CONTROLLING CHAOTIC INSTABILITIES IN BRILLOUIN

FIBER SENSOR BASED ON NEURAL NETWORKS

Tae- Su Jnag

, Kwan-Woong Kim

and Yong K. Kim

School of Electrical Information Communication Engineering, WonKwang University, 570-749, Iksan, Korea

Korea Atomic Energy Research Institute, 305-353, Deajeon, Korea

Keywords: Optical Fibre Sensor, Neural Network, Controlling Chaos, Instability, Optical Processing.

Abstract: In this paper the neuron operation based on neural network in optical fibre system has described. The

inherent feedback wave in optical fibre leads to instabilities in the form of optical chaos. Below threshold

value temporal evolution has periodic and can become chaotic condition. Controlling of chaotic induced

transient instability in Brillouin fibre sensor has been implemented with Kerr nonlinearity with 13GHz

backward shift, IR detector with 20ps impulse response, and ~10ns round trip time in optical network. The

detected sensor has up-shifted in frequency by about ~20MHz from sensing target. The Controlling chaotic

instabilities can lead to stable and periodical states; create optical logic data streams. It can lead to large

optical memory capacity in neural networks.

1 INTRODUCTION

It is well known that optical fibers have potential

usage in applied engineering technology other than

optical communications, such as expanding research

in versatile fiber optic sensors and networks (Cotter,

1983). Our research has also focused on integrating

fiber optic sensors with neural networks to create a

system that is capable of sensing, and controlling

shape or orientation of the medium with respect to

its environment, as a first step in creating a smart

sensor structure (Yong, 2003). Specifically, we have

focused on configuring and developing a fiber

sensing system that behaves as a neural network,

capable of learning by network experience,

predicting future reactions to environmental

changes, and executions as prescribed.

Such a smart sensor system based neural

networks can potentially implement a massively

parallel computational architecture with its attendant

reduction in processing time while managing the

complexity of the system, i.e. the sensing grid

(Lyons and Lewis, 2000). Our fiber sensor network

would learn the correct algorithms by example

during training and have the ability to generalize to

untrained inputs after training is completed under

sensor networks (King and Lyons, 2003).

In general, an artificial neuron in neural networks

can be thought of as a device with multiple inputs

and a single or multiple outputs (Tariq, 1998).

Actually, in equivalent networks, the inputs to a

neuron are weighted signals in neural systems. The

neuron adds the weighted signals, compares the

result with a preset value, and activates if the sum

exceeds threshold. The networks can be explained as

a sensor inputs weighted signals. In the nonlinear

optical phenomenon of stokes wave the system

combined weighted signals produces an output if the

weighted sum is greater than the threshold (Kovalev

and Harrison, 2006).

Optical turbulence in the system and instability

and periodic oscillation are easily seen with hybrid

optically bistable device, which indicated fiber

network configuration, with a delay in the feedback

(Wang et al., 2008). The stability analysis of steady

states is different from the usual criteria applied to

differential equations (Agrawal, 2001). The

instabilities to arise in steady state has implemented

with time dynamics as chaotic (Kuo et al., 2010). A

practical implementation for theoretical scheme has

discussed for neural networks.

In this paper, the implementation for controlling

Chaotic Instabilities in optical fiber has been studied

with neuron operation based on neural network.

177

Jnag T., Kim K. and K. Kim Y..

CONTROLLING CHAOTIC INSTABILITIES IN BRILLOUIN FIBER SENSOR BASED ON NEURAL NETWORKS.

DOI: 10.5220/0003833001770180

In Proceedings of the 1st International Conference on Sensor Networks (SENSORNETS-2012), pages 177-180

ISBN: 978-989-8565-01-3

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

2 NEURON OPERATION BASED

OPTICAL SENSOR NETWORK

Nonlinear effects in optical fiber has emerged as a

versatile tool for the design of active optical devices

for all-optic in-line switching, channel selection,

amplification and oscillation, as well as in optical

sensing, optical communications, optical logic

elements in optical computation and sensing, and a

host of other applications (Lyons and Lewis, 2000).

The paper attempts to present a survey and some of

our own research findings on the nature of optical

fiber scattering in single mode optical fibres and its

device applications (Kuo and More, 2010). In

theory, the backscattering nature of the phenomenon

enables its application as channel selectors and

switches and filters in optical transmission and

communications (Agrawal, 2001). We have been

engaged in the design and implementation of fiber

configurations, such as rings and loop mirrors, with

the purpose of lowering the threshold. We report on

experimental schemes involving fiber ring with

amplifier. These successful devices are being studied

for application as optical logic and neuron elements

for optical switching, and highly versatile sensors

(King and Lyons, 2003).

