Anda di halaman 1dari 8

Journal of Sound and Vibration (1995) 180(1), 177–184

LETTER TO THE EDITOR


APPLICABILITY OF THE PERTURBATION TECHNIQUE TO THE PERIODIC
SOLUTION OF ẍ + ax + bx 2 + gx 3 = 0

M. S. S
Applied Mathematics Division

A. P. B  B. N R
Structural Design and Analysis Division, Structural Engineering Group,
Vikram Sarabhai Space Centre, Trivandrum—695 022, India
(Received 9 February 1994, and in final form 20 June 1994)

1. 
The second order non-linear differential equation
ẍ + ax + bx 2 + gx 3 = 0, (1)
subject to the initial conditions
x = xmax , ẋ = 0 at t = 0, (2)
describes certain phenomena in the theory of non-linear vibrations of plates [1–3]. Here
a, b and g are constants, and overdots denote differentiation with respect to time t. Since
the restoring force function, f(x) = ax + bx 2 + gx 3, in the equation of motion (1), is a cubic
polynomial (with both quadratic and cubic terms) and hence a non-odd function for b % 0,
the behaviour of the oscillation is different for positive and negative amplitudes. Some of
the conventional tools for the analysis of non-linear oscillations, such as the averaging
techniques, the Lindstedt–Poincaré small non-linearity based perturbation scheme, the
method of multiple scales and harmonic balancing, are described in reference [1].
By using the perturbation technique, the non-linear frequency (v) of the plate can be
obtained as a function of the material properties, the dimensions of the plate (i.e., in terms
of a, b and g) and the amplitude of vibration (xmax ), as
v 2 = a[1 + {34 (g/a) − 56(b/a)2}xmax
2
]. (3)
In the perturbation method, the desired quantities are developed in powers of some
parameter which is considered small; then the coefficients of the resulting power series are
determined in a stepwise manner. The method is straightforward, although it becomes
cumbersome for actual computations if many terms in the perturbation series are required
to achieve a desired degree of accuracy. If the coefficients of the non-linear terms in the
differential equation do not involve small parameters, this method becomes inappropriate
[4]. Mickens [5] has indicated that the only generally applicable technique in such a
situation is the method of harmonic balance, which is a fast and efficient calculational
technique for determining approximate analytic solutions. The method of harmonic
balance has often been criticized for not yielding consistent results unless some information
about the behaviour of the system is known a priori [1, 6, 7]. This conclusion has been
drawn by comparing the result of the harmonic balance method with the perturbation
solution, which itself is inappropriate because both the methods give only approximate
results. A possible source of confusion is that, for small oscillations, the harmonic balance
method requires a three-term solution to agree with the (algebraic) perturbation results,
177
0022–460X/95/060177 + 08 $08.00/0 7 1995 Academic Press Limited
178    
but for larger amplitudes, the perturbation method is inappropriate. Gottlieb [8] derived
an approximate frequency–amplitude relation for larger amplitudes based on the method
of harmonic balance and suggested using the perturbation solution for small oscillations.
The potentiality of any approximate solution should be judged by a comparison with the
closed form solution (if it exists) or with the exact numerical integration. The
frequency–amplitude relation from the two-term solution in the method of harmonic
balance provided in reference [9] yields reasonably accurate results for the mixed-parity
non-linear oscillators.
The purpose of this study is to examine the potentialities of the perturbation technique
and the method of harmonic balance by comparing the results with those of exact
integration. It is noted that the perturbation technique for a non-zero value of b in
equation (1) fails to produce the solution or yields an erroneous result by comparison with
the exact solution even though a large number of terms (up to six) are included. On the
other hand, the method of harmonic balance stands apart as an unfailing approximate
method which yields reasonably accurate results.

