2009 Third Asia International Conference on Modelling & Simulation

Design of Microstrip Patch Antennas Using Neural Network

Vivek Singh Kushwah and *Geetam Singh Tomar; Member IEEE

Amity School of Engineering & Technology, New Delhi –India

*Vikrant Institute of Technology & Management Indore 452001 India

E-mail: gstomar@ieee.org & vivek_kushwah@rediffmail.com

ABSTRACT is solved using the electromagnetic formulas of the

This paper presents the general design of
microstrip antennas using artificial neural networks for
rectangular patch geometry. The design consists of
synthesis in the forward side and then analyzed as the
reverse side of the problem. In this work, the neural
network is employed as a tool in design of microstrip
antennas. The Neural network Training algorithms are
used in simulation of results for training the samples to
minimize the error and to obtain the geometric
dimensions with high accuracy for selective band of
frequencies.
Key Words: Microstrip antennas, artificial neural
networks, reverse modeling, patch antenna
1. Introduction
To meet the requirements of high performance
spacecraft, aircraft, satellite and cellular mobile
communication, microstrip antennas are used [1].
Often microstrip antennas are referred as patch
antennas because of the radiating elements are photo
etched on the dielectric substrate [2-3]. This radiating
patch may be square, rectangular, circular, elliptical,
triangular, and any other configuration. In this work,
rectangular microstrip patch antennas are considered
where patch dimensions of rectangular microstrip
antennas are designed for its pattern to be normal to the
patch. Because of their narrow bandwidths and
effectively operation in the vicinity of resonant
frequency, the choice of the patch dimensions for the
specified resonant frequency is very important [4]. In
this work, artificial neural network (ANN) models have
been built for the analysis of microstrip antennas in
various forms such as rectangular, circular, and
equilateral triangle patch antennas [5-7]. The analysis is
defined to obtain resonant frequency for a given
dielectric material and geometric structure. However,
the corresponding synthesis ANN model is built to
obtain patch dimensions of rectangular microstrip
antennas (W, L) as the function of input variables,
which are the height of the dielectric substrate(h),
dielectric constants of the dielectric material (εr, εy),
and the resonant frequency (fr). This synthesis problem
microstrip antennas [8]. Two points are especially
emphasized in this work: the resonant frequency of the
antenna and the condition for good radiation efficiency
[9-10]. Using reverse modeling, an analysis ANN is
built to find out the resonant frequency (upper & lower)
for a given rectangular microstrip antenna system. The
forward and reverse sides of this design problem are
defined as black-ANN boxes then the electromagnetic
background is briefly summarized for building the
synthesis ANN model and is reversed for the analysis
purpose of the given antenna system whose results are
compared with existing designs [11]. The rectangular
microstrip antennas are made of a rectangular patch
with width, W, and length, L, over a ground plane with
a substrate thickness h and dielectric constants εr, εy as
given in Figure 1. Dielectric constants used are in the
range 2.2 ≤ εr ≤ 10. However, the most desirable are the
dielectric constants at the lower end of this range
together with the thick substrates, because they provide
better efficiency and larger bandwidth, but at the
expense of larger element size [5]. The spacing
parameter h is replaced by the effective spacing he, and
the geometric mean εg is used for the dielectric constant
εr :
The effective dielectric constant of the dielectric
material is given as:
For an efficient radiator, a practical width that leads to
good radiation efficiencies is [2]:

