C407X 13
C407X 13
C407X 13
where t denotes the time, the dot denotes derivative with respect to t, and
f (u, u̇, ü) is a known function of u, u̇, and ü. Unlike perturbation techniques,
it is unnecessary to assume the existence of any small/large quantities in the
above equation. Here, we show that the homotopy analysis method can also
be applied to gain limit cycles of multidimensional dynamical systems.
As an example, let us consider a two-dimensional nonlinear dynamical
system governed by (see Kahn [103])
where the dot denotes differentiation with respect to t, µ and are physical
parameters, x and w are two unknown functions. Physically, a limit cycle is
independent of initial conditions. Let T and α = max[x(t)] denote the period
and the maximum value of x(t) of the limit cycle, respectively. Without loss
of generality, we can define t = 0 so that
Define
T
1
δ= w(t)dt (13.5)
T 0
and let
ω = T /2π
denote the frequency of x(t) of the limit cycle. Under the transformations
which provides us with the condition for v(τ ). Note that α, δ, and ω are
unknown.
and
+∞
v(τ ) = [cn cos(nτ ) + dn sin(nτ )] , (13.12)
n=1
where an , bn , cn , and dn are coefficients. The above expressions provide the
so-called rules of solution expression for u(τ ) and v(τ ), respectively.
Under the above rules of solution expressions and from the initial condi-
tions (13.9) and (13.10), it is convenient to choose
u0 (τ ) = cos τ, v0 (τ ) = 0 (13.13)
as the initial guesses of u(τ ) and v(τ ). Here, v0 (τ ) = 0 is chosen because of the
lack of information about v(τ ), especially the relationship between u(τ ) and
v(τ ). Let α0 , δ0 , and ω0 denote the initial guesses of α, δ, and ω, respectively.
Under the rules of solution expression denoted by (13.11) and (13.12) and
from Equations (13.7) and (13.8), we choose the auxiliary linear operators
∂2f
Lu f = +f (13.14)
∂τ 2
and
∂f
Lv f = (13.15)
∂τ
and
(1 − q)Lu [U (τ ; q) − u0 (τ )]
= q u Hu (τ ) Nu [U (τ ; q), V (τ ; q), A(q), ∆(q), Ω(q)] , (13.19)
(1 − q)Lv [V (τ ; q) − v0 (τ )]
= q v Hv (τ ) Nv [U (τ ; q), V (τ ; q), A(q), ∆(q), Ω(q)] , (13.20)
and
1 dn A(q)
αn = , (13.27)
n! d qn q=0
1 dn ∆(q)
δn = , (13.28)
n! d qn q=0
1 dn Ω(q)
ωn = (13.29)
n! d qn q=0
exist for n ≥ 1. Then, using Taylor’s theorem and (13.22), we have the power
series of q in the forms:
+∞
U (τ ; q) = u0 (τ ) + un (τ ) q n , (13.30)
n=1
+∞
V (τ ; q) = v0 (τ ) + vn (τ ) q n , (13.31)
n=1
+∞
A(q) = α0 + αn q n , (13.32)
n=1
+∞
∆(q) = δ0 + δn q n , (13.33)
n=1
+∞
Ω(q) = ω0 + ωn q n . (13.34)
n=1
and
+∞
ω = ω0 + ωn . (13.39)
n=1
Lu [un (τ ) − χn un−1 (τ )]
= u Hu (τ ) Rnu (un−1 , vn−1 , αn−1 , δn−1 , ωn−1 ), (13.43)
Lv [vn (τ ) − χn vn−1 (τ )]
= v Hv (τ ) Rnv (un−1 , vn−1 , αn−1 , δn−1 , ωn−1 ), (13.44)
and
Hu (τ ) = Hv (τ ) = 1. (13.50)
and
R1v = c1,0 + c1,1 cos(2τ ) + c1,2 cos(4τ ), (13.52)
α02 δ0 3α4 µ
ω0 − = 0, ω02 − 1 = 0, δ02 − 0 = 0, (13.53)
4 8
whose solutions are
2
α0 = √
8
, δ0 = 4
6µ, ω0 = 1. (13.54)
6µ
Now the problem is solved in accordance with the rules of solution expression
denoted by (13.11) and (13.12). We now have
and
R1v = c1,1 cos(2τ ) + c1,2 cos(4τ ).
Solving the first-order deformation equations (13.43) and (13.44) under the
conditions noted in (13.45), we have
%&
u1 (τ ) = − u (3 sin τ − sin 3τ ) (13.55)
8
and
µ 1
v1 (τ ) = − 4 v sin 2τ + sin 4τ . (13.56)
6 8
Similarly, first solving a set of linear algebraic equations
% &
2 3u + 4 176µ v − 48 ω1 = 0, (13.57)
(246µ3 )1/8 α1 + (46)3/4 δ1 = 0, (13.58)
(126µ)3/8 α1 − δ1 = 0 (13.59)
and v
Mn
v(τ ) = [cn,k cos(2kτ ) + dn,k sin(2kτ )] ,
k=1
where Mnu and Mnv are integers dependent upon the order n of approximation.
Thus, the frequency of the motion w(t) is twice that of x(t).
THEOREM 13.1
If the solution series (13.35) to (13.39) are convergent, where un (τ ) and vn (τ )
are governed by Equations (13.43) to (13.45) under the definitions (13.46) to
(13.49), and (2.42), they must be the solution of Equations (13.7) to (13.10).
