivative
with respect to p is
dh n p d p n p d p
.

dp
d p 2

(49)

Note that n(p) and d(p) are the numerator and denominator of h,
and that denotes / p in Eq. 49.
Analogously, the derivative with respect to z is written as:

4 p 4 8z 2 16p 2 z 2 24 p 2 z 2 8 2 p 2 z 2 16z 4

dh n z d z n z d z

.
dz
d z 2

32 z 4 16 2 z 4 3 6p 4 p 4 4p 6 5 p 6

These expressions permit precise evaluation of sensitivities. By


numerical calculations,

2 p 6 16p 2 z 2 12 p 2 z 2 8p 4 z 2 16 p 4 z 2 8 2 p 4 z 2
16z 16 z p 4p 2 p p 2 p p
4

8z 8 z 1 p 0.
2

2 4

(43)

Exploiting the fact that the resonance frequencies satisfy quadratic


equation 37, the following cubic equation is derived by dividing
the left-hand side of Eq. 43 by that of Eq. 37:

dh A
0,
dp

dh B
0,
dp

dh A
0,
dz

dh B
0,
dz

(50)

(51)

can be readily verified for various values of the mass ratio ,


where h A and h B indicates h at the resonance points A and B,
respectively. Initially, the discussion depended on the assumption
that any optimal solution yields the same resonance amplitude
h A h B , but now the existence of a trade-off between h A and h B
has been confirmed. Now, the proof for the local optimality of the
solution is intuitive.

3 6z 2 3p 2 2 p 2 3p 4 p 4 p 2 r2z 2 4p 2 z 2
6 p 2 z 2 2 2 p 2 z 2 4z 4 8 z 4 4 2 z 4 p 2 2p 4

Fixed-Points Method

Using the notations provided in the previous section, the conventional results of the fixed-points method are readily expressed
as

p 4 p 6 p 6 p 2 r2p 4 r p 4 r2z 2 8p 2 z 2
4 2 p 2 z 2 2p 4 z 2 8 p 4 z 2 10 2 p 4 z 2 4 3 p 4 z 2

40 p 2 z 4 48 2 p 2 z 4 24 3 p 2 z 4 4 4 p 2 z 4 8z 6
32 z 6 48 2 z 6 32 3 z 6 8 4 z 6 0

z fp
(44)

Here, r, p, and z are the optimal parameters given by Eqs. 38


40. There is one real root and a pair of complex conjugate roots.
Cubic equation 44 is reducible and its real root is given by
Cardanos formula. Thus, the algebraic solutions for the resonance
and anti-resonance frequencies of the exact optimization are derived.
2.10 Sensitivity Analysis. Next, the sensitivities of the amplitude magnification factors of the resonance points are evaluated
with respect to p and z. Sensitivities are important because of the
parametric errors of practical DVAs and the effect on the optimality of the preceding results. For simplicity, G( j ) 2 is abbreviated to h. The sensitivities are evaluated by the general expressions:
dh h h

,
dp p p

(45)

dh h h

.
dz z z

(46)

Journal of Vibration and Acoustics

1
,
1

(52)

3
,
8 1 3

(53)

p fp

4p 2 rz 2 4 p 2 rz 2 2 2 p 2 rz 2 8z 4 8 2 z 4 12p 2 z 4

where pp pf and zz pf 1,8.


The amplitude magnification factor of the fixed points is:
G j P G j Q

2
1 .

(54)

The general equation for the fixed-points frequencies is:

2
1p 2 p 2 p 2 0.
2

(55)

Substitution of p fp and z fp given in Eqs. 52, 53 results in the


simplified expression:

1
1

(56)

The resonance points of the fixed-points method A and B are


located very close to the fixed points P and Q, respectively. These
resonance frequencies satisfy Eq. 43, because it is derived from
N DND 0. Similarly, Eq. 43 is simplified by substituting
the definitions of p fp and z fp given in Eqs. 52 and 53:
OCTOBER 2002, Vol. 124 579

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 11/18/2014 Terms of Use: http://asme.org/terms

Fig. 2 General view of compliance curves 0.05, Optimum parameter values for undamped primary system
,
Fixed-points method
, Primary system z - " -

2020 10 2 3 4 4012 2 3

4 1 3
4 1 4

404 5 2 2
5
1

0.
4 1 5
1 6
1 7
(57)

Since the factorization of this equation is unknown, numerical


methods are required to obtain a solution. It is empirically evident
that the fixed-points method yields two resonance points, and
hence has three real roots and a pair of complex conjugate solutions. These numerical results enable the precise calculation of the
maximum amplitude magnification factor of the fixed-points
method.

