ARTICLE IN PRESS

JOURNAL OF
SOUND AND
VIBRATION
Journal of Sound and Vibration 269 (2004) 409412
www.elsevier.com/locate/jsvi

Letter to the Editor

A classical perturbation technique which is valid for

large parameters
H. Hu
Department of Mathematics and Physics, Xiangtan Polytechnic University, Xiangtan 411201, Hunan,
Peoples Republic of China
Received 8 February 2003; accepted 20 March 2003

There are various perturbation techniques for constructing analytical approximations to the
oscillatory solutions of second order, non-linear differential equations [1,2]. But many of them
apply to weakly non-linear cases only. To overcome the limitations, many novel techniques have
been proposed in recent years. For example, Cheung et al. [3] proposed a modied Lindstedt
Poincare method, and Lim et al. [4] presented a modied Mickens procedure for certain
non-linear oscillators. Recently, He [5] proposed a perturbation technique, which is valid for large
parameters. But unfortunately, o1 and B in Ref. [5] are the functions of Z which is unknown.
The main purpose of this paper is to point out that there already exists a classical perturbation
technique, which is valid for large parameters. In this paper, the following well-known Dufng
equation is considered:
x00 o20 x ex3 0; x0 A; x0 0 0; 1

where e is a parameter.
The solution of Eq. (1) is assumed in the form
xt x0 t ex1 t e2 x2 t ?: 2

The fundamental frequency o2 is given by [69]

o2 o20 eo1 e2 o2 ?; 3

where, at this point, the oi are unknown constants. This expansion has been also employed by
Cheung et al. [3].
Substituting Eqs. (2) and (3) into Eq. (1) gives
x000 ex001 e2 x002 ? o2  eo1  e2 o2  ?x0 ex1 e2 x2 ?
x0 ex1 e2 x2 ?3 0: 4

E-mail address: [email protected] (H. Hu).

0022-460X/03/$ - see front matter r 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0022-460X(03)00318-3
 ARTICLE IN PRESS

410 H. Hu / Journal of Sound and Vibration 269 (2004) 409412

This equation is satised by setting the coefcients of the powers of e equal to zero, resulting in
x000 o2 x0 0; 5

x001 o2 x1 o1 x0  x30 ; 6

x002 o2 x2 o2 x0 o1 x1  3x20 x1 : 7
The initial conditions are replaced by
x0 0 A; x00 0 0; xi 0 0; x0i 0 0; iX1: 8
The solution to Eq. (5) is
x0 A cos ot: 9
Using the normal classical perturbation technique [1,2] and taking into account the initial
conditions given by Eq. (8), we can easily obtain
A3
o1 34A2 ; x1 cos 3ot  cos ot; 10
32o2

3A4 A5
o2  ; x2 cos 5ot  cos ot: 11
128o2 1024o4
Therefore, the second approximate solution to Eq. (1) becomes
eA3 e2 A5
x A cos ot cos 3ot  cos ot cos 5ot  cos ot 12
32o2 1024o4
with
3e2 A4
o2 o20 34 eA2  : 13
128o2
This equation can be rewritten as


o4  o20 34 eA2 o2 128
3 2 4
e A 0: 14
Solving for o2 gives

 q

2 3
o2 12 o20 34 eA2 712 o20 34 eA2 32 e2 A4 : 15

Obviously, if e 0; we should have o2 o20 : Therefore, we get

r
q
o 14 8o20 6eA2 64o20 96o20 eA2 30e2 A4 : 16

To compare the present results with the results in the literature, we take o20 1: Then, Eq. (16)
becomes
q
p
1
o op 4 8 6eA 64 96eA2 30e2 A4 :
2 17
 ARTICLE IN PRESS

