Vous êtes sur la page 1sur 18

Journal of Sound and Vibration (1996) 198(3), 361378

TRANSIENT RESPONSE OF STRUCTURES ON


VISCOELASTIC OR ELASTOPLASTIC MOUNTS:
PREDICTION AND EXPERIMENT
K. Gii:, R. Diroir :Nb G. Frrr:ris
Laboratoire de Mecanique des Structures, ESA CNRS 5006, INSA de Lyon,
20, Avenue Albert Einstein, 69621 Villeurbanne, France
(Received 2 August 1995, and in nal form 4 June 1996)
This paper deals with the response of structures on non-linear elastomer or all-metal
mounts under impact excitations. The mounts are modelled with the tangent and
instantaneous stinesses deduced by using experimental load-deection loops, and with
modal damping factors. The nite element method and a dynamic condensation are used
to predict rst the response of a beam on elastomer or all-metal mounts under hard and
soft impact forces, and then the response of a rigid body on elastomer mounts under base
excitation. Associated experimental tests enable the non-linear model to be validated.
7 1996 Academic Press Limited
1. INTRODUCTION
The viscoelastic behaviour of elastomer mounts and the elastoplastic behaviour of dry-
friction mounts are commonly used for the passive isolation of fragile on-board
equipment. In recent years much eort has been spent on the dynamic modelling of the
suspension and on the prediction of the dynamic behaviour of the structure-suspension
assembly.
Elastomer and metal mounts have a mostly non-linear behaviour. The non-linearities
arising from material behaviour, geometrical design and dry-friction depend on deection,
temperature and forcing frequency, especially for the elastomer mounts [13]. As there is
an interaction between these parameters, the dynamic behaviour of such mounts is
complex and a general model taking into account the type of excitation is dicult to
establish.
The harmonic response of the system can be deduced from the superposition of linear
vibration and large elastic deformation. Padovan [4] has established a theory based on the
superposition of the independent static and dynamic modulus, while Nashif et al. [2],
who have extended the non-linear static equation of Mooney-Rivlin to dynamics,
combined them. They considered that the complex modulus approach is often more
convenient for modelling the linear behaviour of the elastomer material than the
generalized standard or derivative models. They carried out characterization experiments
using the harmonic response of a single-degree-of-freedom system (SDOF). This type of
system has been used in reference [5] to characterize elastomer mounts according to
parameters such as overheating or external temperature; then the harmonic response of
a exible beam on the frequency dependent mount was predicted and it compared well
with the experimental result. In reference [6], a method involving an experimental and
361
0022460X/96/480361 +18 $25.00/0 7 1996 Academic Press Limited
i. cii: r1 :i. 362
numerical treatment was used to establish the harmonic and impact responses of a slender
structure containing frequency dependent viscoelastic materials.
The hysteretic model is often used in the presence of dry-friction or of elastoplastic
behaviour. The original dierential model proposed by Bouc [7], generalized by Wen [8],
is commonly used for non-linear random dynamic analysis [911], but its adaptation to
the prediction of steady state response is dicult and has some disadvantages. The trace
method is based on the polynomial approximation of the hysteretic loop. In the case of
the bilinear system under stationary random excitation, Badrakhan [12] extended the trace
method, keeping the non-linear stiness and approximating the hysteretic damping by an
equivalent viscous damping. Using an identical hypothesis, Ko et al. [13] established
an empirical model with the amplitude displacement dependent parameter of a wire
cable subjected to harmonic excitation. The identication of the hysteretic loop is
performed with an alternative use of the Fast Fourier Transform and of the Chebychev
polynomials.
The restoring force surface method is used for the non-parametric identication of
non-linear systems. This method, developed by Masri and Caughey [14], establishes the
surface of the restoring force of the mount versus the phase plane, the deection, velocity
and acceleration being known each time. Worden [15, 16] proposed a technique to measure
only one signal and then to obtain the two others using an SDOF system. The integration
is better than the dierentiation. Especial attention must be paid to determine the
integration constants and the type of excitation.