2.  –´ 


The idea of the Lindstedt–Poincaré method is to introduce a new independent variable,
say t = vt, where initially v is an unspecified function of o. The new governing equation
(1) will contain v as the coefficient of the second derivative, and one thus has
v 2x0 + ax + bx 2 + gx 3 = 0, x = xmax , x' = 0 at t = 0, (4, 5)
which permits the frequency and amplitude to interact. Primes denote differentiation with
respect to t. One can choose the function v in such a way as to eliminate the secular terms
(i.e., to render the expansion periodic).
Assuming, as the expansion for v,
v 2 = v0 + ov1 + o 2v2 + o 3v3 + o 4v4 + o 5v5 + · · · , (6)
and as the expansion for x
x = ox1 (t) + o 2x2 (t) + o 3x3 (t) + o 4x4 (t) + o 5x5 (t) + o 6x6 (t) + · · · , (7)
and following the procedure as described in reference [1], one obtains the solution of
equation (1) as
x = (oa) cos f + (oa)2{C1 + C2 cos 2f} + (oa)3C3 cos 3f
+ (oa)4{C4 + C5 cos 2f + C6 cos 4f} + (oa)5{C7 cos 3f + C8 cos 5f}
+ (oa)6{C9 + C10 cos 2f + C11 cos 4f + C12 cos 6f} + · · · , (8)
where f = vt + u, (oa) and u are to be determined from the initial conditions (5), the
non-linear frequency v is

$ 6 0 1 0 17 6 01 01 0 17 %
3 g 5 b 3 g 143 b g 335 b
2 2 2 4

v2 = a 1 + − (oa)2 + + − (oa)4 + · · · ,
4 a 6 a 128 a 96 a a 864 a

(9)
and

01 01 01 01
1 b 1 b 1 b 1 g
2

C1 = − , C2 = , C3 = + ,
2 a 6 a 48 a 32 a
    179

0 1 0 10 1 01 0 10 1
19 b 5 b g 59 b 31 b g
3 3

C4 = − + , C5 = − ,
72 a 8 a a 432 a 96 a a

01 0 10 1 01 0 10 1 01
1 b 1 b g 79 b 43 b g 21 g
3 4 2 2

C6 = + , C7 = − − ,
432 a 96 a a 2304 a 768 a a 1024 a

01 0 10 1 01
b b g g
4 2 2
5 5 1
C8 = + + ,
20 736 a 2304 a a 1024 a

01 0 10 1 0 10 1
1529 b 4453 b g 1441 b g
5 3 2

C9 = − + − ,
41 472 a 4608 a a 2048 a a

01 0 10 1 0 10 1
1207 b 1703 b g 1387 b g
5 3 2

C10 = − + ,
6912 a 2304 a a 3072 a a

01 0 10 1 0 10 1
b b g 27 b g
5 3 2
89 7
C11 = + − ,
15 552 a 1440 a a 160 a a

01 0 10 1 0 10 1
b b g b g
5 3 2
1 1 2207
C12 = + + .
41 472 a 3584 a a 1 661 280 a a
The constants v1 , v3 and v5 in equation (6) are found to be zero and hence the odd powers
of o are absent in equation (9). Assuming that o in equation (9) is small compared to unity,
and expanding the frequency parameter v (up to the terms of o 2 ) from the squared
frequency (9), one obtains [1, 6, 7],
v = za{1 + [(9ga − 10b 2 )/24a 2 ](oa)2 }, (10)
Using the initial conditions (5) in equation (8), one obtains u = 0 and
xmax = (oa) + {C1 + C2 }(oa)2 + C3 (oa)3 + {C4 + C5 + C6 }oa)4
+{C7 + C8 }(oa)5 + {C9 + C10 + C11 + C12 }(oa)6 + · · · . (11)
It is possible to establish the relationship between the non-linear frequency (v) and
amplitude (xmax ) only by eliminating oa from equations (9) and (11). Otherwise, for the
specified value of oa, it is possible to determine the amplitude, xmax , from equation (11)
and the corresponding non-linear frequency v from equation (9).
Examination of equation (3) and equation (9) shows that the frequency (v) amplitude
(xmax ) relation (3) suggested in references [10, 11] is obtained by considering the first three
terms in equations (6) and (7), and assuming oa to be the amplitude, xmax . This assumption
may be valid for the case of very small oscillations [12] but, in general, this is not logical
from the mathematical point of view, as revealed by equation (11). Strictly speaking, the
periodic solution of the problem depends on the constants a, b and g in equation (1), and
the maximum amplitude, xmax .
3.   
In accordance with perturbation solution (8), the solution of equation (1) can be
assumed in the form
x = a1 + a2 cos t + a3 cos 2t + a4 cos 3t + a5 cos 4t + a6 cos 5t + · · · . (12)
180    
By considering the first four terms of equation (12) and applying the method of harmonic
balance [1], the frequency (v) corresponding to the amplitude (xmax ) is obtained by solving
the non-linear algebraic equations
a1 + a2 + a3 + a4 = xmax , (13)
aa1 + b(2a + a + a + a ) + g(4a + 6a1 a + 6a1 a +6a1 a + 3a a + 6a2 a3 a4 ) = 0,
1
2
2
1
2
2
2
3
2
4
1
4
3
1
2
2
2
3
2
4
2
2 3