978-0-7695-3648-4/09 $25.00 © 2009 IEEE 720

DOI 10.1109/AMS.2009.12

Authorized licensed use limited to: National Institute of Technology- Delhi. Downloaded on February 05,2021 at 11:20:49 UTC from IEEE Xplore. Restrictions apply.
 where vo is the free-space velocity of light.
Figure 1: Rectangular Microstrip Antenna
The actual length of the patch:
(5)
where ∆L is the extension of the length due to the
fringing effects and is given by:
2. Design for Microstrip Antenna:
In this work, the patch geometry of the
microstrip antenna is obtained as a function of input
variables, which are height of the dielectric material (h),
dielectric constants of the substrate material (εr, εy),
and the resonant frequency (fr), using ANN techniques
(Figure 2). Similarly, in the analysis ANN, the resonant
frequency of the antenna is obtained as a function of
patch dimensions (W, L), height of the dielectric
substrate (h), and dielectric constants of the material
(εr, εy) (Figure 3). Thus, the forward and reverse sides
of the problem will be defined for the rectangular patch
geometry in the following subsections.
2.1 The forward side of the problem: The synthesis
ANN
The input quantities to the ANN black-box in
synthesis (Figure 2) can be ordered as:
• h: height of the dielectric substrate;
• εr, εy : electrical properties of the dielectric substrate,
where εr, εy are the permittivity in the x and y directions
of the dielectric material used in the system,
respectively;
• fr : resonant frequency of the antenna.
The following quantities can be obtained from the
output of the black-box as functions of the input
variables:
• W : width of the patch;
721
Authorized licensed use limited to: National Institute of Technology- Delhi. Downloaded on February 05,2021 at 11:20:49 UTC from IEEE Xplore. Restrictions apply.
• L: length of the patch.
Figure 2 The Synthesis of ANN model
2.2. The reverse side of the problem: The analysis
ANN
In the analysis side of the problem,
terminology similar to that in the synthesis mechanism
is used, but the resonant frequency of the antenna is
obtained from the output for a chosen dielectric
substrate and patch dimensions at the input side as
shown in figure no. 3.
Figure 3 The Analysis of ANN model
2.3 Building Neural Networks for Rectangular
Microstrip Antenna:
In this work, both Multilayer Perceptron
(MLP) and Radial Basis Function (RBF) networks are
used in ANN models. Tthe structures of these ANNs
are described briefly below.
Multilayer Perceptron Networks
MLP are feed forward neural networks trained
with the standard back propagation algorithm. They are
supervised networks and so they require a desired
response to be trained. They learn how to transform
input data into a desired response, and so they are
widely used for pattern classification. With 1 or 2
hidden layers, they can approximate virtually any input-
output map. They have been shown to approximate the
performance of optimal statistical classifiers in difficult
problems. Most neural network applications involve
MLP. The basic MLP building unit is a simple model of
artificial neurons. This unit computes the weighted sum
 of the inputs plus the threshold weight and passes this
sum through the activation function (usually sigmoid).
In a multilayer perceptron, the outputs of the units in
one layer form the inputs to the next layer. The weights
of the network are usually computed by training the
network using the back propagation algorithm [12].

Structures of the neural networks

The MLP network, which has a configuration
of 2 input neurons, 10 and 5 neurons in 2 hidden layers,
and 2 output neurons with learning rate = 0.1, goal =
0.001, was trained for 100 epochs. Hyperbolic tangent
sigmoid and linear transfer functions were used in MLP
training. MLP models were trained with almost all
Figure 4: Rectangular Microstrip patch antenna With
network learning algorithms. In the RBF network, the Microstrip Line Feed for 4.98 Ghz
spread value was chosen as 0.01, which gives the best Simulation is performed using IE3D electromagnetic
accuracy. Both MLP and RBF were trained with 45 simulator where we obtained different graph between
samples and tested with 15 samples determined return loss and frequency. The dimensions of design are
according to the definition of the problem; 5 inputs and
taken in m.m. as: W=17.8, L=13.35, εr=4.7,
1 output were used for the analysis ANN and 4 inputs h=1.588. The return loss of port 1 is given in table No.1
and 2 outputs for the synthesis ANN. The training and and graph is plotted in figure No. 5.
test data of the synthesis and analysis ANN were
obtained from both experimental results given in
previous works [6]. The data are in matrix form
consisting of inputs and target values and arranged
according to the definitions of the problem.

3. IMPLEMENTATION & RESULTS

Design of the rectangular microstrip patch

antenna at 4.98 Ghz resonant frequency using the
following parameters is implemented and results are
plotted according to the results obtained.
Dielectric constant = 4.7
Substrate thickness = 1.588 mm
Now we can calculate the length and width of patch
Figure 5 Frequency Versus Return loss graph (4.98
using the general relations. Length and width of the
Ghz)
patch for 4.98 Ghz frequency is calculated, given by
L=13.35
Here 4.93 Ghz is the upper cutoff frequency and 5.03
W=17.8
Ghz is the lower cutoff frequency. In the Next step we
Now we can determine the position of feed point by
have changed the length of the patch while keeping the
using IE3D Electromagnetic Simulator. Microstrip
width constant and Vice –versa and note down the
patch with microstrip line feed is shown in figure 4.
resultant upper and lower cutoff frequencies (f1, f2).
The width Wo of microstrip line is calculated using the
Again with the help of IE3D Electromagnetic Simulator
following relationship.
obtained the various graphs between frequency and
Zo= 120Π / [√(εreff)
return losses. With dimensions in m.m. as: W=17.8,
{Wo/h+1.393+.667ln(Wo/h+1.444)}]
L=13.55, εr=4.7, and h=1.588 plot is shown in figure 6
Here
and in table 2.
Zo = 50 Ω load,
Wo = width of the microstrip line
h = thickness of the substrate
εreff = effective dielectric constant
Here
Wo = 2.96 mm and width of the hole section is 4.56
mm

722

Authorized licensed use limited to: National Institute of Technology- Delhi. Downloaded on February 05,2021 at 11:20:49 UTC from IEEE Xplore. Restrictions apply.
 4. ANALYSIS OF NEURAL NETWORK FOR
MICROSTRIP PATCH ANTENNA

As shown in the above neural network architecture, it

Figure 6: Frequency Vs Return loss graph (4.91 Ghz)
As mentioned above with the help of IE3D
electromagnetic simulator we obtained the different
frequency range for different dimensions. Results are
given in the following table.
εr =4.7, h=1.588
In the next step we have built the neural network
for rectangular microstrip patch antenna [14]. For the
analysis problem of neural network we will take width
(W) & length (L) at the input side of ANN black box
while upper and lower frequencies are taken as output
quantities. ANN is trained using training algorithms and
transfer function so that error can be minimized. Here
we have used Matlab neural network GUI toolbox.
While working on GUI, type the command” nntool” in
the command window. Then following Network/Data
manager window appears
consists of three layers.
The three-layer network has one output layer (layer 3),
one hidden layer (layer 2) and one input layer (layer 1).
Input and output layer consists of two neurons .inputs
W & L are applied at the input neurons while outputs f1
and f2 are obtained from output neurons. A constant
input 1 is fed to the biases for each neuron. Note that
the outputs of each intermediate layer are the inputs to
the following layer. A layer that produces the network
output is called an output layer. Below Figure shows
the neural network training graph results for the
analysis of microstrip patch antenna. it is clear from this
figure ,training performs in 21 epochs . We pass the full
set of input samples through the neural network to
compute the least squared error function we will use in
the back propagation of the errors step. Each such pass
is called an epoch. From this figure it is clear that error
minimizes from 101 to nearly 10-4.

Figure 8: Neural network architecture for microstrip

patch antenna

Figure 7 Network/Data manager window

Now in the next step click on the new data, and

substitute the input & target quantities in matrix form.
After that we have to create a neural network by
clicking on New network. Here we are applying
feedforward backpropagation algorithm and Trainlm
(Levenberg –Marquardt back propagation training
function).and supported transfer functions are purelin Figure 9: training graph results for analysis of
(Linear transfer function) and tansig (hyperbolic microstrip patch antenna
tangent sigmoid transfer function).

723

Authorized licensed use limited to: National Institute of Technology- Delhi. Downloaded on February 05,2021 at 11:20:49 UTC from IEEE Xplore. Restrictions apply.