The backscattering nature of nonlinear process in

optical fiber sensor network and the existence of a

threshold provide potential optical device functions,

such as optical switching, arithmetic, and neural net

functions in neural networks (Traq and Habib, 1998).

The inputs to the network are the fibre optic sensor

signal outputs, and the network outputs are the

control signals for actuation controls. The true

advantage of this system for application to sensor

structures lays both in its capability to analyze

complex sensor signal patterns. The experimental

feature based on neural networks in hardware

implementation is shown in real time in Figure1.

Figure 1: Optical fibre generates multiple Stokes waves

for optical logic networks.

An artificial neuron can be thought of as a device

with multiple inputs and single or multiple outputs.

The neuron adds the weighted signals, compares the

result with a preset value, and activates if the sum

exceeds threshold values (W.B. Lyons and Lewis,

2000). In the nonlinear optical phenomenon, the

combined weighted signals also produce an output if

the weighted sum is greater than the threshold. Our

research has also focused on a simplified multi-

layered ward neural network; the processing node

between interconnects in sensor networks, where

weighted sums are fed to a threshold decision

processing elements. The implemented frequencies

scale have v

> v

3 CONROLLING OF CHAOTIC

INSTABILITY

Conversion of optical fibre chaos induced instability

to periodic effect is inspired by theory in nonlinear

dynamics. The basic idea lies in the stabilization of

unstable periodic orbits embedded within an

optically chaotic attractor (Kovalev and Harrison,

2006).

The bit stream orbits are very dense; a successful

control may therefore serve as a generator of rich

forms of periodic waves, thus turning the presence

of chaos to advantage (Wang et al., 2008). The

experimental setup for controlling chaotic instability

based on optical fibre network has shown in Figure

Figure 2: Schematic diagram for controlling chaos induced

instability.

There are two fiber ring operations, one passive

and one backwards active. In the passive ring, the

laser beam circulates the ring, with the output

characteristics governed by the finesse of the ring.

The neuron active ring is governed by the finesse

and the nonlinear scattering phenomenon in the

fiber. Through acousto-optic coupling, a laser pump

induces an acoustic wave in the fiber, which scatters

part of the laser beam backwards as a Stokes wave.

As configured, this Stokes wave is circulated in the

ring, and continues to be amplified by the laser

SENSORNETS 2012 - International Conference on Sensor Networks

178

(pump). The ring is a network resonator or oscillator

and amplifier. The enhanced sensor has clearly

demonstrated in Fig.1 in the form of line narrowing.

The sensor signal will act as a stokes wave for the v

and as a pump wave for the v

when v

- v



and v

- v



. It is to be noted, since neuron is a

backscattering process, threshold g = g

L/A = 21

for a straight fiber, has lowered in a fiber ring to

approximately 0.1. On increasing the pump strength

in the vicinity of the threshold g  4, a stoke signal

emerges from stochastic high-frequency noise to

exhibit randomly amplitude-modulated periodic

oscillations at the fundamental periods 2T

Figure 3: Optical pulse induced instabilities in function of

time (μsec/div) at threshold (a), high above threshold (b).

A stabilized proves laser has yielding ~12.1GHz

backward scattering shift, 25GHz IR Photo-detector

(20ps impulse response) connected to a optical

analyzer. The temporal repetition rate of which

corresponds to a pulse round-trip time in the fibre-

ring taken to be less than ~10nsec. The R is the

mirror reflectivity and B is beam splitter. The narrow

gain spectrum and relatively small frequency shift of

the sensor process will allow the use of the same

oscillator format for signals under identical fibres.

The sensor signal is up-shifted in frequency by about

10MHz from the sensing signal in an amplifier

format. It’s has been implemented based on neural

network to sensed it. The optical sensor pulse train

amplitudes remain unstable, particularly just below

pump threshold. When the observation is made

using a long time scale (~100usec/div), the target

pulse output exhibits randomly distributed trains of

periodic pulses. Partial stabilization of amplitude

fluctuations has achieved as laser pump power

approaches maximum value with ~nm source pump.

These experimental features are shown in real time

in Figure 3 (a) and (b). It’s shown that the optical

pulse induced on instabilities as function of time. In

the immediately above threshold, chaotic

instabilities have occurred towards turbulence. In

high above threshold, chaotic instability has

periodically turbulence. The discrete lines have

widths of less than 1MHz, with a resolution of 90/58

= 1.55MHz. We propose to employ continuous

optical feedback for control in which coherent

interference of the chaotic optical signal itself in

achieve signal differences.