(14)
aa2 + b(2a1 a2 + a2 a3 + a3 a4 ) + g(12a a + 12a1 a2 a3+12a1 a3 a4 + 3a
1
4
2
1 2
3
2

+ 6a2 a32 + 6a2 a42 + 3a22a4 + 3a32a4 ) − a2 v 2 = 0, (15)


aa3 + b(a + 4a1 a3 + 2a2 a4 ) + g(6a1 a + 12a a +12a1 a2 a4 + 6a a + 6a2 a3 a4
1
2
2
2
1
4
2
2
2
1 3
2
2 3

+ 3a33 + 6a3 a42 ) − 4a3 v 2 = 0, (16)


aa4 + b(2a1 a4 + a2 a3 ) + g(12a a + 12a1 a2 a3+a + 6a a4 + 3a2 a + 3a a ) − 9a4 v = 0.
1
4
2
1 4
3
2
2
2
2
3
2
3 4
2

(17)
The iterative procedure adopted for the set of non-linear algebraic equations (13)–(17)
is as follows. Let the set of non-linear algebraic equations be represented by
F (z ) = {Fk (v 2, a1 , a2 , a3 , a4 )} = 0. (18)
To find the numerical solution, the following iterative procedure can be applied:
z i + 1 = z i − F (z i )/F (z i ) (19)
Here F (z ) is the Jacobian matrix of F . The initial values of z assumed are as follows:
z 0 = {v 2, a1 , a2 , a3 , a4 }T = {V 2, j1 , j2 , j3 , j4 }, (20)
where
V 2 = a + b(2j1 + j3 ) + 34 gj2 (j2 + j4 ), (21)
j1 = { bj (a + gj + gj2 j4 )}/j5 ,
3
2
2
2
3
4
2
2
1
2 (22)
j3 = −bj2 {12 aj2 + j4 (a + 32gj22)}/j5 , (23)
j4 = j (−2ab + gj5 )/4j6 ,
3
2
2
j5 = −3a − 3gj (2a + gj ),
2 2
2
9
8
2
2 (24, 25)
j6 = b 2j22 (a + 32 gj22 ) + j5 (8a + 214 gj22 ), j2 = xmax . (26, 27)
The initial values (20) are obtained by assuming that the values of j1 , j3 , and j4 in equations
(13)–(17) are small compared to j2 , and neglecting the coefficient terms involving products
of these constants. Equation (19) is then iterated until =z i + 1 − z i = is sufficiently small, thus
resulting in the values of v 2, a1 , a2 , a3 and a4 for the specified values of xmax . The two-term
solution in the method of harmonic balancing is obtained by putting a3 = a4 = 0 and
solving equations (13)–(15). Similarly, by putting a4 = 0 and solving equations (13)–(16),
the three-term solution is obtained.

4.    


In reference [1], the solution of the problem was obtained by using the first three terms
of equation (12) in the method of harmonic balance. Upon assuming a small value of A2 ,
and neglecting the terms higher than O(A22 ), the squared frequency parameter (v 2 ) is
expressed explicitly in terms of A2 , which results in expression (3) if A2 is identified with
xmax . This can also be verified from the initial value of the squared frequency (V 2 ) given
T 1
Comparison of frequency ratio (v/za) for the constants a = 0·02927, b = 0·0, g = 5·0239 in equation (1)
(h = 0·1 mm)
Perturbation technique
ZXXXXXXXCXXXXXXXV
Amplitude Exact Three-term series Six-term series Harmonic balance method
ratio, integration ZXXXCXXXV ZXXXCXXXV ZXXXXXXXXCXXXXXXXXV
xmax /h [13] ea/h v/za ea/h v/za Two-term Three-term Four-term
0·2 1·0254 0·1996 1·0253 0·1996 1·0254 1·0254 1·0254 1·0254
0·4 1·0975 0·3967 1·0966 0·3972 1·0976 1·0982 1·0982 1·0975
0·6 1·2072 0·5890 1·2028 0·5930 1·2088 1·2097 1·2097 1·2073
0·8 1·3447 0·7750 1·3316 0·7913 1·3539 1·3505 1·3505 1·3449
1·0 1·5023 0·9535 1·4732 1·0045 1·5392 1·5124 1·5124 1·5028

T 2
Comparison of frequency ratio (v/za) for the constants a = 0·5908, b = −1·1044, g = 5·295
(h = 0·5 mm)
Perturbation technique
   

ZXXXXXXXCXXXXXXXV
Amplitude Exact Three-term series Six-term series Harmonic balance method
ratio, integration ZXXXCXXXV ZXXXCXXXV ZXXXXXXXXCXXXXXXXXV
xmax /h [13] ea/h v/za ea/h v/za Two-term Three-term Four-term
−2·0 2·8610 −2·5616 2·6925 † — 2·9160 2·9129 2·8656
−1·5 2·2641 −2·0492 2·2359 † — 2·3053 2·3004 2·2674
−1·0 1·6918 −1·3512 1·6549 † — 1·7202 1·7122 1·6939
−0·5 1·1920 −0·5905 1·1541 −0·5996 1·3024 1·2056 1·1967 1·1926
0·001 1·0 0·001 1·0 0·001 1·0 1·0 1·0 1·0
0·5 1·1239 0·4341 1·0860 0·4446 1·1381 1·1160 1·1246 1·1237
1·0 1·4929 0·7731 1·2527 0·7534 1·5565 1·4930 1·5022 1·4926
1·5 1·9976 1·0523 1·4334 0·9034 1·8975 2·0162 2·0222 1·9985
2·0 2·5616 1·2910 1·6085 0·9923 2·143 2·5987 2·6025 2·5641
† The real value does not exist.
181
182

T 3
Comparison of frequency ratio (v/za) for the constants a = 1·0688, b = 5·0716, g = 10·621 (h = 0·5 mm)
Perturbation technique
ZXXXXXXXCXXXXXXXV
Amplitude Exact Three-term series Six-term series Harmonic balance method
ratio, integration ZXXXCXXXV ZXXXCXXXV ZXXXXXXXXCXXXXXXXXV
xmax /h [13] ea/h v/za ea/h v/za Two-term Three-term Four-term
−2·0 2·3315 −1·0038 † † — 2·3695 2·3757 2·3341
−1·5 1·7106 −0·8345 † † — 1·7260 1·7372 1·7109
−1·0 1·1641 −0·6330 † † — 1·1506 1·1717 1·1621
−0·5 0·8928 −0·3770 0·7722 −0·4160 0·8771 0·8724 0·8883 0·8925
0·001 1·0 0·001 1·0 0·001 1·0 1·0 1·0 1·0
   

