Microstrip antenna optimization using artificial neural network

Abstract

In the modern era of wireless communication there is a need for an antenna with reduced size, higher bandwidth (to support high data rate), fewer losses and ability to operate at high frequency. The microstrip antenna is one of the most favorable candidates in this context and is compatible with today’s wireless scenario because of their smaller size, ability to operating at high frequency, and ease of installation. The microstrip antenna is used in mobiles, satellite communication devices, radars etc. Because of the versatility of the microstrip patch, a lot of research work have been done and is going on today. The tool(software) used for designing these antennas in today’s scenario is the EM simulation software. Although it is precise, it is very time consuming and complex example of which are CST(computer simulation technique), IE3D etc. The primary objective of the thesis is to use a neural network as a tool to reduce the consumed time and computational complexity in designing of an antenna. The Neural network has favorable properties in this context including the ability to highly non linear input output relationships, the property of generalization and less time consuming. It is turning out to be a very efficient tool in Antenna optimization. There are different kinds of neural techniques that can be used based on the application. The most commonly used neural techniques are feedforward with back propagation algorithm, which operates on the gradient descent method, and the RBF (radial basis function) NN, which work based on the principle of distance between the input and the weight vector. These two techniques are widely used and accepted in the application of the neural network in the field of antenna. The thesis has two main objectives. The first one is to optimize the microstrip patch dimension using the neural network and, the second one is to predict the directivity of a planar array by using neural network. The size of the antenna plays a central part in today’s wireless scenario because of less space availability and the portability issue. The work is present here to optimize the dimension of microstrip patch using RBF neural network. Further, in order to make use of neural network more versatile the neural network is applied to a microstrip planar array for the purpose of directivity estimation. Since directivity plays a key role to steer a beam in a particular iv direction, so to enhance directivity a microstrip planar array is designed, then to estimate the EM simulated directivity the Feed forward and RBF neural network is utilized. It is found that RBF neural network produces favorable results but the Feed forward network with back propagation algorithm produces more accurate and precise optimization.