|
N - l
(2-27)
The coefficient matrix is a Vandermonde matrix [V] whose determinant is given as:
det (V) =
(2-28)
2 Identification of Nonlinearity Using First-order
3 7
Therefore, if all eigenvalues are distinct, then
exists and the unknown
can be uniquely determined.
When the damping matrix of the system does not satisfy the condition set in equation (2-
24), then the damping matrix cannot be diagonalised using the real modes of the
corresponding conservative system and the modes of the damped system will in general
become complex. Since the extent of departure of a given damping matrix from that of
equation (2-24) (often referred as nonproportionality) determines how complex the modes
of the system will be, the quantification of such departure becomes necessary in order to
study complex modes. Some research work on the quantification of nonproportionality of
a given damping matrix and the complexity of modes has been reported and the
relationship between complexity of modes and nonproportionality of damping matrix has
been investigated
The degree of complexity of a certain mode depends on the closeness of the natural
frequencies of the system. In the case when all the modes of the system are well
separated, even though the damping matrix is nonproportional
damping for
example), the modes will not be substantially complex. Theoretically, considerably
complex modes can only occur when modes become close. This effect of mode spacing
on the complexity of modes is to be discussed below based on the perturbation theory.
In order to illustrate the effect of mode spacing on the complexity of modes, a hysteretic
damping model is assumed in the analysis although the relationship between the hysteretic
and viscous damping models will be discussed later. Also, assume that the structural
damping matrix [D] is of second order in its Euclidean norm sense when compared with
the system’s stiffness matrix [K], then to first order approximation, the
modeshape of
the damped system
can be expressed in terms of the modal parameters of the
corresponding conservative system and damping matrix
as:
(2-29)
In
case when the
mode is well isolated, then
(which is a scalar), will
be of second order compared with
and, therefore,
will be effectively real.
However, if there are close modes, say mode r and mode
then when
Identification of Nonlinearity Using First-order
3 8
will no longer be of second order of
and
will become
considerably complex.
In the above discussion, both viscous and hysteretic damping models have been used.
The relationship between these two damping models and the complexity of modes need to
be discussed.
In the case of viscous damping, the eigenvalue problem of the system becomes quadratic
as:
(2-30)
While the standard eigenvalue problem is in the form of
(2-3 1)
where [A] and [B] can be complex matrices in general. In order to solve the quadratic
eigenvalue problem given in equation
some mathematical transformations are
required, namely:
By solving equation
the eigenvalues and the so-called ‘A-normalised’
eigenvectors (normalised to the system’s generalised mass matrix [A]) of the system can
be obtained.
In the case of hysteretic damping, the eigenvalue problem becomes:
(2-32)
Compared with equation
the solution to equation (2-32) is standard and since
and
+ i[D]) in this case, the eigenvectors for the hysteretic damping
case are therefore
(normalised to the mass matrix [M] of the system).
Because of the different normalisation procedures used when the different damping
models are considered, the corresponding eigenvectors are apparently quite different,
even for the case of proportionally damped systems although, in fact, they differ only by
Identifkation of Nonlinearity Using First-order
3 9
a complex scaling factor. These differences in amplitude as well as phase angles of the
corresponding eigenvectors often cause confusion to analysts and it is therefore necessary
to establish the relationship between the ‘A-normalised’ and the mass-normalised
eigenvectors. In the case of proportional damping, the corresponding
mode
eigenvectors of hysteretically-
to [M]) and viscously-damped
to
[A]) systems can be expressed, in theory, as
=
(where
is a complex
scaling factor). Substitute
into the ‘A-normalisation’ condition for the
mode together with
and
(2-33)
can be calculated as:
(2-34)
From equation
it can be seen that in the case of proportional damping, the
normalised’ modeshape for the case of viscous damping,
is the corresponding
modeshape
for the case of hysteretic damping scaled by a factor of
and a phase rotation of
For the case of nonproportional
damping, the relationship between these two damping models has been investigated in
2.4.2 NUMERICAL EXAMPLE OF COMPLEX MODES
As discussed above, when the damping distribution of the structure is nonproportional,
complex modes exist. However, the degree of complexity of a mode is mainly dependent
on the closeness of the structure’s natural frequencies. order to illustrate these points, a
numerical case study was carried out.
The system used in the numerical study is the 4DOF mass-spring system shown in
The mass matrix
stiffness matrix [K] and hysteretic damping matrix [D]
of the system are:
2 Identification of Nonlinearity Using Fist-order
4 0
=
Fig. 2.11 A 4DOF Mass-spring System
0.800 0.000 0.000
0.000 1.005 0.000 0.000
0.000 0.000 1.000
0.000
1
0 . 0 0 0
.
0.000 0.000 0.000 0.800
3.00 -1.00 -1.00 0.00
N/m).
0.00 0.00 0.00
The calculated eigenvalue matrix
and eigenvector matrix
are:
0.0000 0.0000
0.0000
0.0000 0.0000 0.0000
1
1
2 Identification of Nonlinearity Using First-order
4 1
=
1 SO”)
1
176”)
1 . 0 ”)
1
From the eigenvalue and eigenvector matrices, it can be seen that modes 1 and 2 are quite
well separated and, as a result, their modeshapes are effectively real. While for modes 4
and 3, since they are very close in natural frequency, their modeshapes become quite
complex when the damping is nonproportional, as it is in this case. Physically, complex
modes can be explained as a kind of travelling wave which transfers energy from one part
of the structure to another during vibration.
2.4.3
Do'stlaringiz bilan baham: |
