Artificial intelligence in the design of microstrip antenna

Artificial Intelligence in the
Design of Microstrip
Antenna

By:
Raj Kumar Thenua
Vandana V. Thakare
Department of Electronics & Instrumentation Engineering
AEC, Agra, UP

Outline
 Introduction
 Methodology
 Design of a microstrip line feed rectangular Microstrip Antenna using
IE3D EM Simulator
 Analysis of a microstrip line feed rectangular Microstrip Antenna
using ANN
 Application
 Conclusion
 Future scope of the work
 Results
 References

2 8/9/2012

Introduction
 Accurate RF/Microwave design is crucial for the current
upsurge in VLSI, telecommunication and wireless technologies
 Design at microwave frequencies is significantly different from
low-frequency and digital designs
 Substantial development in RF/microwave CAD techniques
have been made during the last decade
 Further advances in CAD are needed to address new design
challenges
 Fast and accurate models are key to efficient CAD
 Neural network based modeling and design could significantly
impact high-frequency CAD

3 8/9/2012

A Illustration Example
MICRO STRIP PATCH ANTENNA

 radiating metallic patch on a ground substrate
 patch can take different configurations but rectangular
and circular patches are the most popular one because of
ease of analysis and fabrication and their attractive
radiation characteristics

4 8/9/2012

Example Antenna

Figure 1.0

5 8/9/2012

Justification for Present Work
 Antenna design is a very complex problem .
 Spacecrafts,aircrafts,missiles and satellite
applications require antenna in small ,size
,weight,cost and easy to install.
 Mobile ,radio and other wireless communications
also demands such specific antennas.
 To fulfill such requirements Microstrip patch
antennas are used.

6 8/9/2012

Methodology
 Development of an ANN model in Mat Lab Neural
Network Tool Box for the calculation of patch
dimensions for Microstrip Antenna .
 The data for training the network is generated
using IE3D a Electro Magnetic Simulator.
 As an example a microstrip line feed rectangular
Microstrip patch Antenna is being considered
and designed using simulator for a particular
resonating frequency i.e. 4.9 GHz.
 Validation of ANN model

7 8/9/2012

IE3D Electro Magnetic Simulator

• Computer Aided Simulation
 Integrated Electromagnetic three Dimensional
(IE3D) Software
 Developed by Zeland Inc., United States

 Design Dimensions can be milli, micro and so on.

 Simulation Time – Few Minutes

 Output Result can be obtained in the form of patch
dimensions , VSWR, Return loss, Gain Directivity
,Radiation efficiency,etc.

8 8/9/2012

Design parameters
 designed in IE3D Simulator
 With dielectric constant Єr = 4.7
 Substrate thickness h = 1.588mm
 Length L= 6.6 mm
 Width W = 8.8 mm
 Length of the feed l = 2 mm
 Width of the feed w = 0.5mm
 Resonating frequency fr = 4.9 GHz

9 8/9/2012

Inset feed Microstrip Antenna

Figure 2.0

10 8/9/2012

Relations to calculate different parameters of
rectangular patch antenna

 The effective dielectric constant of the dielectric
material is given by

11 8/9/2012

Contd…

 For an efficient radiator, a practical width that leads
to good radiation efficiencies is given by:

 where vo is the free-space velocity of light.

12 8/9/2012

Contd..

 The actual length of the patch:

where ∆L is the extension of the length due to the
fringing effects and is given by:
13 8/9/2012

Contd…

 ∆L is given by

14 8/9/2012

Microstrip Antenna Designed at
4.9GHz

Figure 3.0

15 8/9/2012

IE3D Electromagnetic Simulator
to Generate Simulated Data

efficiency
h
gain
W
IE3D Input impedance
L
SOFT
Feed WARE VSWR
dimensions
Єr Return loss
frequency band

