15-10-2012, 05:39 PM
BANDWIDTH ANALYSIS BY INTRODUCING SLOTS IN MICROSTRIP ANTENNA DESIGN USING ANN
BANDWIDTH ANALYSIS.pdf (Size: 241.43 KB / Downloads: 57)
Abstract
Many applications of microstrip antenna are rendered by
their inherent narrow bandwidth. In this paper, a new approach is
proposed to design inset feed microstrip antenna with slots in it to
improve the antenna bandwidth. This paper deals with the design
of slotted microsrip antenna on a substrate of thickness 1.588mm
that gives wideband characteristics using ANN. The illustrated patch
antenna gives enhanced bandwidth as compared to antenna with out
slots of the same physical dimensions. In the present work an Arti¯cial
Neural Network (ANN) model is developed to analyse the bandwidth
of the example antenna. The Method of Moments (MOM) based
IE3D software has been used to generate training and test data for
the ANN. The example antenna is also designed physically with glass
epoxy substrate with "r = 4:7 for few results for testing the arti¯cial
neural network model. The di®erent variants of training algorithm of
MLPFFB-ANN (Multilayer Perceptron feed forward back propagation
Arti¯cial Neural Network) and RBF-ANN (Radial basis function
Arti¯cial Neural Network) has been used to implement the network
model. The results obtained from arti¯cial neural network when
compared with experimental and simulation results, found satisfactory
and also it is concluded that RBF network is more accurate and fast as
compared to di®erent variants of backpropagation training algorithms
of MLPFFBP.
INTRODUCTION
Arti¯cial Neural Networks (ANNs) are suitable models for microwave
circuit optimization and statistical design. Neuro models are
computationally much more e±cient than EM models. Once they are
trained with reliable learning data obtained from a \¯ne" model by
either EM simulation or measurement, the Neuro models can be used
for e±cient and accurate optimization and design within the region
of training. The Arti¯cial neural networks (ANNs) provide fast and
accurate models for microwave modeling, simulation, and optimization.
The past decades has seen a phenomenon growth in the development
of new tools for microwave CAD. ANNs are computational tools that
learn from experience (training), generalize from previous examples to
new ones, and abstract essential characteristics from input containing
irrelevant data. ANN application to the ¯eld of microwaves is very
recent. A number of papers [1{10, 12] indicates how ANNs can be used
e±ciently in analysing and synthesizing various microwave circuits.
Multi Layer Perceptron Feed Forward Back
Propagation (MLPFFBP)
MLP networks are feed forward networks trained with the standard
back propagation algorithm as shown in Figure 4. They are supervised
networks, and also they required a desired response to be trained. With
one or two hidden layers they can approximate virtually any input
output map. The weights of the network are usually computed by
training the network using the back propagation algorithm.
In this paper, three layer multilayer feed-forward back propagation
arti¯cial neural network with one hidden layer and trained by di®erent
variants of back propagation training algorithms is used to design
microstrip antenna. ANN structure (number of layers, number of
neurons in each layer, neurons activation function, learning algorithm
and training parameters) is not known in advance. Hence the network
model is analysed with di®erent number of hidden layers in the
structure and also the numbers of processing elements are also varied to
acquire the accuracy. Hence it is concluded that three layer MLP with
one hidden layer and 25 processing elements in the hidden layer is the
optimum network structure for the proposed problem. The network
is trained with 5 di®erent training algorithms to achieve the required
degree of accuracy and hence compared for network performance.
RBF Networks
Radial basis function network is a feed forward neural network with
a single hidden layer that uses radial basis activation functions for
hidden neurons. RBF networks are applied for various microwave
modeling purposes. The RBF neural network has both a supervised
and unsupervised component to its learning. It consists of three layers
of neurons | Input, hidden and output. The hidden layer neurons
represent a series of centers in the input data space. Each of these
centers has an activation function, typically Gaussian. The activation
depends on the distance between the presented input vector and the
centre. The further the vector is from the centre, the lower is the
activation and vice versa. The generation of the centers and their
widths is done using an unsupervised k-means clustering algorithm.
The centers and widths created by this algorithm then form the weights
and biases of the hidden layer, which remain unchanged once the
clustering has been done. A typical RBF network structure is given
in Figure 5. The parameters cij and "ij are centers and standard
deviations of radial basis activation functions.
Conclusion
The inset fed microstrip patch antenna is a versatile structure which
can be modi¯ed by the addition of simple slots in the design
structure to overcome selected limitations inherent to conventional
patch antennas. The antenna can provide improved bandwidth
enhancement, under certain conditions, while maintaining many of the
desirable features of conventional patches. It must be emphasized that
improving a particular characteristic of the antenna may result in the
degradation of one or more of the other performance characteristics.