Numerical Examples

Fig. 3 Close-ups of point A and B 0.05, Optimum parameter values for undamped primary system
, Fixedpoints method
, Primary system z - " -

4.1 Compliance Curves. Figure 2 shows the compliance


curves for 0.05. Two curves, corresponding to the exact optimization and fixed-points methods, overlap each other and are
difficult to distinguish. The regions proximal to the resonance
points A and B are shown in detail in Fig. 3. The amplitude magnification factors of the left and right resonance points with exact
optimization are G( j A ) G( j B ) 6.407921. Conversely,
G( j P ) G( j Q ) 6.403124, G( j A ) 6.407492, and
G( j B ) 6.408443 for the fixed-points method. The fixedpoints method results in a lower left resonance point A and a
higher right resonance point B , while the average is nearly equal
to that of the exact optimization. The tuning ratio p fp of the fixedpoints theory is highly accurate. The deterioration is considered to
originate mainly from errors in the damping coefficient z fp of the
fixed-points method. Figure 4 shows that the variation between
the exact optimization and fixed-points method increases in proximity of the antiresonance point.
4.2 Optimum Parameter Values. Table 1a shows the
tuning ratios p derived using the fixed-points method and the exact optimization for various values of the mass ratio , and b
shows the dimensionless damping coefficient z calculated by these
two methods. The variations of p and z are relatively trifling when
the mass ratio is less than 0.01. The variation of z tends toward
0.5% at 0.1, while the variation of p remains very small.
4.3 Resonance Amplitude Magnification Factors. Table 2
shows the resonance amplitude magnification factors calculated
580 Vol. 124, OCTOBER 2002

Fig. 4 Close-up of point C 0.05, Optimum parameter


values for undamped primary system
, Fixed-points
method

by the exact optimization and the fixed-points methods. The fixedpoints method is highly precise. For example, the ratio is only
0.023% for a relatively large mass ratio 0.1. The ratios remain
between 0.5% and 2.3% for larger mass ratios 1 to 10, although
Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 11/18/2014 Terms of Use: http://asme.org/terms

Table 1 Optimum frequency ratio and damping coefficient

approach applied to the undamped primary system is considered


impractical in the damped case. Therefore, a numerical approach
is used in preference to an algebraic method.
The equation f 1 (p,z,r)0 is rewritten as zg 1 (p,r). Substituting this into f 2 (p,z,r)0 results in f 2 (p,g 1 (p,r),r)0. The
condition of a double root f 2 (p,r)0 with respect to p is written
as f 2 / p0. The expression is evaluated by implicit differentiation and rearranged as:
f 3

f1 f2 f1 f2

0.
p z
z p

(58)

Note that f 2 (p,g 1 (p, r),r)0 is equivalent to the simultaneous


equations f 1 (p,z,r)0, f 2 (p,z,r)0 , providing a set of three
equations including Eq. 58. Equation 58 forces the singularity
of the Jacobian matrix of f 1 (p,z), f 2 (p,z) .
The equations f 1 0, f 2 0 are rather simple. Conventional
formula manipulation systems such as Mathematica readily compute the partial derivative of f 1 and f 2 with respect to p and z 12.
Then, Eq. 58 is readily derived as a polynomial.
f 1 2 1 2 p 2 rr 2 z 2 4 p 2 rZzp 2 1r r 1 p 2
2p 2 rp 2 Z 2

(59)

f 2 1r 2 4 1 2 z 2 4 1 4 z 4 8 z 2 1 2 z 2 1 Z
4 2 13 1 z 2 1 Z 2 16 zZ 3 4Z 4 2 r1
4 1 3 z 2 8 1 zZ4 1 Z 2 p 2

Table 2 Maximum amplitude magnification factor

(60)

The Newton-Raphson method is applicable in solving Eqs.


59, 60 and 58 simultaneously. The algebraic exact solution to
the undamped case is a valid initial value for the iteration. If the
iteration does not converge, then the continuation method is required, in which Newton-Raphson method is repeatedly applied
with the gradual increment of Z from zero.
Figure 5 shows the compliance curves with various values of
the primary system damping coefficient. In each case, the numerical optimization is compared to the diversion of the exact solution
to the undamped case. For example, by numerical methods taking

such high mass ratios are impractical for usual applications. The
fixed-points method is considered highly accurate as a design
method for the DVA.

Damped Primary System

The next task is to extend the preceding method to a damped


primary system 13,14. The assumption that the optimum solution satisfies the condition G( j A ) G( j B ) is inherited from
the undamped case. In this case, the equations originate from the
coincidence of the resonance amplitude magnification factor f 1
0 and f 2 0 include the primary system damping coefficient Z.
An eighth order algebraic equation corresponding to the quartic
equation 24 is generated by a similar procedure. However, the
resultant is more complicated compared to Eq. 31. The algebraic
Journal of Vibration and Acoustics

Fig. 5 Reduction of maximum amplitude magnification factors


by numerical optimizations 0.05. Iterative solutions that
take accounts of the primary system damping are displayed for
several values of the primary system damping ratio
. The
additional curves are provided only for comparison purposes,
where the optimum parameter values for undamped primary
system are diverted to the damped cases
.

OCTOBER 2002, Vol. 124 581

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 11/18/2014 Terms of Use: http://asme.org/terms

accounts of a damping coefficient Z0.01, the maximum amplitude decreases by 8.5%. Importantly, the improvement in accuracy is much greater than the difference between the fixed-points
method and the exact algebraic optimization of the undamped
case.

