Sie sind auf Seite 1von 6

First Steps in Vibration Analysis Using ANSYS

Ulrike Zwiers

Spring 2007

Referring to the two-mass-spring system sketched below, the basic steps in performing
a vibration analysis with ANSYS are outlined. To validate the results of the modal
analysis, the natural frequencies, amplitude ratios and the particular solution are also
computed analytically. Furthermore, ANSYS is used to determine the system response
resulting from an harmonic excitation.

1 Equations of Motion and Closed-Form Solution


Formulating the balance of linear momentum for the free-body diagrams shown in
Figure 2 yields the following equations of motion:
m1 ẍ1 = −k1 x1 + k2 (x2 − x1 )
m2 ẍ2 = −k3 x2 − k2 (x2 − x1 )
The corresponding matrix equation reads as
M ẍ + Kx = 0 ,
with the mass matrix M and the stiffness matrix K being given by
   
m1 0 k1 + k2 −k2
M= and K = ,
0 m2 −k2 k2 + k3
respectively.

m1 m2
k1 k2 k3

Figure 1: Two-mass-spring system

1
1 EQUATIONS OF MOTION AND CLOSED-FORM SOLUTION 2

m1 m2
k1 x1 k2 (x2 − x1 ) k2 (x2 − x1 ) k3 x2

x1 x2

Figure 2: Free-body diagrams

To simplify the further notation, the masses and stiffness coefficients are assumed in
the form
m2 = m1 = m , k3 = k1 = k , and k2 = k1 /5 = k/5 .
Consequently, the mass and stiffness matrices may be written as
   
1 0 k 6 −1
M =m and K = ,
0 1 5 −1 6
respectively.
As for the analytical solution, an ansatz function of the form x(t) = x̂est may be chosen.
However, since the considered system is undamped, it is known in advance that the two
roots s1 and s2 will turn out to be purely imaginary. Thus, one may express the solution
directly in terms of the natural frequency ω as x(t) = x̂eiωt . Substituting this ansatz
into the matrix equation of motion yields
(K − ω 2 M )x̂ = 0 .
To obtain solutions other than the trivial solution x̂ = 0, the determinant of the matrix
of the coefficients must vanish, i.e., det(K − ω 2 M ) = 0, which gives for the problem at
hand the characteristic equation
 2
6 1
k − ω m − k2 = 0 .
2
5 25
This equation is quadratic in ω 2 and may therefore be solved, for example, by using
the so-called pq-formula resulting in
k 7k
ω12 = and ω22 = .
m 5m

The amplitude vector x̂ is not uniquely defined, but the ratios of the amplitude com-
ponents corresponding to the frequencies ω1 and ω2 can be determined, namely,
x̂1 x̂1
=1 and = −1 ,
x̂2 x̂2
2 MODAL ANALYSIS USING ANSYS 3

respectively. Thus, one mode represents an in-phase vibration, while the other one
represents an out-of-phase vibration.
As an example for forced vibrations, the first mass m1 is now assumed to be harmoni-
cally excited in horizontal direction by an external force of the form

F = F̂ sin Ωt ,

where Ω represents the excitation frequency. The governing equations of motion may
now be expressed in matrix notation as

M ẍ + Kx = f sin Ωt ,

with
 

f= .
0

The steady-state response is given by the particular solution of this inhomogeneous


differential equation for which an ansatz function of the form

x = x̂ sin Ωt

is chosen. Substituting this ansatz into the equation of motion yields

(K − Ω2 M )x̂ = f .

For the simplifying assumptions introduced before (m2 = m1 = m, k3 = k1 = k,


k2 = k1 /5 = k/5), the amplitudes are computed as
6

5
k − Ω2 m F̂
x̂1 = ,
2 m 2 − k2
6

5
k − Ω 25

k
5

x̂2 = 2 .
6 k2
5
k − Ω2 m − 25

2 Modal Analysis Using ANSYS


A modal analysis is performed to determine the vibration characteristics (i.e., the nat-
ural frequencies and mode shapes) of a structure. In ANSYS, a modal analysis is
also the starting point for other, more detailed, dynamic analyses, such as a harmonic
response or a transient analysis.
First, the type of analysis is to be specified by using the command line ANTYPE,MODAL.
The modal analysis options may then be specified by the command MODOPT. The mass
and stiffness matrices arising in the modal analysis of elastic structures are usually
large but sparse. The corresponding eigenvalue problem is typically of order 105 − 106 ,
3 HARMONIC ANALYSIS USING ANSYS 4

