NPTEL – Mechanical Engineering – Nonlinear Vibration

Module 3 Lecture 4

The method of multiple scales

In method of multiple scales, the original time is written in terms of different time scales which
are considered to be multiple independent variables, or scales, instead of a single variable. Here,
the new independent variables ( Tn , n = 1,2, ) of time are written using the book-keeping
parameter ε as
Tn = ε nt (3.4.1)

Hence, the derivatives with respect to t can be written in terms of the partial derivatives with
respect to the Tn as follows.
d dT0 ∂ dT ∂
= + 1 +  = D0 + ε D1 +  (3.4.2)
dt dt ∂T0 dt ∂T1
d2
D02 + 2ε D0 D1 + ε 2 ( D12 + 2 D0 D2 ) + ....
= (3.4.3)
dt 2

Let us apply this method to the Duffing equation with quadratic and cubic nonlinearities

Example 3.4.1:
x + ω02 x + εα 2 x 2 + εα 3 x3 =
 0 (3.4.4)

Similar to previous method here, one may assume that the solution of (3.4.4) can be represented
by an expansion having the form

x(t ; ε ) = εx1 (T0 , T1 , T2 ,......) + ε 2 x 2 (T0 , T1 , T2 ,......) + ε 3 x 3 (T0 , T1 , T2 ,......) + ..... (3.4.5)

We note that the number of independent time scales needed depends on the order to which the
( )
expansion is carried out. For example for O ε 3 , one may consider T0 , T1 , and T2 . Substituting
(3.4.3) and (3.4.5) into (3.4.4) and equating the coefficients of ε , ε 2 , and ε 3 to zero, one obtains
the following sets of equations.

Order of ε 1
D02 x1 + ω02 x1 =
0 (3.4.6)
Order of ε 2

D02 x2 + ω02 x2 =
−2 D0 D1 x1 − α 2 x12 (3.4.7)
Order of ε 3

Joint initiative of IITs and IISc – Funded by MHRD Page 12 of 67

 NPTEL – Mechanical Engineering – Nonlinear Vibration

D02 x3 + ω02 x3 =
−2 D0 D1 x2 − D12 x1 − 2 D0 D2 x1 − 2α 2 x1 x2 − α 3 x13 (3.4.8)
The solution of (3.4.6) can be written as
=x1 A(T1 , T2 ) exp(iω0T0 ) + A exp(−iω0T0 ) . (3.4.9)
Here A is an unknown complex function and A is the complex conjugate of A . Substituting
(3.4.9) into (3.4.7) leads to
− 2iω0 D1 A exp(iω0T0 ) − α 2  A2 exp(2iω0T0 ) + AA + cc
D02 x2 + ω02 x2 = (3.4.10)
 
Secular term
Here cc denotes the complex conjugate of the preceding terms. The particular solution of
(3.4.10) has a secular term containing the factor T0 exp ( iω0T0 ) . To have a bounded solution this
term has to be eliminated. Hence one can obtain
dA
D=
1A = 0 (3.4.11)
dT1
Therefore A must be independent of T1 . With D1 A = 0 the particular solution of (3.4.10) can be
written as
α 2 A2 α
= x2 exp(2iω0T0 ) − 22 AA + cc (3.4.12)
3ω0 2
ω0
Substituting the expression for x1 and x2 from equation (3.4.9) and (3.4.12) into (3.4.8) and
recalling that D1 A = 0 we obtain
 10α 22 − 9α 3ω02 2 
−  2iω0 D2 A −
D02 x3 + ω02 x3 = A A exp ( iω0T0 )
 3ω02 
 
Secular Term (3.4.13)
3α 3ω + 2α 3
2 2
− 0
A exp(3iω0T0 ) + cc
2

3ω02
To eliminate the secular terms from x3 , we must put
9α 3ω02 − 10α 22 2
2iω0 D2 A + A A= 0 (3.4.14)
3ω02
To solve Eq. (3.4.14), it is convenient to write A in the polar form as
1
A = a exp(i β ) (3.4.15)
2
where a and β are real function of T2 . Substituting (3.4.15) into (3.4.14) and separating the
result
into real and imaginary parts, we obtain
10α 22 − 9α 3ω02 3
ω a′ = 0 and ω0 aβ ′ + a =0 (3.4.16)
24ω02
where the prime denotes the derivative with respect to T2 . As a′ = 0 , a is a constant and
10α 22 − 9α 3ω02 3 9α 3ω02 − 10α 22 2
β′ = − = a or β a T2 + β 0 (3.4.17)
24ω02ω0 a 24ω03