Figure 4.0
16 8/9/2012

Analysis ANN ModelA

h
W
Єr
ANN
F1 L

F2

Figure 5.0

17 8/9/2012

Neural Network Model
 The ANN model is a system with input vector s’
representing the circuit design parameters:
height of substrate= h
dielectric constant = Єr
cut off frequencies F1 and F2
 And the output vector r’ representing the Patch
dimensions.
Length of the Patch = L
Width of the Patch = W

18 8/9/2012

Neural Network Architecture
 Three layer network structure has been considered
 Input layer will have four neurons to accept input
parameters h ,Єr, F1 and F2.
 Output layer will have two neurons to output patch
dimensions.
 The hidden layer will have number of neurons
depending upon design accuracy.
 The radial basis function network is considered for
the network architecture.
 The network will be trained using radial basis
function

19 8/9/2012

RBF Network
 Feed forward neural networks with a single hidden layer
that use radial basis activation functions for hidden
neurons are called radial basis function networks.
 RBF networks are applied for various microwave
modeling purposes.
 RBF can approximate any regular function.
 Trains faster than any multi-layer perceptron.
 It has just two layers of weights.
 Input is non-linear and output is linear.
 No saturation while generating outputs

20 8/9/2012

Architecture of RBF Network

x1

y1
x2

y2
x3
output layer
input layer
(linear weighted sum)
(fan-out)

hidden layer
(weights correspond to cluster centre,
output function usually Gaussian)
21 8/9/2012

RBF Functions
 Gaussian Activation Function
1
j x exp X j j X j j 1...L
 Output Layer: is a weighted sum of hidden inputs
L

k (x) jk . j (x)
j 1
X is a multi dimensional input vector with elements xi and j is
the vector determining the center of basis function j and
has elements ji.

22 8/9/2012

Contd..
 The distance measured from the cluster centre is
usually the Euclidean distance.
n
rj ( xi wij ) 2
i 1

23 8/9/2012

MAT Lab Tool Box

 In order to develop the ANN model MAT LAB
neural network tool box has been used.

24 8/9/2012

Network Training
 Two kinds of training algorithms
- Supervised and Unsupervised
- RBF networks are used mainly in supervised
applications
- In this case, both dataset and its output is known.
- The model is trained with the set of 200 samples
 Clustering algorithms (k-mean)
 The centers of radial basis functions are initialized
randomly.
 For a given data sample Xi the algorithm adapts its
closest center

25 8/9/2012

Network Testing
 The performance of the network is tested by a
second set of a sample vectors pairs which are not
included in training data set but must be in the
specified given range.
 If the unknown sample pairs are modeled correctly
the network is likely to represent a valid model.
 The model is tested for around 26 values and found
satisfactorily.

26 8/9/2012

Application
 After training and testing the model is ready to be
used as a simulator for the calculation of patch
dimensions for the Microstrip antenna.
 The model can be reused in the design process
many times without the cost of EM Simulations.
 The network is capable of predicting the output for
any given input in the trained region inexpensively.

27 8/9/2012

Results
S. No. F1 GHz F2 GHz W mm L mm W mm L mm
(IE3D) (IE3D) (RBF) (RBF)
1 4.93 5.03 17.8 13.35 17.84 13.33

2 4.86 4.95 17.8 13.55 17.83 13.53

3 4.82 4.91 17.8 13.65 17.84 13.66

4 4.8 4.89 17.8 13.85 17.83 13.84

5 4.77 4.85 17.8 14.05 17.84 14.02

6 4.73 4.81 17.8 14.15 17.83 14.17

7 4.71 4.79 17.8 14.25 17.82 14.26

8 4.69 4.76 17.8 14.35 17.83 14.36
28 8/9/2012

W mm L mm W mm L mm
S. No. F1 GHz F2 GHz
9 4.66 4.73 17.8 14.45 17.84 14.46

10 4.78 4.87 18.3 13.85 18.29 13.86

11 4.65 4.73 18.3 14.35 18.31 14.37

12 4.61 4.7 18.8 14.35 18.82 14.36

13 4.49 4.55 18.8 14.85 18.83 14.84

14 4.47 4.55 19.3 14.85 19.32 14.83

15 4.37 4.41 19.3 15.35 19.31 15.33

16 4.35 4.41 19.8 15.35 19.82 15.37

29 8/9/2012

W mm L mm W mm L mm
S. No. F1 GHz F2 GHz
17 4.31 4.37 20.3 15.85 20.29 15.84

18 4.29 4.35 20.3 16.35 20.31 16.36

19 4.27 4.33 20.8 16.35 20.84 16.36

20 4.26 4.33 20.8 16.85 20.82 16.86

21 4.21 4.27 21.3 16.85 21.33 16.84

22 4.16 4.21 21.3 17.35 21.29 17.36

23 4.14 4.19 21.8 17.35 21.83 17.37

24 4.02 4.04 21.8 17.85 21.83 17.86

25 3.99 4.01 22.3 17.85 22.33 17.83

26 3.92 3.93 22.3 18.35 22.31 18.36
30 8/9/2012

Conclusion
 The neural network developed in this work models
the patch dimensions calculator for microstrip line
feed rectangular Microstrip patch antenna.
 The radial basis function network is giving the best
approximation to the target values
 The values obtained from ANN are very close to
simulation readings .
 The error between output of ANN and IE3D is very
very small.
 The developed model for resonant structure
Microstrip Antenna validate the modeling approach.

31 8/9/2012

Future Scope
 Working with the same concept and design
analysis ,different microwave and RF devices
could be designed .
 Different analysis and synthesis ANN model
could be developed for other performance
parameters of the microwave circuits like input
impedance ,directivity ,gain, VSWR, return
loss etc.
32 8/9/2012

References
 [1] Q. J. Zhang, K. C. Gupta, Neural Networks for RF and Microwave Design,
Artech House Publishers, 2000.
 [2] R. K. Mishra, Member, IEEE, and A. Patnaik , ANN Techniques in
Microwave technology .
 [3] A. H. Zaabab, Q.J. Zhang, M. Nakhla, ”Analysis and Optimization of
Microwave Circuits & Devices Using Neural Network Models”’ IEEE MTT-
S Digest 1994, pp 393- 396
 [4] C.A. Balanis, Antenna Theory, John Wiley & Sons, Inc., 1997.
 [5] D.M. Pozar, Microstrip Antenna , Proc. IEEE, Vol. 80, pp.79-81,
 [6] F. Wang, V.K. DevabhaktunI, and Q.J. Zhang,” A
hierarchical neural network approach to the development of library of
neural models for microwave design”,
IEEE Intl. Microwave Symp. Digest, pp. 1767-1770, Baltimore, MD, 1998.

33 8/9/2012

Contd..
 [7] F. Peik, G. Coutts, R.R. Mansour ,COM DEV, Cambridge, ON, Canada,
“Application of neural networks in microwave circuit modelling” , Electrical
and computer Engineering,1998,IEEE Canadian Conference , vol-2 ,24-28 May
1998,pages:928-931
 [8] S. Devi , D.C. Panda, S.S. Pattnaik, “A novel method of using Artificial neural
networks to calculate input impedance of circular microstrip antenna”,
Antennas and Propagation Society International Symposium, Vol. 3, pp. 462 – 465,
16-21 June 2002.
 [9] R.K. Mishra, A. Patnaik, “Neural network-based CAD model for the design of
square-patch antennas”, Antennas and Propagation, IEEE Transactions, Vol. 46,
No. 12, pp. 1890 – 1891, December 1998.
 [10] A. Patnaik, R.K. Mishra, G.K. Patra, S.K. Dash, ”An artificial Neural network
model for effective dielectric constant of microstrip line,”
IEEE Trans. On Antennas Propagat., vol. 45, no. 11, p. 1697, Nov. 1997.
 [11] Simon Haykin, Neural Networks second edition pHI

34 8/9/2012

Artificial intelligence in the design of microstrip antenna

More Related Content

What's hot

Viewers also liked

Similar to Artificial intelligence in the design of microstrip antenna

Recently uploaded

In this document

Artificial intelligence in the design of microstrip antenna