5. SYNTHESIS OF NEURAL NETWORK FOR
MICROSTRIP PATCH ANTENNA
for the synthesis problem of neural network we will
take upper and lower frequencies at the input side of
ANN black box while width(W) & Length(L) are taken
as output quantities. We have trained the neural
network by using training algorithms and transfer
function.
Figure shows the neural network training graph results
for the synthesis of microstrip patch antenna. it is clear
from this figure ,training performs in 100 epochs . We
pass the full set of input samples through the neural
network to compute the least squared error function we
will use in the back propagation of the errors step. Each
such pass is called an epoch. From this figure it is clear
that error minimizes from 102 to nearly 10-2 .
Figure 10: training graph results for synthesis of
microstrip patch antenna
6. CONCLUSION:
In this design procedure, synthesis is defined
as the forward side and then analysis as the reverse side
of the problem. Therefore, one can obtain the geometric
dimensions with high accuracy, which is the length and
the width of the patch in our geometry, at the output of
the synthesis network by inputting resonant frequency,
height and dielectric constants of the chosen substrate.
Furthermore, in our work, the synthesis can also be
applied into anisotropic dielectric substrate. In this
work, the analysis is considered as a final stage of the
design procedure, therefore the parameters of the
analysis ANN network are determined by the data
obtained reversing the input-output data of the synthesis
network. Thus, resonant frequency resulted from the
724
Authorized licensed use limited to: National Institute of Technology- Delhi. Downloaded on February 05,2021 at 11:20:49 UTC from IEEE Xplore. Restrictions apply.
synthesized geometry is examined against the target in
the analysis ANN network. Finally, in this work, a
general design procedure for the microstrip antennas
and filter is suggested using artificial neural network.
REFERENCES
[1] I.J. Bahl, P. Bhartia, “ Microstrip Antennas”, Dedham,
MA, Artech House, 1980.
[2] D.M. Pozar, “Microstrip Antennas”, Proc. IEEE, Vol. 80,
pp.79-81, January, 1992.
[3] S¸. Sa˘gıro˘glu, K. G¨uney, “Calculation of resonant
frequency for an equilateral triangular microstrip antenna
using artificial neural Networks”, Microwave Opt.
Technology Lett., Vol. 14, pp. 89-93, 1997.
[4] D. Karabo˘ga, K. G¨uney, S¸. Sa˘gıro˘glu, M. Erler,
“Neural computation of resonant frequency of electrically thin
and thick rectangular microstrip antennas”, Microwaves,
Antennas and Propagation, IEE Proceedings-Vol. 146, No. 2,
pp. 155 – 159, April 1999.
[5] K. G¨uney, S¸. Sa˘gıro˘glu, M. Erler, “Generalized neural
method to determine resonant frequencies of various
microstrip antennas”, International Journal of RF and
Microwave Computer-Aided Engineering, Vol. 12, No. 1, pp.
131-139, January 2002.
[6] 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.
[7] 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.
[8] K. G¨uney, N. Sarıkaya, “Artificial neural networks for
calculating the input resistance of circular Microstrip
antennas”, Microwave and Optical Technology Letters, Vol.
37, No. 2, pp. 107-111, 20 April 2003.
[9] G. Angiulli, M. Versaci, “Resonant frequency evaluation
of microstrip antennas using a neural-fuzzy approach”,
Magnetics, IEEE Transactions, Vol. 39, No. 3, pp. 1333 –
1336, May 2003.
[10] R.K. Mishra, A. Patnaik, “Neurospectral computation for
input impedance of rectangular microstrip antenna”,
Electronics Letters, Vol. 35, No. 20, pp. 1691 – 1693, 30
Sept. 1999.
[11] R.K. Mishra, A. Patnaik, “Designing rectangular patch
antenna using the neurospectral method”, Antennas and
Propagation, IEEE Transactions, Vol. 51, No. 8, pp. 1914 –
1921, Aug. 2003.
[12] Q. J. Zhang, K. C. Gupta, Neural Networks for RF and
Microwave Design, Artech House Publishers, 2000.