Joint initiative of IITs and IISc – Funded by MHRD Page 13 of 67

 NPTEL – Mechanical Engineering – Nonlinear Vibration

Here β 0 is a constant. Now using T2 = ε 2t from (3.4.15) we find that

1  9α 3ω02 − 10α 22 2 2 
A a exp i ε a t + i β  (3.4.18)
24ω03
0
2  
Substituting Eq. (3.4.18) in the expressions for x1 and x2 in Eqs. (3.4.9), (3.4.12) and (3.4.5),
one obtains
ε 2 a 2α 2  1 
=x ε a cos(ωt + β 0 ) − 1 − cos(2ωt + 2 β 0 )  + O(ε 3 ) (3.4.19)
2ω0  3
2

 9α ω 2 − 10α 2 
Here ω = ω0 1 + 3 0 4 2 ε 2 a 2  + O(ε 3 ) (3.4.20)
 24ω0 
This solution is in good agreement with the solution obtained using the Lindstedt-Poincare’
procedure. The method of multiple scales though a little more involved, has advantage over the
Lindstedt-Poincare method, for example it can treat damped systems conveniently ( Nayfeh and
Mook 1979).

Example 3.4.2: Find the expression for the frequency-response curve for a nonconservative
system using method of multiple scales.

Solution:
Consider the governing equation of motion of a nonconservative system which can be given by

. (3.4.21)
Following standard procedure of method of multiple scales one may write

u ( t; ε ) =
u0 + ε u1 (T0 ,T1,T2 ,......) + ε 2u2 (T0 ,T1,T2 ,......) + ε 3u3 (T0 ,T1,T2 ,......) + ..... (3.4.22)

Substituting (3.4.3) and (3.4.22) into (3.4.21) and equating the coefficients of ε 0 , ε 1 and ε 2 to
zero,
one obtains the following sets of equations.

(3.4.23)
0 1 −2 D0 D1u0 + f ( u0 , D0u0 )
D u +ω u =
2 2
0 1 (3.4.24)
D02u2 + ω02u2 =
−2 D0 D1 x2 − D12 x1 − 2 D0 D2 x1 − 2α 2 x1 x2 − α 3 x13 (3.4.25)
D u + ω u F (u0 , u1 , un −1 ) for n ≥ 2
= 2
0 n
2
0 n (3.4.26)

The solution of Eq. (3.4.23) can be given by

Joint initiative of IITs and IISc – Funded by MHRD Page 14 of 67
 NPTEL – Mechanical Engineering – Nonlinear Vibration

=u0 A(T1 , T2 ) exp ( iω0T0 ) + A exp ( −iω0T0 ) (3.4.27)

Substituting Eq. (3.4.27) in Eq. (3.4.24) following equation is obtained.
−2iω0 D1 A exp ( iω0T0 ) + 2iω0 D1 A exp ( −iω0T0 ) +
D02u1 + ω02u1 =
(3.4.28)
f  A exp ( iω0T0 ) + A exp ( −iω0T0 ) , 2iω0 ( A exp ( iω0T0 ) − A exp ( −iω0T0 ) ) 
One may use Fourier series to write the forcing function as follows.
∞
f = ∑ f ( A, A ) exp ( inω T )
n = −∞
n 0 0 (3.4.29)
2π
ω0
ω0
=
where, f n ( A, A ) ∫ f exp ( −inω0T0 )dT0 (3.4.30)
2π 0
Hence to eliminate secular term from Eq. (3.4.28) one may write

2π
ω0
1
=
2iD1 A ∫ f exp ( −iω0T0 )dT0 (3.4.31)
2π 0

For a first order approximation, one may consider A to be a function of T1 only and can write
A in its polar form as
A (T1 ) = a (T1 ) exp ( i β (T1 ) )
1
(3.4.32)
2
Substituting (3.4. 32) in (3.4.31) one can write
2π
ω0
1  1
2iD1  a (T1 ) exp
= ( i β (T1 ) )  ∫ f exp ( −iω0T0 )dT0 (3.4.33)
2  2π 0
2π
ω0
dβ
exp ( i β (T1 ) ) − a exp ( i β (T1 ) ) =∫ f exp ( −iω0T0 )dT0
da 1
Or, i (3.4.34)
dT1 dT1 2π 0
2π
ω0
da dβ 1
=
Or, i −a ∫ f exp ( −iω0T0 )dT0
dT1 dT1 2π exp ( i β (T1 ) ) 0
2π
ω0
f exp ( −iω0T0 ) exp ( −i β (T1 ) )dT0
1
=
2π ∫0
(3.4.35)
2π 2π
ω0 ω0