There are dierent analytical and numerical methods to solve non-linear dynamic
systems containing a low number of degrees of freedom, particularly dynamic stationary
analysis methods. In the frequency domain the incremental method of Lau et al. [17, 18],
integrated to the harmonic balance, Timoshenko et al. [19], the IHB method, is
often employed to predict the stationary response of strongly non-linear structures
under arbitrary excitation. There is also the Galerkin-Newton-Raphson (GNR) method
developed by Ferri and Dowell [20] which is adapted to multi-harmonic excitation. Specic
techniques are necessary if the response is dicult to predict in the frequency domain.
Cameron and Grifn [21] developed a method based on frequency/time domain alternance
(AFT) which is applied to an SDOF system containing a non-linear elastic/perfectly-plastic
model under multi-frequency excitation. Such an alternance using the Galerkin-Levenberg-
Markardt approach was presented by Wong et al. [22] to analyze the multi-harmonic
stationary response of a system with amplitude displacement dependent parameters. A
step-by-step incremental linearization technique (SILFD), performed at each integration
step in the frequency domain through an FFT algorithm, was presented by Venancio-Filho
and Claret [23] for an SDOF system with displacement dependent stiness and frequency
dependent viscous damping, subjected to an arbitrary excitation.
Current computation facilities permit the easy use of the nite element method (FE) to
predict the linear or non-linear response of a structure modelled with several thousands
of DOF. In addition, the reduction methods such as the modal or sub-structure methods
decrease computer time consumption. The pseudo-modal method can be used to predict
the frequency response of a system with frequency dependent characteristics [24]. This
method has been extended to the impact response of a ski with the updating of the modal
stiness matrix [6]. Setio et al. [25, 26] introduced the non-linear mode and modal
superposition method to predict the steady state response of a multi-DOF system with
displacement dependent stiness under harmonic excitation.
In this paper, the experimental quasi-static force deection curves enable the calculation
of the tangent and instantaneous stinesses of elastomer and all-metal mounts which are
used to compute the static equilibrium position while only instantaneous stiness is
s1ric1irrs oN NoN-iiNr:r xoiN1s 363
Figure 1. Tested elastomer mount.
employed to predict the impact response. An FE model, a dynamic condensation and the
Newmark method are used to predict the time response of structures with localized
non-linearities due to the two kinds of mounts, under force or base excitations.
Experimental investigations carried out on a rigid structure and on a exible structure
validate the predicted response.
2. MODELLING OF THE NON-LINEAR MOUNTS
The elastomer mount, shown in Figure 1, has a viscoelastic behaviour while the all-metal
mount, shown in Figure 2, has an elastoplastic behaviour. They have overheating, external
temperature, frequency and deection dependent behaviour [3]. Under a steady state
temperature condition, it has been shown that the deection amplitude has a predominant
eect, especially for the all-metal mount which is, in addition, almost independent of
forcing frequency. In order to predict the static equilibrium position under preload and
the transient response of the structure, the quasi-static load-deection curves are
investigated to evaluate the stiness of the mounts. The classical experimental test consists
of picking-up the fth load-deection loop, the rst four loops allowing the overheating
of the mount. Axial and transverse tests are performed. A linear curve tting of the
load-deection curves denes the tangent stiness k
T
and the instantaneous stiness k.
The procedure is illustrated in Figure 3.
Figure 2. Tested all-metal mount.
i. cii: r1 :i. 364
Figure 3 (a) Load-deection loop of a mount. (b) Load-deection curve. (c) Linear curve tting.