H. Hu / Journal of Sound and Vibration 269 (2004) 409412 411

The exact frequency of the periodic motion of the Dufng equation is given by [10]
p Z !1
p=2
p 1 eA2 dy eA2
oe p ; m 18
2 0 1  m2 sin2 y 21 eA2

and the corresponding approximate frequencies obtained by Lim et al. [4] and He [5] are
q
p
1
oL 8 32 25eA2 1024 1472eA2 433e2 A4 19

and
q
p
1
oH 8 6eA 64 96eA2 42e2 A4 ;
4
2 20

respectively.
For comparison, the exact frequency obtained by integrating Eq. (18) and the approximate
frequencies op ; oL and oH computed by Eqs. (17), (19) and (20), respectively, are listed in Table 1.
Table 1 indicates that formula (17) is more accurate than formulas (19) and (20), and can give
excellent approximate frequencies for both small and large parameters and oscillation amplitude.
We also have
q
p Z
op 6 30 p=2 dy
lim p
eA2 -N oe 2p 0 1  0:5 sin2 y
q
p
6 30
1:8541 0:9997: 21
2p
In summary, a classical perturbation technique has been used to solve the Dufng equation. We
nd that the parameter e does not need to be small. For any value of e; the maximal relative error
of the second approximate frequency with respect to the exact solution is less than 0.03%.

Table 1
Comparison of approximate frequencies with the corresponding exact frequency for the Dufng equation
eA2 oe Eq. (18) [10] oL Eq. (19) [4] oH Eq. (20) [5] op Eq. (17) present
0.2 1.07200 1.07200 1.07276 1.07200
0.4 1.13891 1.13889 1.14144 1.13891
0.6 1.20173 1.20170 1.20656 1.20173
0.8 1.26118 1.26112 1.26859 1.26118
1 1.31778 1.31767 1.32789 1.31776
2 1.56911 1.56873 1.59278 1.56905
5 2.15042 2.14912 2.20688 2.15018
10 2.86664 2.86408 2.96097 2.86613
100 8.53359 8.52220 8.88639 8.53110
1000 26.8107 26.7734 27.9465 26.8025
10000 84.7245 84.6088 88.3257 84.7013
 ARTICLE IN PRESS

412 H. Hu / Journal of Sound and Vibration 269 (2004) 409412

Why does this classical technique work for large parameters? It is an interesting question and
needs further study. The normal classical perturbation method [1,2,10] uses the expansion
o o0 eo01 e2 o02 ? 22
instead of expansion (3), which is the important difference of the normal classical perturbation
method with respect to this innovative classical perturbation method. For example, we have the
rst approximate frequency (for o0 1)
q
o 1 38 eA2 : 23
But the normal classical perturbation method [1,10] gives
o 1 38 eA2 : 24
Formula (23) can give good approximate frequencies for both small and large parameters [4,5].
q
But for large parameters, formula (24) is not valid. 1 34 eA2 E1 38eA2 only when e is small.
Maybe using expansion (22) is the reason why the normal classical perturbation method is not
able to provide accurate result when the parameter is large.

Acknowledgements

The work was supported by the Natural Science Foundation of Hunan (01JJY3002) and
Xiangtan Polytechnic University Scientic Research Fund (E20103).

References

[1] R.E. Mickens, Oscillations in Planar Dynamic Systems, World Scientic, Singapore, 1996.
[2] A.H. Nayfeh, D.T. Mook, Nonlinear Oscillations, Wiley, New York, 1979.
[3] Y.K. Cheung, S.H. Chen, S.L. Lau, Amodied LindstedtPoincare method for certain strongly non-linear
oscillators, International Journal of Non-linear Mechanics 26 (1991) 367378.
[4] C.W. Lim, B.S. Wu, A modied Mickens procedure for certain non-linear oscillators, Journal of Sound and
Vibration 257 (2002) 202206; doi:10.1006/jsvi.2001.4233.
[5] J.H. He, A new perturbation technique which is also valid for large parameters, Journal of Sound and Vibration
229 (2000) 12571263; doi:10.1006/jsvi.1999.2509.
[6] P. Hagedorn, Non-linear Oscillations, Clarendon Press, Oxford, 1981 (see Section 1.1.1).
[7] S.S. Rao, Mechanical Vibrations, 2nd Edition, Addison-Wesley, Reading, 1990 (see Section 13. 4).
[8] W.T. Thomson, Theory of Vibration with Applications, 2nd Edition, Prentice-Hall, Englewood Cliffs, NJ, 1981
(see Section 12. 6).
[9] F.S. Tse, I.E. Morse, R.T. Hinkle, Mechanical Vibrations: Theory and Applications, 2nd Edition, Allyn and
Bacon, Boston, 1978 (see Section 8-8).
[10] A.H. Nayfeh, Introduction to Perturbation Techniques, Wiley, New York, 1981.