=
1
2π ∫ (
f exp −i (ω0T0 + β (T1 ) )=
dT0
1
2π
) ∫ f exp ( −iφ )dT0
0 0
2π
ω0
1
f ( cos φ − i sin φ )dT0 where=
φ ω0T0 + β (T1 )
2π ∫0
=

Separating the real and imaginary parts one may write

Joint initiative of IITs and IISc – Funded by MHRD Page 15 of 67
NPTEL – Mechanical Engineering – Nonlinear Vibration

2π
da 1
sin φ f ( cos φ , −aω0 sin φ )dφ
2πω0 ∫0
=
− (3.4.36)
dT1

2π
dβ 1
cos φ f ( cos φ , −aω0 sin φ )dφ
2πω0 a ∫0
= − (3.4.37)
dT1
The first order approximation solution can be written as

u ( t; ε ) =u0 = A(T1,T2 )exp ( iω0T0 ) + A exp ( −iω0T0 )

1 1
= a exp ( i β ) exp ( iω0T0 ) + a exp ( −i β ) exp ( −iω0T0 )
2 2
= a ( exp ( iω0T0 + i β ) + exp ( −iω0T0 − i β ) )
1 (3.4.38)
2
= a ( exp ( iφ ) + exp ( −iφ=) ) a cos=
1
φ a cos (ω0t + β ) + O ( ε )
2
Exercise Problems:
1. Derive the frequency-amplitude relation for the following systems using multiple scales
(a) u + ω02u + 2εµ u =0
(b) u + ω02u + εα 2u 2= ε f cos Ωt
(c) u + ω02u + 2εµ u + εα u 3= ε f cos Ωt
(d) u + ω02u + 2εµ u + εα 2u 2 + εα 3u 3= ε f cos Ωt
(e) u + ω02u + 2εζ u + εα u 3 + ε f cos Ωtu =
0

Joint initiative of IITs and IISc – Funded by MHRD Page 16 of 67

 NPTEL – Mechanical Engineering – Nonlinear Vibration

Module 3 Lecture 5

METHOD OF MULTIPLE SCALES APPLIED TO FORCED VIBRATION

In this lecture the method of multiple scales is applied to a forced vibration system. One may
follow similar procedure as in the previous lecture. But in this case additional secular terms will
arise which will give different resonance conditions. In the following example the primary
resonance condition for the forced Duffing equation is illustrated. It may be noted that unlike
linear system in case of nonlinear system multiple equilibrium solution will arise.

Example 3.5.1: Find the frequency-amplitude relation for primary resonance condition for the
forced Duffing equation.
u + ω02u + 2εµ u + εα u=
3
ε K cos Ωt (3.5.1)
Solution

For primary resonance condition, the frequency of external excitation Ω should be nearly equal
to that of natural frequency ω0 of the system. Hence, to show the nearness of Ω to ω0 , one may
use a detuning parameter σ , and by using book-keeping parameter it can be written that
Ω= ω0 + εσ (3.5.2)
Now expanding u using the book-keeping parameter and different time scales one may write
u ( t ; ε ) =u0 (T0 , T1 ) + ε u1 (T0 , T1 ) + ....... (3.5.3)
Now substituting Eqs.(3.4.3) and (3.5.3) in Eq. (3.5.1) and separating the like power of ε ,
following equations are obtained.
D02u0 + ω02u0 =
0 (3.5.4)

(3.5.5)
The solution of Eq. (3.5.4) can be given by
u0 A (T1 , T2 ) exp ( iω0T0 ) + A (T1 , T2 ) exp ( −iω0T0 ) .
=
(3.5.6)
Substituting (3.5.6) in Eq. (3.5.5) one obtains
D02u1 + ω02u1 =−  2iω0 ( D1 A exp ( iω0T0 ) + µ A exp ( iω0T0 ) ) + 3α A2 A exp ( iω0T0 ) 
1
−α A3 exp ( 3iω0T0 ) + f exp i (ω0T0 + σ T1 )  + cc (3.5.7)
2
Or,
D02u1 + ω02u1 −  2iω0 ( A' + µ A ) + 3α A2 A  exp ( iω0T0 ) − α A3 exp ( 3iω0T0 )
 