Let x be the deection in the applied force direction. The restoring force F
r
of the
mount between the deections at points i and j is (a list of symbols is given in the
Appendix)
F
r
=ax +b. (1)
At any point (F
r
, x) of segment ij let the instantaneous stiness k be
k =F
r
/x =a +b/x (2)
and the tangent stiness k
T
be
k
T
=k +(1k/1x)x =a. (3)
The constants a and b are dened as
a =(k
j
x
j
k
i
x
i
)/(x
j
x
i
), b =(k
i
k
j
)x
i
x
j
/(x
j
x
i
) (4, 5)
with
k
i
=(F
r
)
i
/x
i
and k
j
=(F
r
)
j
/x
j
. (6)
The quasi-static load-deection curve, its approximation and the instantaneous stiness
k of the elastomer mount are shown in Figures 47, and those of the all-metal mount in
Figures 811. Figures 4, 5 and Figures 8, 9 concern the axial behaviour while the transverse
behaviour is illustrated by Figures 6, 7 and 10, 11. It should be noted that the transverse
loops shown in Figures 6 and 10 are performed with two mounts.
s1ric1irrs oN NoN-iiNr:r xoiN1s 365
Figure 4 Elastomer mount: measured load-deection loop. (a) Traction. (b) Compression.
3. DYNAMIC CONDENSATION
The nite element model of an industrial structure uses a large number of DOF.
The updating of non-linear parameters at each step of the response requires techniques
such as dynamic condensation to reduce the number of equations and the computer
time consumption. In the case of a linear structure on non-linear mounts, the reduction
is performed on the DOF describing the structure while all the DOF of the suspension
are kept. The Craig and Bampton method [27] is well adapted to this problem. With no
external force, the static equilibrium of the linear structure, schematically represented
in Figure 12, is governed by a set of N equations which can be separated into two
sub-systems,
$
[K
str
ii
[K
str
ci
]
[K
str
ic
]
[K
str
cc
]
%6
{x
i
}
{x
c
}
7
=
6
{0}
{0}
7
, (7)
Figure 5. Elastomer mount: axial linear curve tting (), and instantaneous stiness ().
i. cii: r1 :i. 366
Figure 6. Elastomer mount: measured transverse load-deection loop of two identical elastomer mounts.
Figure 7. Elastomer mount: transverse linear curve tting (), and instantaneous stiness ().
where {x
c
} and {x
i
} are respectively the interface and interior nodal displacement vectors.
System (7) expresses the interior DOF as a function of the interface DOF which in fact
correspond to the connection of the mounts:
{x
i
} =[K
str
ii
]
1
[K
str
ic
]{x
c
}. (8)
The matrix [F
str
c
] of the static displacement shapes permits a change of the basis,
6
{x
i
}
{x
c
}
7
=[F
str
c
]{x
c
}, (9)
Figure 8. All-metal mount: measured traction-compression load-deection loop.
s1ric1irrs oN NoN-iiNr:r xoiN1s 367
Figure 9. All-metal mount: axial linear curve tting (), and instantaneous stiness ().
with
[F
str
c
] =
$
[K
str
ii
]
1
[K
str
ic
]
[I]
%
, (10)
[I] being the identity matrix. The rst lower dynamic modes of the linear structure built
at the mount connections, computed with
[M
str
ii
]{x
i
} +[K
str
ii
]{x
i
} ={0}, (11)
yield the matrix [F
str
n
] which is extended to the whole DOF of the system:
[F
str
n
] =
$
[F
str
i
]
[0]
%
. (12)
The reduction matrix [F
CB
] can be established,
[F
CB
] =[F
str
n
, F
str
c
] =
$
F
str
i
]
[0]
[K
str
ii
]
1
[K
str
ic
]
[I]
%
, (13)
Figure 10. All-metal mount: measured transverse load-deection loop.
i. cii: r1 :i. 368
Figure 11. All-metal mount: transverse linear curve tting (), and instantaneous stiness ().
