Professional Documents
Culture Documents
Dynamic Analysis of Mechanical Systems With Clearances
Dynamic Analysis of Mechanical Systems With Clearances
DUBOWSKY
Senior Engineer,
Dynamic Analysis of Mechanical Systems
With Clearances3
O p t i c a l Technology Division,
Perkin-Elmer Corp.,
Danbury, Conn.
Introduction
X2-X|
hEKHAPS the most significant result of this investiga-
tion is t h a t a simple clearance model, the impact pair, will ex-
hibit the nonlinear behavior common to complex systems con- JPERIOD OF I t h PERIOD OF i
taining clearances. It is hoped t h a t insight into the behavior CYCLE CYCLE
of such systems can be obtained by studying the characteristics
Fig. 5 Free-vibration motion of Impact Pair
of the impact pair. The following sections contain a discussion
of the dynamic response of the impact pair in the case of free
vibrations, constant-load operation, and under displacement-
forced excitation. required excessive computer time to yield sufficient accuracy,
We proceed directly to the integration of the equations of while the less-direct method would result in an efficient software
motion derived in P a r t I. package.
*-(-». - . -
STATIC
COMPRESSION^
x 2 -x,
>r^
L-^
.006
U n = 16,000 RAD/SEC
r =.001 INCHES
£=.025 X 2 -X| IN INCHES PER SECOND
P = 0. POUNDS MODE I
BEHAVIOR
.003 "2AI
2 3 4 5 IN INCH,
TIME IN MILLISECONDS
0 5.0 10.0
IMPACT VELOCITY IN INCHES PER SECOND
Fig. 10 Vibration frequency versus impact velocity at surface (— r),
constant-load operation
8 12 16 20
TIME IN MILLISECONDS
U n = 16,000 RAD/SEC
£ =.025
X,= 1.00 INCH
io!ooo (7
Ui -
- . 0 0 0 ( N 0 CLEARANCE)
ill 1 I J l l i i i in J I
Approximate Solutions
T h e applicability of the approximate nonlinear techniques of
control engineering to the solution of the impact pair equations
was determined next. The most promising method appeared
to be the well-known describing-function technique.
The describing-function method can be used in two ways to
predict the dynamic response of nonlinear systems. One, it can
0 I 2 3 4 5 6 7 8 9 10
be used in conjunction with the Nyquist criterion to determine
CLEARANCE r, IN MILS (IO- 3 inches)
the stability of a freely vibrating system. This includes predict-
Fig. 14 Force-amplification versus clearance, case of forced-motion ing the existence of limit cycles, their frequencies, and ampli-
excitation tudes. Two, the describing function can be used in very simple
systems to determine the frequency response to sinusoidal forcing
able clearance-to-driving-amplitude ratio of 0.001. In the top functions. The former application is the more common one,
curve of this figure, the three curves representing the driven and will be used here for the free-vibration case, while the second
member and the two impact boundaries are indistinguishable, method is applied to the displacement-forced case. Both of these
while the relative-motion curve clearly shows the nature of the techniques are limited by the assumption associated with the
motion. This curve also shows that even for relatively low describing function, namely that when the input of the nonlinear
damping (zeta equal to 0.025) and a high frequency ratio (10 element is a sine wave the output is assumed to be sinusoidal,
percent), the motion strongly resembles a series of constant-load but amplified by the gain factor, 0D, with a phase shift, <j>.
cases of alternating direction. The importance of permitting The magnitude of the gain, GD, and phase shift are functions
multiple impacts on one surface is clear from these figures. I t of the frequency and amplitude of the input sine wave and these
was found t h a t solutions could be obtained for all driving fre- two variables constitute the describing function. This assump-
quencies and that the fundamental-frequency component of tion results in all of the higher harmonics of the nonlinear element
these solutions was always the driving frequency. The forbidden response being taken as zero. Such a system can be represented
regions predicted by the Kobrinskii analysis were not in evidence. by the block diagram shown in Fig. 16, where G(s) and H(s) are
It can be seen from Fig. 11 that the maximum surface com- the transfer functions of the linear forward and feedback elements
pression, X cml , x , is significantly higher than the static surface expressed as functions of the Laplacian variables.
compression; this results in significant stress amplification. Fig. The expressions for GD and </> for a clearance or dead zone are
13 represents a plot of the surface compression as a function of well-known [11] and are given by:
frequency for several different values of clearances. These curves
show t h a t the surface compression, and therefore the internal R
GD{e) = - - V l - R*/e (23)
forces, increase with frequency. In addition, it can be seen t h a t IT t e
increasing the clearance will greatly increase the dynamic forces, <$> = o
u.
H (s)
k + cS
K> 7^£GO.*0
mS 2
-X? (s)
S Q -30 - X| =
-
IE =>
=> 1 - '°'°'X^^V^// DESCRIBING FUNCTION
« | -40 - y?j/ SOLUTION -
5?
5
-50
-
AX
-60
(J n = l 6 , 0 0 0 RAD./SEC.
£ =.025
X, = 1.000
I I I I 1111 I—I I I 11 H
.000 10.000 (j
APPROXIMATE
DESCRIBING
FUNCTION
SOLUTION
EXACT SOLUTION
NO CLEARANCE (SOLUTIONS
COINCIDENT)
X=-00l
"I
r=.000l
X|
Equation (26) is a transcendental expression for 6 which can be possibly useful result was obtained in an attempt to find an ap-
solved both numerically and approximately to yield e as a func- proximate solution to the transcendental equation (26) for the
tion r, the clearance; w„ and f; and the dynamic parameters, u results presented in Fig. 19. In this approximation, the de-
and Xi, the forcing frequency and amplitude. scribing-function value given by equation (23) was approximated
The accuracy of the describing-function approach to this prob- by:
lem, equation 26, was determined by comparing its response to
the exact solution. The results are shown in Fig. 19. Clearly 4£ (27)
OB 1
the describing function does not adequately predict the force 67T
amplification due the clearance effects. The reason can be seen
for the inadequacy of the describing function from Fig. 13, in With this approximation equation (26) becomes a quadratic equa-
which it is apparent t h a t a good part of the stress amplification is tion and thus solvable directly for the relative-motion amplitude: