EMPIRICAL STUDY OF ANN BASED PREDICTION OF

RESONANT FREQUENCY AND BANDWIDTH OF

CIRCULAR SLOT LOADED POLYGON FSS

AUTHOR NAME

• M. PANDA

• MD RABIUL HOSSAIN

• MD EHTESHAM HUSSAIN

• SAMRIDDHA SAMANTA

• RUBABA RAHMAN

Presented byRUBABA RAHMAN

OUTLINE What is FSS

Elements of FSS Design

Types of FSS

Applications of FSS.

Work Done By The Candidate

Conclusions

WHAT IS FREQUENCY SELECTIVE SURFACE ???

A frequency-selective surface (FSS) is any thin, repetitive surface designed to reflect, transmit or absorb electromagnetic fields based on frequency. In this sense, an FSS is a type of optical filter or metal-mesh optical filters in which the filtering is accomplished by virtue of the regular, periodic pattern on the surface of the FSS.

ELEMENTS OF FSS DESIGN

Element Geometries:Patch-type elements-capacitive effect.Aperture-type element-Inductance Effect.

PATCH-TYPE ELEMENTS

The patch-array produces a capacitive response, Band Stop Filter

APERTURE-TYPE ELEMENT

Array of slots is inductive- Band Pass filter

TYPES OF FSSFSS (Based on shapes by MUNK,[1])– 4 Classes.

1. The center connected or N-poles

2. The loop Types

3. Solid interiors or plate types

4.Combinations of 1,2,3

APPLICATIONS OF FREQUENCY SELECTIVE SURFACES ..• Frequency selective surfaces have been most commonly used in

the radio frequency region of the electromagnetic spectrum• use in applications as diverse as the aforementioned microwave

oven, antenna radomes and modern metamaterials.• Sometimes frequency selective surfaces are referred to simply as

periodic surfaces and are a 2-dimensional analog of the new periodic volumes known as photonic crystals.

ARTIFICIAL NEURAL NETWORK (ANN)

Artificial Neural Networks (ANNs) are a family of models inspired by biological neural network i.e. the central nervous system of animals, in particular the brain and are used to estimate or approximate functions that can depend on a large number of inputs and are generally unknown. Artificial neural networks are generally presented as systems of interconnected neurons which exchange messages between each other. The connections have numeric weights that can be tuned based on experience, making neural nets adaptive to inputs and capable of learning.

PROPOSED STRUCTURE OF FSS

The geometry of the single layer FSS unit cell structure shown in Fig which consists of polygon with centrally circular slot loaded. The dimension of unit cell structure is clearly mentioned in the said figure. For x, y direction periodicity the length is 11mm & 16mm respectively. The unit cells are placed in repeated structure in 2D Array form

3 mm

10 mm3

mm

12 m

m

11 mm

16 m

m

SIMULATION RESULT USING ANSOFT

Computer based extensive numerical method is used for designing the parameter of FSS by using different commercial software. Numerical methods are Finite Element Method (FEM), Method of Moment (MOM), Finite Difference Time Domain (FDTD) method etc. For extensive simulation ANSOFT simulation tool is adopted here.

BAND WIDTH VARIATION FOR DIFFERENT Y DIRECTION PERIODICITY

RESONANT FREQUENCY FOR DIFFERENT Y DIRECTION PERIODICITY

ARTIFICIAL ANN FOR PREDICTION OF BW AND RF

Back Propagation (BP) is best on MLN perception. Four layer architecture is used here. One input layer, one output layer and two hidden layer is used for the proposed ANN model. For input layer of ANN model the periodicity in x-direction and y-direction are taken. The output of ANN model is the bandwidth and resonating frequency.

PROPOSE ANN MODEL FOR ESTIMATION OF BAND WIDTH AND RESONANT FREQUENCY

DATA SEQUENCE FOR ANNSl no. Periodi

city in x-

direction

(mm)

Periodicity in

y-directio

n(mm)

1 11 18.5

2 11.5 17

3 12.5 17

4 13.5 18

5 14.5 19.5

Sl No. ANSOFT simulation B.W (GHz)

ANNB.W

(GHz)

1 4.63 4.75

2 5.49 5.38

3 5.28 5.4

4 4.38 4.28

5 3.53 3.66

Sl No. ANSOFT

Simulation RF (GHz)

ANNRF(GHz

)

1 10.35 10.43

2 10.17 10.27

3 9.98 10.1

4 9.88 9.76

5 9.62 9.57

INPUT Output for BW Output for RF

CONCLUSION • All results are quite satisfactory. • Results obtained by proposed ANN structure are compared with ANSOFT

simulation tool results. • In both cases result are quite identical and error value is in acceptance range. • For further future study authors intend to take new model with solid or slit

loaded structure with different dimensions.• Apart from Neural Network authors concentrate on different types of soft

computing tool like PSO,GA, ant colony optimization.

THANK YOU