to permit the change of basis
6
{x
i
}
{x
c
}
7
=[F
str
CB
]{x
CB
}, (14)
to phrase the structures reduced mass and stiness matrices:
[m
str
CB
] =[F
CB
]
T
[M
str
][F
CB
], [k
str
CB
] =[F
CB
]
T
[K
str
][F
CB
]. (15, 16)
In addition, let [m
sus
CB
] and [k
sus
CB
(x
CB
)] be the reduced mass matrix and the reduced variable
stiness matrix, respectively, so that:
[m
sus
CB
] =[F
CB
]
T
[M
sus
][F
CB
], [k
sus
CB
(x
CB
)] =[F
CB
]
T
[K
sus
(x
CB
)][F
CB
]. (17, 18)
Finally, the dynamic characterization of the structure-suspension system expressed in the
CB base is
([m
str
CB
] +[m
sus
CB
]){x
CB
} +([k
str
CB
] +[k
sus
CB
(x
CB
)]){x
CB
} ={0}. (19)
The static equilibrium is computed with the iterative process of the Newton-Raphson
method using the tangent [k
sus
TCB
] and instantaneous [k
sus
CB
] stinesses of the suspension,
([k
str
CB
] +[k
sus
TCB
(x
i
CB
)]){Dx
CB
} ={F
S
} ([k
str
CB
] +[k
sus
CB
(x
i
CB
)]{x
i
CB
}, (20)
Figure 12. Schema of the structure with localized non-linearities.
s1ric1irrs oN NoN-iiNr:r xoiN1s 369
with {Dx
CB
} the displacement increment at the actual iteration, {x
i
CB
} the displacement at
the previous iteration, and {F
S
} the applied static force.
4. PREDICTION OF THE NON-LINEAR TIME RESPONSE
Here the structure-suspension system is subjected to constant concentrated {F
C
}
or density forces and to dierent forces {F(t)} which are commonly time dependent.
The FE model uses the classical beam, shell, and solid elements for the structure, and
a two node variable stiness element with three translations per node for the mount.
Half of the mount mass is distributed at each node. It is possible to introduce additional
mass and stiness matrices.
It is convenient to reduce the computation basis once again, by using the modal
reduction method in particular to introduce modal damping factors. The modal basis [F
p
]
deduced from the conservative system (19) at the static equilibrium position permits a
change of base:
{x
CB
} =[F
p
]{p}. (21)
Consequently, equations (19), in the presence of forces, become
[m]{p} +[c]{p } +([k
str
] +[k
sus
(p)]){p} ={ f
c
} +{ f(t)}, (22)
with {p} the modal parameters, [m] =[F
p
]
T
([m
str
CB
] +[m
sus
CB
])[F
p
] the modal mass
matrix, [k
str
] =[F
p
]
T
[k
str
CB
][F
p
] the modal stiness matrix of the structure, [k
sus
(p)] =
[F
p
]
T
[k
sus
CB
(x
CB
)][F
p
] the modal stiness matrix of the suspension, [c] the constant
modal damping matrix, { f
c
} =[F
p
]
T
[F
CB
]
T
{F
c
}, and { f(t)} = [F
p
]
T
[F
CB
]
T
{F(t)} the modal
force vectors.
In the modal basis, the modal damping factors are assumed to be the diagonal
coefcients of matrix [c],
c
j, j
=zm
j, j
k
j, j
/Q
j
, (23)
where m
j, j
and k
j, j
are the computed modal mass and stiness parameters while the
Q-factor Q
j
is evaluated by using an experimental sine wave excitation. The non-linear
time response of the structure-suspension system subjected to dierent types of excitation
is computed by using the Newmark method. At each time step Dt, the Newmark method
solves the linear system
(4[m]/Dt
2
+2[c]/Dt +[k
str
] +[k
sus
(t +Dt)]){p(t +Dt)}
={ f
c
} +{ f(t +Dt)} +(4[m]/Dt
2
+2[c]/Dt){p(t)}
+(4[m]/Dt +[c]){p (t)} +[m]{p(t)}, (24)
which, with the relationships
{p (t +Dt)} ={p (t)} +(2/Dt)({p(t +Dt)} {p(t)}), (25)
{p(t +Dt)} ={p(t)} +(2/Dt)({p (t +Dt)} {p (t)}), (26)
predicts {p(t +Dt} from {p(t)}, {p (t)}, {p(t)}.