Secular term

1
+ f exp i (ω0T0 + σ T1 )  + cc (3.5.8)
2
  
Nearly secular term

Joint initiative of IITs and IISc – Funded by MHRD Page 17 of 67

 NPTEL – Mechanical Engineering – Nonlinear Vibration

In Eq. (3.5.8), the term containing exp ( iω0T0 ) is a secular term and term containing
exp i (ω0T0 + σ T1 )  is a nearly secular term as it will approach to a secular term when σ → 0 . To
have a bounded solution these two terms should be eliminated by imposing the following
condition.
1
2iω0 ( A '+ µ A ) + 3α A2 A − f exp ( iσ T1 ) =
0 (3.5.9)
2
1
Substituting A = a exp ( i β ) in Eq. (3.5.9) and separating the real and imaginary parts, the
2
following first order differential equations are obtained.
1 f
a' = −µ a + sin (σ T1 − β ) (3.5.10)
2 ω0
3α 1 f
a β ' =a 3 − cos (σ T1 − β ) (3.5.11)
8 ω0 2 ω0
One may write these two equations in their autonomous form by substituting= γ σ T1 − β . The
resulting equations are
1 f
a' = −µ a + sin γ (3.5.12)
2 ω0
3α 3 1 f
aγ ' =−aσ a + cos γ (3.5.13)
8 ω0 2 ω0
Equations (3.5.12) and (3.5.13) are known as the reduced equations and can be used for finding
the response and stability of the system. By analytically or numerically solving these equations
one may obtain the amplitude and phase of the response of the system. The first order response
of the system can be given by
=u a cos (ω0t + β ) + O(ε ) (3.5.14)
It may be noted that for steady state, the amplitude and phase of the system do not depend on the
time and hence the time derivative terms i.e., a′ and γ ′ should be equal to zero. Hence, for steady
state one can write
1 f
µa = sin γ (3.5.15)
2 ω0
3α 3 1 f
aσ − a = − cos γ (3.5.16)
8 ω0 2 ω0
Now squaring and adding Eqs. (3.5.15) and Eq. (3.5.16), the following closed form equation is
obtained.
  3 α 2  2
2
f2
µ + σ −
2
a   a =2 (3.5.17)
  8 ω0   4ω0

It may be noted that this equation is a 6th order polynomial in terms of a and is quadratic in
terms of detuning parameter σ . Hence, solving the quadratic equation, one can obtain the
following relation for the frequency response curve.

Joint initiative of IITs and IISc – Funded by MHRD Page 18 of 67

 NPTEL – Mechanical Engineering – Nonlinear Vibration

(3.5.18)

Example 3.5.2: Apply method of multiple scales to the following nonlinear parametrically
excited system
q + 2 ε ζ q + q + ε ( α1 q 3 + α 2 q 2 q + α 3 q 2 q ) − εf1 cos ( 2 ωτ ) q -εk1 (1 + cos ( 2 ωτ ) ) qq
 2 =0
(3.5.19)

This equation contains parametric term f1 cos ( 2 ωτ ) q and nonlinear damping term

k1 (1 + cos ( 2ωτ ) ) qq
 2 , along with cubic geometric ( α q3 ) and inertial ( α q 2 q + α q 2 q )
1 2 3
nonlinear terms.

Solution: In this method the displacement q can be represented in terms of different time scales
(T0 , T1 ) and a book keeping parameter ε as follows.
q (τ; ε) q0 (T0 , T1 ) + ε q1 (T0 , T1 ) + O ( ε 2 ) .
= (3.5.20)
Here, T0 = τ , and T1 = ετ . The transformation of first and second time derivatives are given by

d = D + εD + O(ε 2 ) and d 2 = D 2 + 2εD D + O(ε 2 ) .

0 1 0 0 1
dτ dτ 2
∂ ∂
where, D0 = , and D1 = .
∂T0 ∂T1
Substituting Eqs. (3.5.20) in (3.5.19) and equating the coefficient of like powers of ε , yields the
following equations.
Order ε 0 : D0 2 q0 + q0 =
0, (3.5.21 )
Order ε 1 : D0 2 q1 + q1 =−2 D0 D1 q0 − 2 ζ D0 q0 −α1 q03 − α 2 ( D02 q0 ) q0 − α 3 ( D0 q0 ) q02
2

+ f1 cos ( 2ωT0 ) q0 + k1 (1 + cos ( 2ωT0 ) ) ( D0 q0 ) q02 . (3.5.22)

General solutions of Eq. (3.5.21) can be written as
=
q0 A(T1 , T2 ) exp(iT0 ) + A(T1 , T2 ) exp(−iT0 ) . (3.5.23)
Substituting Eq. (3.5.23) into Eq. (3.5.22) leads to