Figure 4: Transiently controlled instabilities at threshold

(a) and high above threshold (b). The examples of

sequence of suppression are assigned by ‘1’.

If suppressing by attractor proves to control

chaos then, suppressing under natural chaos can be

exploited as a means of sensing structural chaos in

systems. The examples of sequence of suppression

are assigned by ‘low level’ and ‘high level’ states

since previous results. These states has implemented

with TDM phase analysis and modulation. multi-

CONTROLLING CHAOTIC INSTABILITIES IN BRILLOUIN FIBER SENSOR BASED ON NEURAL NETWORKS

179

stable periodic states, as shown in Figure 4 (a) and

(b), can lead to logic ‘low level’ or ‘high level’ and

can in principle create large memory capacity as

input data bit streams in digital network systems. Its

implementation also still requires much engineering

improvements, such as arriving target at a spatial

resolution speckle, and suppression of its tendency

to chaos. We will focus on a more realistic case of N

weighted pumps, i.e., one v

, one v

and v

, v

…



. The optical sensor based neural networks is

typically configured into an array for the sensor

networks in optical fiber. This sensor concept can be

used to form either adaptive sensor arrays which are

similar to our researched neural network system, or

used simply as an embedded sensor inside structures

or materials.

4 CONCLUSIONS

Controlling of chaotic instabilities in optical fiber

sensor networks has been implemented under chaos-

induced transient instability in optical systems.

Controlling also leads to the possible logic theory

with ‘low level’ or ‘high level’, as logic ‘0’ and ‘1’

with stable and periodic states. It’s used for neural

networks as a neural net. It is theoretically possible

to apply the multi-stability data regimes as an optical

large memory device for multi encoding-decoding

messages. It can be also applied for complex data

transmission in TDM networks and other optical

communications. It can be possible to create large

optical memory capacity.

REFERENCES

Cotter, D., 1983. “Stimulated Brillouin scattering in

Monomode Optical Fiber.” J. Opt. Com. 4, 10-19.

Yong, K, Kim., Choon, B, Park., 2003. “Study of chaotic

and instability effect of optical fiber using on the

internet.” SPIE. 5246, 648-655.

W. B. Lyons and E. Lewis, 2000, “Neural networks and

pattern recognition techniques applied to optical fibre

sensors,” Transaction of the Institute of Measurement

and Control 22(5), 385-404.

D. king, W. B. Lyons, C. Flanggan and E. Lewis, 2003,

“An optical fibre ethanol concentration sensor

utilizing Furier transform signal processing analysis

and artificial neural network pattern recognition,” J.

Opt. A: Pure Appl.Opt. 5, S69-S75.

Tariq, S., Habib, M, K., 1998. “Neural operation using

stimulated Brillouin scattering in optical fiber.” Opt.

Eng. 37, 1823-1826.

Agrawal, G, P., 2001. “Nonlinear Fiber Optics, 3rd,

Academic press”, London.

V. I. Kovalev and R. G. Harrison, 2006. “Suppression of

stimulated Brillouin scattering in high-power single

frequency fiber amlifiers,” Opt. Lett.31, 161-163.

C. Wang, F. Zhang, Z. Tong, T. Ning, and S. Jian., 2008.

“Suppression of stimulated Brillouin scattering in

high-power single-frequency multicore fiber

amplifiers,” Opt. Fiber Technol. 14, 328-338.

J. B. Coles, B. P. P. Kuo, N. Alic, S. More, C. S. Bres, J.

M. Chavez Boggio, P. A. Andrekson, M. Karlsson,

and S. Radic., 2010. “Bandwidth efficient phase

modulation techniques for stimulated Brillouin

scattering suppression in fiber optic parametric

amplifiers,” Opt. Express 18, 18138-18150.

Yang Jiang, J. Yu, Binfchen Han, Li Zhang, Wenrui Wang,

Lital Zhang, and Enze Yang., 2009. “Millimeter-wave

subcarrier generation utilizing four-wave mixing and

dual-frequency Brillouin pump suppression,” Opt. Lett.

48(3), 030502-1 – 030502-2.

S. Mussot, M. Le Parquier, P. Szriftgiser., 2010. “Thermal

noise for SBS suppression in fiber optical parametric

amplifiers,” Opt. Communication. 283, 2607-2610.

SENSORNETS 2012 - International Conference on Sensor Networks

180