If the system is assumed to be at rest at the static equilibrium position, the iterative
process can be initialized. The acceleration is computed by using the system
[m]{p(t
0
)} ={ f
c
} +{ f(t
0
)} ([k
str
] +[k
sus
(p(t
0
))]){p(t
0
)}, (27)
i. cii: r1 :i. 370
and the DOF of the structure by using the relationship
6
x
i
x
c7
=[F
CB
][F
p
]{p}. (28)
The Newmark method used, known as the average acceleration method, is unconditionally
stable in the case of linear systems [28]. This implicit integration scheme employed for
non-linear analysis [28] is one of the most efcient methods [29]. An initial dynamic
computation gives the highest frequencies of the system and the time step can be chosen.
The stiness [k
sus
(t)] takes the place of [k
sus
(t +Dt)] in equation (24), in order to avoid
an additional iterative process inside a time step. The small variation of the suspension
stiness inside small time-step justies this substitution.
4.1. rorcr rxci1:1ioN
The impact force applied to DOF l at the time t
i0
can be described by a half-sine series:
F(t) = s
n
i =1
a
i
sin V
i
(t t
i0
). (29)
The unit component of the vector {F
l
} expressed in the basis of the structure,
{F
l
}
T
={0 0 1 0 0}, (30)
corresponds to the DOF l. The two successive basis transformations give the vector { f
l
}
so that
{ f
l
} =[F
p
]
T
[F
CB
]
T
{F
l
}, (31)
which, in the case of the simultaneous force applied at m-DOF, enables expression of
the force vector { f(t)} as
{ f(t)} = s
m
l =1
{ f
l
}F(t). (32)
Thus the force vector is rapidly built at each time step. Relationship (32) can be extended
for any type of time dependent force F(t).
4.2. n:sr rxci1:1ioN
Here the base is rigid and the acceleration vector {G} contains only translation
components. Equation (24) remains valid if vectors {p}, {p }, {p} are considered with
regard to the moving base; i.e., as relative quantities. Let the force due to the base
acceleration be
{g(t)} =[m]{g(t)}, (33)
where the acceleration {g(t)}, dened by
{G(t)} =[F
CB
][F
p
]{g(t)}, (34)
is calculated by solving the linear system:
[F
p
]
T
[F
CB
]
T
[F
CB
][F
p
]{g(t)} =[F
p
]
T
[F
CB
]
T
{G(t)}. (35)
s1ric1irrs oN NoN-iiNr:r xoiN1s 371
Figure 13. Experimental set-up of the force excitation test.
5. APPLICATIONS
5.1. rorcr rxci1:1ioN
A 0375 m long steel beam, shown in Figure 13, is built-in at one of its ends and
connected onto the elastomer or all-metal mount at L
P
=0335 m. Its rectangular cross
section is 004 m wide and 0004 m high. A proximity probe, with a linear measurement
range of 02 mm is located at L
D
=0067 m, while a piezo-electric force transducer is stuck
onto the beam at L
F
=0360 m. The soft impact is performed on the force transducer with
a tennis ball and the hard impact with an hammer.
The structure is modelled with 12 nite beam elements, the force transducer with an
additional mass element and the mount with a variable stiness element. The damping is
introduced by the Q-factors measured with a sine wave excitation of the beam-mount
system. The rst six modes are kept to dene the CB basis.
Figure 14. Measured soft impact.
i. cii: r1 :i. 372
Figure 15. (a) Vertical displacement response of the beam: prediction (); experiment (); (b) predicted
vertical displacement response of the beam at the mount location (), and instantaneous stiness ().
Figure 16. Measured soft impact.