Concluding Remarks

It has been shown that the algebraic exact solution exists for the
minimization of the resonance amplitude magnification factor by
the dynamic vibration absorber attached to the undamped primary
system. The exact algebraic expressions for the resonance and
antiresonance frequencies have been obtained. Initially, a trade-off
relation between two resonance points was assumed, but a sensitivity analysis numerically identified the trade-off relation. The
local optimality of the solution then became apparent. This approach was extended to the damped primary case, but complementary numerical solution was still required. The fixed-points
method was shown to be highly accurate, especially for small
mass ratios of less than or around unity. This method is applicable
to the linearized model of the passive gyroscopic damper whose
dynamical characteristics are similar to that the rotational dynamic
vibration absorber. Their algebraic solution of optimum parameters has been obtained under the assumption of undamped primary system. The algebraic approach is also applicable to an air
damped dynamic vibration absorber, which is modeled by a threeelement type system. New expressions for the optimum parameters have been derived 15 by the extension of the method described in the present paper.

References
1 Brock, J. E., 1946, A Note on the Damped Vibration Absorber, ASME J.
Appl. Mech., 134, p. A-284.
2 Ormondroyd, J., and Den Hartog, J. P., 1928, The Theory of the Dynamic
Vibration Absorber, Trans. ASME, 507, pp. 922.

582 Vol. 124, OCTOBER 2002

3 Den Hartog, J. P., 1956, Mechanical Vibrations 4th ed., McGraw-Hill, New
York.
4 Korenev, B. G., and Reznikov, L. M., 1993, Dynamic Vibration Absorbers:
Theory and Technical Applications, John Wiley & Sons, New York.
5 Ikeda, T., and Ioi, T., 1977, On Dynamic Vibration Absorbers for Damped
Vibration Systems, Trans. Jpn. Soc. Mech. Eng., 43369, pp. 17071715.
6 Soom, A., and Ming-San, Lee., 1983, Optimal Design of Linear and Nonlinear Vibration Absorbers for Damped Systems, ASME J. Vibr. Acoust., 1051,
pp. 1121193.
7 Haddad, W. M., and Razavi, A., 1998, H2 , Mixed H2 /H , and H2 /L1 Optimally Tuned Passive Isolators and Absorbers, ASME J. Dyn. Syst., Meas.,
Control, 1202, pp. 282287.
8 Nishihara, O., and Matsuhisa, H., 1997, Design of a Dynamic Vibration
Absorber for Minimization of Maximum Amplitude Magnification Factor
Derivation of Algebraic Exact Solution, Trans. Jpn. Soc. Mech. Eng., Ser.
C, 63614, pp. 3438 3445.
9 Nishihara, O., Asami, T., and Watanabe, S., 2000, Exact Algebraic Optimization of a Dynamic Vibration Absorber for Minimization of Maximum Amplitude Response 1st Report, Viscous Damped Absorber, Trans. Jpn. Soc.
Mech. Eng., Ser. C, 66642, pp. 420 426.
10 Asami, T., Nishihara, O., and Watanabe, S., 2000, Exact Algebraic Optimization of a Dynamic Vibration Absorber for Minimization of Maximum Amplitude Response 2nd Report, Hysteretic Damped Absorber, Trans. Jpn.
Soc. Mech. Eng., Ser. C, 66644, pp. 1186 1193.
11 Azuma, T., Nishihara, O., Honda, Y., and Matsuhisa, H., 1997, Design of a
Passive Gyroscopic Damper for Minimization of Maximum Amplitude Magnification Factor, Preprint of JSME in Japanese, No. 974-2, pp. 5354.
12 Wolfram, S., 1991, MathematicaA System for Doing Mathematics by Computer Second Edition, Addison-Wesley, Reading, MA.
13 Asami, T., and Hosokawa, Y., 1995, Approximate Expression for Design of
Optimal Dynamic Absorbers Attached to Damped Linear Systems 2nd Report,
Optimization Process Based on the Fixed-Points Theory, Trans. Jpn. Soc.
Mech. Eng., Ser. C, 61583, pp. 915921.
14 Nishihara, O., Asami, T., and Kumamoto, H., 1999, Design Optimization of
Dynamic Vibration Absorber for Minimization of Maximum Amplitude Magnification Factor Consideration of Primary System Damping by Numerical
Exact Solution, Preprint of JSME in Japanese, No. 99-7 I, pp. 365368.
15 Asami, T., and Nishihara, O., 1999, Analytical and Experimental Evaluation
of an Air-Damped Dynamic Vibration Absorber: Design Optimizations of the
Three-Element Type Model, ASME J. Vibr. Acoust., 1213, pp. 334 342.

Transactions of the ASME

Downloaded From: http://vibrationacoustics.asmedigitalcollection.asme.org/ on 11/18/2014 Terms of Use: http://asme.org/terms

Você também pode gostar