and several hundred to thousand eigenvectors are often required as the frequency range
of interest for the modal analysis increases. Several algorithms for solving such large
eigenvalue problems have been implemented in ANSYS. Of course, the modal analysis
of simple problems like the two-mass-spring system considered here does not require
any special solution method, thus, the default setting (Block-Lanczos algorithm) may
be used, while for more complex problems, other algorithms might be more suitable.
However, in any case, the number of modes to extract must be specified, which is also
done by the command MODOPT.
The ANSYS-element type MASS21 models a point mass in space, while simple spring-
damper-combinations are represented by the element type COMBIN14. The mass and
stiffness parameters are to be input as Real Constants.
The spring lengths can be arbitrarily selected since they are used only to define the
spring direction.
After successful computation, the results of a modal analysis may be reviewed in the
General Postprocessor (\post1). The natural frequencies are listed in the Results Sum-
mary SET,LIST, where each set is associated with a natural mode. For a certain mode
(selected by using the SET-command), the corresponding mode shape may be animated
(via the ANMODE-command).
The unit of frequencies computed by ANSYS is Hz specifying the number of cycles per
second, whereas the frequencies determined in the previous section represent angular
frequencies specifying the number of radians per second. Since there are 2π radians in
one revolution, the frequency fn and the angular frequency ωn are related through
ωn
fn = .

Thus, for validation of the results, it is essential to designate frequencies in the correct
unit.
The enclosed APDL-input file modal schwingerkette.txt documents the modal analy-
sis of the two-mass-spring system at hand.

3 Harmonic Analysis Using ANSYS


An harmonic response analysis (ANTYPE,HARMIC) yields solutions of time-dependent
equations of motion associated with linear structures undergoing steady-state vibration.
To this end, all loads and displacements are assumed to vary sinusoidally at the same
known frequency.
As for the two-mass-spring system introduced above, the first mass m1 is now assumed
to be harmonically excited in horizontal direction by an external force as discussed in the
first section. The excitation frequency Ω is now supposed to vary within a certain fre-
quency range (to be defined by the HARFRQ-command), for which an amplitude-versus-
frequency plot should be generated using the Time-History Postprocessor (\post26).
4 REFERENCE RESULTS 5

To get an adequate response curve, solutions at specific intervals are to be determined,


i.e., a suitable number of substeps is to be specified (using the NSUBST-command). Here,
it is essential to specify a stepped loading (KBC,1), since otherwise, the load amplitude
is gradually increased with each substep. In addition, harmonic analysis options may
be specified by the command HROPT.
The enclosed APDL-input file harm schwingerkette.txt documents the harmonic re-
sponse analysis of the two-mass-spring system at hand.

4 Reference Results
The results obtained from the modal analysis and the harmonic analysis performed in
ANSYS are validated by comparing them against the corresponding analytical solutions
derived in the first section. To this end, the following parameters are chosen

m = 10 kg , k = 2000 N/m and F̂ = 100 N . (1)

It is pointed out that the units of these parameters can be arbitrarily chosen as long as
they are consistent.

Figure 3: Harmonic response of the two-mass-spring system


4 REFERENCE RESULTS 6

The two natural angular frequencies of the considered system are ω1 = 14.1421 [s−1 ]
and ω2 = 16.7332 [s−1 ], which agrees perfectly with the natural frequencies computed
by ANSYS: f1 = 2.2508 [Hz] and f2 = 2.6632 [Hz]. These values are also confirmed by
the amplitude-versus-frequency plot shown in Figure 3.
As a further validation, the amplitudes x̂1 and x̂2 at a certain excitation frequency
should be compared. At Ω = 2.4 [Hz], for example, both the analytical solution and
the ANSYS computation yield concordantly x̂1 = −0.0875 [m] and x̂2 = −0.2776 [m].

Das könnte Ihnen auch gefallen