5.1.1. Beam/elastomer mount system
The signal of the soft impact performed, displayed in Figure 14, can be considered as
a half sine series. The predicted response of the beam is compared in Figure 15(a) to the
measured displacement given by the proximity probe. Figure 15(b) shows the predicted
s1ric1irrs oN NoN-iiNr:r xoiN1s 373
displacement of the beam at the mount location and the non-linear mount stiness versus
the time of the analysis. The non-linear zone is reached.
5.1.2. Beam/all-metal mount system
In the impact and response graphs the previous presentation is used. Figures 16, 17(a)
and 17(b) are for the soft impact analysis, and Figures 18, 19(a) and 19(b) for the hard
impact. As previously, the predicted and measured displacement of the beam compare well.
5.2. n:sr rxci1:1ioN
The tested system used in reference [30] is schematically represented in Figure 20.
The 0140 0120 0080 m parallelepiped structure weighing 1025 kg is non-linearly
supported by four vertical elastomer mounts.
A signal data processing station controls the base acceleration produced by using
a 80 kN electrodynamic shaker. The vertical base acceleration, shown in Figure 21(a),
picked-up with a piezo-electric transducer, is approximated by a half sine series, displayed
Figure 17. (a) Vertical displacement response of the beam: prediction (); experiment (); (b) predicted
vertical displacement response of the beam at the mount location (), and instantaneous stiness ().
i. cii: r1 :i. 374
Figure 18. Measured hard impact.
Figure 19. (a) Vertical displacement response of the beam: prediction (); experiment (); (b) predicted
vertical displacement response of the beam at the mount location (), and instantaneous stiffness ().
s1ric1irrs oN NoN-iiNr:r xoiN1s 375
Figure 20. Schematic of the parallelepiped structure on four elastomer mounts. Values (m): a =0140;
b =0120, c =0080; H=0008.
in Figure 21(b). The piezo-electric accelerometer located at the top centre of the structure
gives the vertical response. By using measured sine wave transmissibilities in the X-,
Y- and Z-directions, the rst six viscous damping factors are evaluated at approximately
0025.
Four solid parabolic brick elements model the structure, and one variable stiness
element models each mount. Only the rst six modes are kept in the basis.
The gravity eect computed with the Newton-Raphson method establishes the static
equilibrium position. The comparison of the predicted and measured absolute acceleration
responses is shown in Figure 22(a). The relative displacement and the non-linear mount
stiness versus the analysis time is displayed in Figure 22(b).
6. CONCLUSIONS
Techniques for predicting the transient response of linear structures on two types
of non-linear suspensions have been developed. They are based on the experimental
modelling of the mounts, on a reduction FE model of the structures and on experimental
validations. The modelling of the mount has been carried out mainly on non-linear
stiness, because, the eect of non-linear stiness is more important than the eect
of damping. The damping is taken into account by using the modal damping factor.
In future studies it will be convenient to distinguish the mount damping from the structural
damping.
Figure 21. (a) Measured base acceleration; (b) half sine series approximation of the base acceleration.
i. cii: r1 :i. 376
Figure 22. (a) Absolute vertical acceleration response of the structure on four elastomer mounts: prediction
(); experiment (); (b) predicted relative vertical displacement response of the structure on four elastomer
mounts (), and instantaneous stiness ().
The experimental characterizations of elastomer and all-metal mounts permit the
denition of the quasi-static load deection curves and the establishment of a mount nite
element with account taken of the instantaneous and tangent stinesses. In the complete
FE model, the structure is condensed to the interface DOF while all the DOF of the
suspension are kept. The FE method with the Craig and Bampton reduction is well
adapted to the time response of linear structures with localized non-linearities.
The predicted and measured transient responses of rigid and exible structures on the
two types of mounts correspond well. Thus, it has been shown that the instantaneous
stiness and modal damping factors, both deduced from simple experiments, efciently
model the non-linear behaviour of the two types of mounts under soft or hard impact
excitation. Under these force or base excitations, the role of the damping is of secondary
importance.
