Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

SPP-2010-172

Numerical simulation for a Duffing oscillator driven by white


noise with constant and time-varying noise strengths
Kendrick Agapito1* and Jose Perico Esguerra1
1
National Institute of Physics, Diliman, Quezon City
*kagapito@nip.upd.edu.ph

Abstract
We present a numerical analysis of the Duffing oscillator driven by white
noise with constant and sinusoidally varying noise strengths. For each „noise
strength history‟ the Runge-Kutta method was used to generate trajectories
corresponding to 200 to 500 realizations of the noise for the driven Duffing
oscillator. The plot of variance versus time was obtained and analyzed. The
results obtained for 200, 300, 400, and 500 realizations were in good
agreement. It is observed that for both the constant noise strength and
sinusoidal noise strength cases the variance exhibits a generally increasing
trend but with an additional oscillatory fluctuation. However the nature of the
oscillatory fluctuation is different for the two cases. For the constant noise
strength case the amplitude of the oscillatory fluctuation is increasing; while
for the case with sinusoidal noise strength the amplitude of the oscillatory
fluctuation is modulated. The variance is generally larger for noise strength
that oscillates with a frequency ω = 1.32 than for ω = 100 and ω = 0.01. ©
2010 Samahang Pisika ng Pilipinas
Keywords: Duffing, noise, numerical
1. Introduction
Interest in stochastic differential equations (SDE) or differential equations involving random variables or
noise have grown since the explanation of the phenomenon of Brownian motion by Einstein, Smoluchowski and
Langevin [1]. Today, special attention is being given to nonlinear stochastic differential equations of the form
(1)
where is the noise. Though there have been many studies in the field where interesting properties have
been obtained (e.g. stochastic resonance [2]), not much has been done in studying the effects of introducing a
time-varying noise in (1). Additionally motivated by the fact that the Duffing oscillator described by the
equation
(2)
is the simplest nonlinear differential equation that may be used as a benchmark for studying SDE with higher
orders of nonlinearities, we present a numerical analysis of the Duffing oscillator driven by white noise with
constant and sinusoidally varying noise strengths. It might be possible in principle to control the noise strength
history of a system. For example, in the case of Brownian motion, if a medium is cooled or heated its viscosity
as well as the root mean square speed of the Brownian particles in the medium change in response to the
changing temperature of the medium [3] – these effects may be modeled by using time varying noise strengths.
By performing computational experiments, we intend to identify qualitative behavior that may in the future be
the subject of further analysis through other methods (e.g. perturbative and variational calculations, stability and
bifurcation analysis etc.). In this work we focus on the evolution of the variance of solutions to the noise-driven
Duffing equation.

2. Methodology
We take into consideration the Duffing oscillator with additive noise,
, (3)
where is an additive noise with properties
. (4)
The correlator given in (4) describes an uncorrelated noise, more commonly known as white noise, with
noise strength D. Python language was used to generate random numbers that will serve as the additive noise
with a Gaussian distribution
. (5)
The noise strength was chosen to be a constant value D(t)=D and a sinusoidal function of time
D(t)=D1+D2sin(ωt). The same language was used to numerically solve the nonlinear Duffing equation by way

28th SPP Physics Congress, October 25-27, 2010, Antipolo City - 1


SPP-2010-172

of the Runge-Kutta method. Two-hundred noise realizations for every noise strength history were used to obtain
the variance of the displacement x(t).

3. Results and Discussion


Figure 1 shows the displacement versus time plots of a realization of the noise-driven Duffing oscillator for
various noise strengths. Notice that for constant noise strength, the frequency of oscillation increases as the
noise strength increases. Meanwhile, Figure 2 shows the variance of the displacement versus time for the noise-
driven Duffing oscillator for different number of realizations. It can be seen in Figure 2 that 200 to 500
realizations yield nearly identical results for the variance. The variance exhibits a generally increasing trend but
with an additional oscillatory fluctuation for both the constant noise strength case (Figure 2) and the sinusoidally
varying noise strength case (Figure 3). The frequencies used for the sinusoidally varying noise strength in the
figure correspond to a low frequency (ω = .01), a high frequency (ω = 100) and the approximate frequency of
the mean trajectory [5]. It may be observed from the figure that the variance for the sinusoidally varying noise is
generally larger than that for the constant noise. Moreover, among the three frequencies presented, the
frequency ω = 1.32 gave the largest values.

(a) (b)

(c) (d)

(e) (f)

Figure 1. x-t plot for the solution of one noise realization of the noise-driven Duffing oscillator when
(a-c) D = 1.0, 5.0, 10.0 and (d-f) D = 5.0 + 5.0 sin (ωt), ω=0.1, 1.32, 100.0. All solutions are
compared with the “noise-less” Duffing oscillator (shown as the blue curve).

28th SPP Physics Congress, October 25-27, 2010, Antipolo City - 2


SPP-2010-172

Figure 2. Variance-time plot for the solution of the noise-driven Duffing oscillator
for D=1.0 and for 200, 300, 400 and 500 realizations.

Figure 3. Variance-time plot for the solution of the noise-driven Duffing oscillator
for one constant noise strength and three sinusoidally varying noise strengths
(D = 5.0 + 5.0 sin (ωt), ω=0.1, 1.32, 100.0). 300 realizations were used.

4. Conclusions and Recommendations


We have presented a numerical analysis of the noise-driven Duffing oscillator for constant and sinusoidally
varying noise strengths. We observe that for both the constant noise strength and sinusoidal noise strength cases,
the variance exhibits a generally increasing trend but with an additional oscillatory fluctuation. However the
nature of the oscillatory fluctuation is different for the two cases. For the constant noise strength case the
amplitude of the oscillatory fluctuation is increasing; while for the case with sinusoidal noise strength the
amplitude of the oscillatory fluctuation is modulated. Understanding through analytical calculations why the
variance seems to be highest for ω = 1.32 and why the amplitude of the additional oscillatory fluctuation is
modulated is a future challenge. The discovery of this new type of time evolution for the variance is the main
result of this paper.

References
1. M. Gitterman, The Noisy Oscillator, Chapters 2 & 3, World Scientific, Singapore, 2005
2. B. Andό, S. Graziani, Stochastic Resonance Theory and Application, Kluwer Academic, Boston, 2000
3. L. E. Reichl, A Modern Course in Statistical Mechanics, Chapter 4, Wiley & Sons, Canada, 1998.
4. P. Amore, N. Sanchez, “Development of accurate solutions for a classical oscillator”, Journal of Sound and
Vibration 300 (2007) 345-351
5. Wen-Xian Xie, Li Cai, Wei Xu, “Numerical simulation for a Duffing oscillator driven by colored noise
using nonstandard difference scheme”, Physica A 373 (2007) 183-190
6. A. Rößler, Runge-Kutta Methods for the Numerical Solution of Stochastic Differential Equations, Shaker
Verlag, Germany, 2003

28th SPP Physics Congress, October 25-27, 2010, Antipolo City - 3

You might also like