Joint initiative of IITs and IISc – Funded by MHRD Page 19 of 67

 NPTEL – Mechanical Engineering – Nonlinear Vibration

D0 2 q1 + q1 − ( 2 i A′ + 2iζA + ( 3α1 − 3α 2 + α3 − ik1 ) A2 A ) exp(iT0 )+νA3exp ( 3iT0 )


secular term

 
f1  ik ik
+ik1 A exp ( 3iT0 ) +
3
A exp i (2ω − 1)T0  + 1 A3 exp i (2ω + 3)T0 + 1 A2 A exp i (2ω + 1)T0 (3.5.24)
2   2 2
 nearly secular term 0 
ik ik
− 1 A 2 A exp i (2ω − 1)T0 + 1 A3 exp i (3 − 2ω)T0 + cc
2
    2 
nearly secular term nearly secular term

Here, ν = −α1 + α 2 + α 3 . One may observe that any solution of Eq. (3.5.24) will contain secular
or small divisor terms when non-dimensional frequency of magnetic strength ( ω ) is nearly equal
to 1 which is known as simple resonance case. In this case, one may present detuning parameter
σ to express the nearness of ω to 1, as
ω = 1 + ε σ , and σ = O(1) (3.5.25)
Substituting Eq. (3.5.25) into Eq. (3.5.24), one may obtain the following secular or small divisor
terms.
−2 i A′ exp(iT0 ) − 2iζA exp(iT0 ) − ( 3α1 − 3α 2 + α3 − ik1 ) A2 A exp(iT0 )
f1 k k
A exp ( 2σT1 ) + i 1 A3 exp ( −2σT1 ) − i 1 A 2 A exp ( 2σT1 ) =
+ 0. (3.5.26)
2 2 2
1 ( i β(T ) )
Putting A equal to a (T1 ) e 1 into Eq. (3.5.26) and separating the real and imaginary terms,
2
one may find the reduced equations as given below.
da k f
= −ζ a + 1 a 3 + 1 a sinγ , (3.5.27)
dT2 8 4
dγ  ω −1  3  α3  3 1 3 f1
= a 2a  −  α1 − α 2 +  a + a k1 sin γ + a cosγ . (3.5.28)
dT2  ε  4 3  4 2
da d γ
For steady state as , are equal to 0,the above equations reduces to
dT2 dT2
 k f 
a  −ζ + 1 a 2 + 1 sinγ  =0 (3.5.29)
 8 4 
 ω −1  3  α3  2 1 2 f1
2  −  α1 − α 2 +  a + a k1 sin γ + cosγ = 0 (3.5.30)
 ε  4 3  4 2
One may observe from the Eq. (3.5.29) that, the system possesses both trivial ( a = 0 ) and
nontrivial ( a ≠ 0 ) responses. The nontrivial response can be obtained by numerically solving
Eqs. (3.5.29) and (3.5.30). Later it can be studied that the stability of the steady state response
can be determined by finding the eigenvalues of the Jacobian matrix obtained by perturbing Eqs.
(3.5.27) and (3.5.28).
Joint initiative of IITs and IISc – Funded by MHRD Page 20 of 67
 NPTEL – Mechanical Engineering – Nonlinear Vibration

The first order non-trivial steady state response is given by

1 
= q a cos  ( ωτ − γ )  (3.5.31)
2 
Exercise Problems
1. Find the expression for the transition curve for a parametrically excited system given by the
following equation of motion.
u + ω02u + 2εµ u + ( ε f cos Ωt ) u =
0
2. The equation of motion for a bimaterial beam with alternating magnetic field and thermal
loads (G. Y. Wu, Journal of Sound and Vibration 327(2009)197-210) can be given by
Ω 2 q + q + 2 Ω  k1 + k2 (1 + cos ( 2 τ ) ) q 2  q − f1 cos ( 2 τ ) q =0
Derive the expression for the frequency response by using method of multiple scales.

References for further reading

One may read higher order method of multiple scales from Rahman and Burton(1989) and
Dwivedy and Kar (1999) .
1. Z. Rahman and T. D. Burton, On higher order method of multiple scales in nonlinear
oscillations-periodic steady state response, Journal of Sound and Vibration 133, 369-379, 1989.
2. S. K. Dwivedy and R. C. Kar Nonlinear response of a parametrically excited system using
higher–order method of multiple scales, Nonlinear Dynamics, 20, 115-130, 1999.
3. A. H Nayfeh and D. T. Mook, Nonlinear oscillations, New York, Willey Interscience,1979.
4. H. Boyaci and M. Pakdemirli, A comparison of different versions of the method of multiple
scales for partial differential equations, Journal of Sound and Vibration, 204(4),595-607, 1997.

Joint initiative of IITs and IISc – Funded by MHRD Page 21 of 67