ACKNOWLEDGMENTS
The authors would like to thank Vibrachoc S.A. and the French Ministry of Research
and Technology for providing technical and nancial support.
REFERENCES
1. J. H:rris and A. S1rvrNsoN 1986 Rubber Chemical Technology 59, 740763. On the role of
non-linearity in the dynamic behaviour of rubber components.
s1ric1irrs oN NoN-iiNr:r xoiN1s 377
2. A. D. N:snir, D. I. G. JoNrs and J. P. H:NbrrsoN 1985 Vibration Damping. New York:
John Wiley and Sons.
3. K. Gii: 1994 The` se de Doctorat: Institut National des Sciences Appliquees de Lyon.
Comportement dynamique de structures localement non line aires. Cas des suspensions.
4. J. P:bov:N 1987 International Journal of Computers and Structures 27, 249257. Finite element
analysis of steady and transiently moving/rolling non-linear viscoelastic structure.
5. R. Diroir, K. Gii:, M. Poxiri 1992 Mecanique, Materiaux, Electricite 444, 3840.
Comportement dynamique de structures exibles sur suspension passive non line aire:
calcul-mesure.
6. C. Uiricn, G. Frrr:ris, M. L:i:NNr and J. L:croix 1993 ASME Proceedings of 14th Biennial
Conference on Mechanical Vibration and Noise. Vibration, Shock, Damage, and Identication
of Mechanical Systems, Albuquerque, New Mexico 64, 913. Prediction of the frequency and
time responses of composite structures: Application to skis.
7. R. Boic 1987 Proceedings of the Fourth Conference on Non-linear Oscillation, Prague,
Czechoslovakia, 315. Forced vibrations of a mechanical system with hysteresis. (Abstract.)
8. Y. K. WrN 1976 Journal of the Engineering Mechanics Division, ASCE 102, 249263. Method
for random vibration of hysteretic system.
9. Y. K. WrN 1990 Journal of Applied Mechanics 47, 150154. Equivalent linearization for
hysteretic system under random excitation.
10. T. T. B:nrr and M. N. Noori 1986 Journal of Vibration, Acoustics, Stress, and Reliability
in Design 108, 411420. Modelling general hysteresis behaviour and random vibration
application.
11. J. B. Ronrr1s and P. D. Si:Nos 1990 Random Vibration and Statistical Linearization. New York:
John Wiley and Sons.
12. F. B:br:in:N 1987 International Journal of Non-linear Mechanics 22, 315325. Rational study
of hysteretic system under stationary random excitation.
13. J. M. Ko, Y. Q. Ni and Q. L. Ti:N 1992 The International Journal of Analytical and Experimental
Modal Analysis 7, 111127. Hysteretic behaviour and empirical modelling of a wire-cable
vibration isolator.
14. S. F. M:sri and T. K. C:icnr. 1979 Journal of Applied Mechanics 46, 433447. A non-
parametric identication technique for non-linear dynamic problems.
15. K. WorbrN 1990 Mechanical Systems and Signal Processing 4, 295319. Data processing and
experiment design for the restoring force surface method, Part I: integration and dierentiation
of measured time data.
16. K. WorbrN 1990 Mechanical System and Signal Processing 4, 321344. Data processing and
experiment design for restoring force surface method, Part II: choice of excitation signal.
17. L. S. L:i, Y. K. CnriNc and S. Y. Wi 1982 Journal of Applied Mechanics 49, 849853.
A variable parameter incrementation method for dynamic instability of linear and non-linear
elastic system.
18. L. S. L:i, Y. K. CnriNc and S. Y. Wi 1983 Journal of Applied Mechanics 50, 871876.
Incremental harmonic balance method with multiple time scales for a periodic vibration of
non-linear systems.
19. S. TixosnrNio, D. H. YoiNc and Jr. W. Wr:vrr 1974 Vibrations problems in engineering.
New York: John Wiley and Sons.