0·5 1·1038 2·3272 † 0·4846 0·9018 1·1537 1·1275 1·1093


1·0 1·7876 2·8967 † 0·6020 1·0558 1·8368 1·8261 1·7926
1·5 2·4576 3·2039 † 0·6622 1·1957 2·5161 2·5104 2·4631
2·0 3·1241 3·4312 † 0·7039 1·3159 3·1952 3·1917 3·1306
† The real value does not exist.
    183
2
in equation (21) by neglecting terms higher than O(j ). From the above observation, it
2
is stated in reference [1] that, to obtain a consistent solution by using the method of
harmonic balance, ‘‘one needs either to know a great deal about the solution a priori or
to carry enough terms in the solution and check the order of the coefficients of all the
neglected harmonics’’, and also one is cautioned ‘‘not to use this technique’’. These
conclusions are based on the assumption that the perturbation method provides the exact
periodic solution of equation (1).
In order to verify the adequacy of the perturbation method as well as the method of
harmonic balance, the solutions of equation (1) with the initial conditions (2) by these
methods have been obtained and compared with existing results obtained by exact
integration [13]. In Tables 1, 2 and 3 are shown comparisons of the frequency parameter
v for different values of the constants a, b, and g and the amplitude xmax .
Concerning Table 1, for the case b = 0, the three-term and six-term perturbation
solutions, and the two-term, three-term and four-term harmonic balance method solutions
are found to be in good agreement with the exact integration results. In fact, the six-term
perturbation method solution and the four-term harmonic balance method solution are
very close to the exact solution. The two-term harmonic balance method results for an
amplitude ratio of 0·8 or higher are better than the six-term perturbation technique results.
When b % 0, the squared frequency in the three-term perturbation solution becomes
negative for some specified amplitudes and the frequency becomes purely imaginary and
the method fails for those amplitudes (xmax ). Concerning Tables 2 and 3, the six-term
perturbation method results do not match well with the exact solution. The three-term
harmonic balance method results are better than the two-term results for small amplitude
ratios. For all the cases examined here, the solution obtained by taking four terms in the
method of harmonic balance is found to be very close to the exact solution. The relative
error between the two-term method of harmonic balance solution and the exact solution
is found to be within 6%. As the number of terms in the assumed solution (12) of equation
(1) increases, the frequency parameter will be close to the exact. However, for all practical
purposes, the two-term method of harmonic balance solution yields reasonably accurate
results. The method of harmonic balance is a fast and efficient calculational technique for
determining approximate analytic periodic solutions.
Furthermore, based on this study, it can be concluded that the Lindstedt–Poincaré
perturbation technique [1] may be applicable for small amplitudes. However, it is not
immediately obvious whether the specified amplitude is small or large, and the periodic
solution of the problem depends on the constants a, b and g in equation (1) and on the
initial conditions (2). Many terms in the perturbation series are required to achieve a
desired degree of accuracy.


1. A. H. N and D. T. M 1979 Nonlinear Oscillations. New York: Wiley-Interscience.
2. C. Y. C 1980 Nonlinear Analysis of Plates. New York: McGraw-Hill.
3. M. S 1987 Applied Mechanics Reviews 40, 1553–1561. Nonlinear vibration
analysis of plates.
4. R. E. M 1981 An Introduction to Nonlinear Oscillations. Cambridge: Cambridge
University Press.
5. R. E. M 1984 Journal of Sound and Vibration 94, 456–460. Comments on the method of
harmonic balance.
6. A. S. A and K. H 1984 Journal of Sound and Vibration 95, 525–530. An intrinsic
method of harmonic analysis for nonlinear oscillations (a perturbation technique).
7. K. H and R. L 1991 International Journal of Nonlinear Mechanics 26, 727–740. An
intrinsic multiple-scale harmonic balance method for nonlinear vibration and bifurcation
problems.
184    
8. H. P. W. G 1992 Journal of Sound and Vibration 152, 189–191. On the harmonic balance
method for mixed-parity non-linear oscillators.
9. A. V R and B. N R 1994 Journal of Sound and Vibration 170,
571–576. Some remarks on the harmonic balance method for mixed-parity nonlinear oscillations.
10. R. C and B. B. R 1975 Journal of Sound and Vibration 40, 393–408. Large deflection
vibration of angle-ply laminated plates.
11. C. Y. C 1982 Fibre Science and Technology 17, 123–131. Large amplitude vibrations of
laminated rectangular plates.
12. R. E. M 1993 Journal of Sound and Vibration 167, 564–567. Construction of a
perturbation solution to a mixed-parity system that satisfies the correct initial conditions.
13. M. S. S, A. V R, S. R. R. P and B. N R 1992 Journal
of Sound and Vibration 159, 540–545. Large amplitude vibration of laminated hybrid composite
plates.

Anda mungkin juga menyukai