Proof: If the solution series (13.35) and (13.36) are convergent, then
lim um (τ ) = 0, lim vm (τ ) = 0.
m→+∞ m→+∞
From Equation (13.43) and using the definitions (2.42) and (13.14), we then
have
+∞
u Hu (τ ) Rnu (un−1 , vn−1 , αn−1 , δn−1 , ωn−1 )
n=1
+∞
= Lu [un (τ ) − χn un−1 (τ )]
n=1
m
= lim Lu [un (τ ) − χn un−1 (τ )]
m→+∞
n=1
= lim Lu [um (τ )]
m→+∞
= Lu lim um (τ )
m→+∞
= 0,
Substituting (13.46) and (13.47) into the above expressions and simplifying
them, we have, due to the convergence of the series (13.37) to (13.39), that
+∞ 2 ⎡ ⎤
2 +∞
+∞
d ⎣ ⎦
ωi u j (τ ) + uj (τ )
i=0
dτ 2 j=0 j=0
+∞ ⎡ ⎤
+∞
d ⎣
= ωi uj (τ )⎦
i=0
dτ j=0
⎧ ⎫
⎪ +∞ 2 ⎛+∞ ⎞2 +∞ ⎪
⎨ ⎬
× 1− αi ⎝ uj ⎠ [δk + vk (τ )]
⎪
⎩ ⎪
⎭
i=0 j=0 k=0
and
+∞ ⎡ ⎤
+∞
d ⎣
ωi vj (τ )⎦
i=0
dτ j=0
⎧ ⎫
⎪+∞ 2 +∞ 4 ⎛+∞ ⎞4 ⎪
⎨ +∞
⎬
= − δk + vk (τ ) −µ αi ⎝ uj ⎠ .
⎪
⎩ ⎪
⎭
k=0 k=0 i=0 j=0
u = v =
TABLE 13.2
The [m, m] homotopy-Padé approximations of ω, α, and δ
when = 1/5, µ = 3 by means of u = v = −3/4 and
Hu (τ ) = Hv (τ ) = 1.
[m, m] ω α δ
[1, 1] 0.96889 1.39354 2.05977
[2, 2] 0.96977 1.41413 2.08345
[3, 3] 0.96968 1.41414 2.08735
[4, 4] 0.96968 1.41399 2.07015
[5, 5] 0.96968 1.41398 2.07016
[6, 6] 0.96968 1.41399 2.07015
[7, 7] 0.96968 1.41399 2.07015
2.5
d
2
w,a,d
1.5 a
1
w
when e = 1/5, m = 3
0.5
0
-2.5 -2 -1.5 -1 -0.5 0
-h
FIGURE 13.1
The -curves of ω, α, and δ at the 10th order of approximation when = 1/5
and µ = 3 by means of Hu (τ ) = Hv (τ ) = 1. Dashed line: δ; dash-dotted line:
α; solid line: ω.
1.5
0.5
x ’ (t)
-0.5
-1
-1.5
-2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
x(t)
FIGURE 13.2
x − ẋ plane projection of the limit cycle when = 1/5 and µ = 3. Solid line:
fifth-order approximation by means of u = v = −3/4 and Hu (τ ) = Hv (τ ) =
1; symbols: numerical result.
2.5
w(t)
1.5
1
-1.5 -1 -0.5 0 0.5 1 1.5
x(t)
FIGURE 13.3
x − w plane projection of the limit cycle when = 1/5 and µ = 3. Solid line:
fifth-order approximation by means of u = v = −3/4 and Hu (τ ) = Hv (τ ) =
1; symbols: numerical result.
2.5
w(t)
1.5
1
-1.5 -1 -0.5 0 0.5 1 1.5
x ’ (t)
FIGURE 13.4
ẋ − w plane projection of the limit cycle when = 1/5 and µ = 3. Solid line:
fifth-order approximation by means of u = v = −3/4 and Hu (τ ) = Hv (τ ) =
1; symbols: numerical result.
2.5
a
2
w,a,d
1.5
1
w
0.5
when e = 3/4, m = 1/6
0
-1.5 -1 -0.5 0
-h
FIGURE 13.5
The -curves of ω, α, and δ at the 10th order of approximation when = 3/4
and µ = 1/6 by means of Hu (τ ) = Hv (τ ) = 1. Dashed line: δ; dash-dotted
line: α; solid line: ω.
1
x’(t)
-1
-2
-3
-3 -2 -1 0 1 2 3
x(t)
FIGURE 13.6
x − ẋ plane projection of the limit cycle when = 3/4 and µ = 1/6. Solid
line: 20th-order approximation by means of u = v = −3/4 and Hu (τ ) =
Hv (τ ) = 1; Symbols: numerical result.
1.75
1.5
w(t)
1.25
0.75
0.5
-3 -2 -1 0 1 2 3
x(t)
FIGURE 13.7
x − w plane projection of the limit cycle when = 3/4 and µ = 1/6. Solid
line: 20th-order approximation by means of u = v = −3/4 and Hu (τ ) =
Hv (τ ) = 1; symbols: numerical result.
1.75
1.5
w(t)
1.25
0.75
0.5
-3 -2 -1 0 1 2 3
x’(t)
FIGURE 13.8
ẋ − w plane projection of the limit cycle when = 3/4 and µ = 1/6. Solid
line: 20th-order approximation by means of u = v = −3/4 and Hu (τ ) =
Hv (τ ) = 1; symbols: numerical result.