20. A. A. Frrri and E. H. Dovrii 1988 Journal of Sound and Vibration 124, 207224. Frequency
domain solutions to multi-degree of freedom, dry friction damped systems.
21. T. M. C:xrroN and J. H. GrirriN 1989 Journal of Applied Mechanics 56, 149154. An
alternating frequency/time domain method for calculating the steady-state response of
non-linear dynamic systems.
22. C. W. WoNc, Y. Q. Ni and J. M. Ko 1993 Modal Analysis: The International Journal of
Analytical and Experimental Modal Analysis 8, 6378. Steady-state dynamic response of
structures with non-linear hysteretic isolators.
23. F. VrN:Ncio-Finio and A. M. Ci:rr1 1989 Proceedings of Damping 89, West Palm Beach,
Florida II, FDD-1-11. Non-linear analysis with frequency-dependent damping.
24. H. T. Znoi, J. Drr H:coii:N, G. Frrr:ris and M. L:i:NNr 1988 59th Shock and Vibration
Symposium IV, 265271. Prediction of modal characteristics and harmonic response of
viscoelastically damped structures.
25. S. Sr1io, H. D. Sr1io, C. H. L:x:rqir and L. Jrzrqiri 1990 Proceedings of the 15th
International Seminar on Modal Analysis, Belgium, 889899. Introduction of non-linear modes
to compute the steady-state response of MDOF systems.
i. cii: r1 :i. 378
26. S. Sri1o, H. D. Sri1o and L. Jrzrqiri 1992 The International Journal of Analytical and
Experimental Modal Analysis 7, 7593. Modal analysis of non-linear multi-degree of freedom
structures.
27. R. R. Cr:ic and M. C. C. B:xi1oN 1968 AIAA Journal 6, 13131321. Coupling of substructure
for dynamic analysis.
28. K. J. B:1nr 1982 Finite Element Procedures in Engineering Analysis. Englewood Cliffs, New
Jersey: Prentice Hall Inc.
29. Y. M. Xir 1996, Journal of Sound and Vibration 192, 321331. Assessment of time integration
schemes for non-linear dynamic equations.
30. R. Diroir, J. Drr H:coii:N, M. Poxiri and C. G:rNirr 1992 ASME Proceedings of 13th
Biennial Conference on Mechanical Vibration and Noise. Structural Vibration and Acoustics,
Miami, Florida DE 34, 171176. Shock and sine response of rigid structures on non-linear
mounts.
APPENDIX: LIST OF SYMBOLS
x suspension deection
k suspension instantaneous
stiffness
kT suspension tangent stiffness
Fr restoring force
f natural frequency, Hz
V excitation frequency, rad/s
{xc } interface nodal displacement
vector
{xi } interior nodal displacement
vector
[I] identity matrix
{0} vector of 0s
[0] matrix of 0s
[F
str
c
] matrix of static displacement
shapes
[F
str
n
] matrix of xed interface mode
shapes
[M
str
] structure mass matrix
[M
sus
] suspension mass matrix
[K
str
] structure stiffness matrix
[K
sus
(x)] suspension stiffness matrix
{Fc } external constant excitation
force
{F(t)} external time function excitation
force vector
{G(t)} base acceleration vector
CB Craig-Bampton representation
[xCB ] Craig-Bampton displacement
vector
[FCB ] reduction matrix
[mCB ] reduced mass matrix
[kCB ] reduced stiffness matrix
{p}, {p }, {p} modal co-ordinate, velocity and
acceleration vector
[Fp ] mode shape matrix
{m} modal mass matrix
{k} modal stiffness matrix
{c} modal damping matrix
{ fc } modal external constant exci-
tation force vector
{ f(t)} modal external time function
excitation force vector
{g(t)} modal base acceleration vector
t time
Dt numerical time step
Qj quality factor of the jth mode

Vous aimerez peut-être aussi