Performance Tuning of Spade Card Antenna using Mean Average Loss of Backpropagation Neural Network

The microstrip antennas have different dimensions to get the desired performance, especially for microstrip antennas that have complex components and dimensions with the performance: the range of frequency at 2.4 GHz until 3.6 GHz, Maximum Power of Gain value is 5.83 dB and the minimum value is 3 dB and Maximum Directivity Value is 6.22 and the minimum value is 3.32. in consequence, needs to fill the demand for a new and the corresponding design as solvent to adaptive matching as tuner the frequency on antenna design that needs requires a complex mathematical method and simulation. This paper has the novel design to tune the performance of spade card microstrip antenna that can operate on the single, dual or multiband and able to produce circular or linear polarization using Backpropagation Neural Network in order to obtain an optimum design with a backpropagation algorithm as a solution to simplify the design process. As a result, after 20000 epochs the training loss is around 0.044 and the testing loss is around 0.058. The model has a good performance despite only using a few numbers of training data. Keywords—Spade card antenna; mean average loss; neural network; performance tuning antenna


I. INTRODUCTION
The progress of telecommunication technology has developed very rapidly. This can be seen from the increasing need to obtain information, whether the information in the form of sound, data, pictures, or video, with communication equipment that can be used anywhere and anytime. The most developed technological advancement today is wireless communication, which can support the implementation of a global telecommunication system. One example of the application of wireless communication is WLAN (Wireless Local Area Network) that can meet the needs of access to information and communication that can be used anywhere and anytime [1]. Compared to the Wired Local Area Network that used to use the cable as its transmission medium, WLAN technology is easier in terms of installation, practical and efficient [2].
In the application of wireless communication, the role of the antenna as an electromagnetic component is necessary because of the antenna attend as a means to emit and take electromagnetic waves in which the equipment formation signal is contained. The antenna is a very important component to support the wireless communication system [3]. The latest developments in wireless communication systems require antenna characteristics that have relatively small, flexible and practical shapes.
The microstrip antenna is one kind of the antenna with the exact quality of the required telecommunication circuit. The microstrip antenna is made on a specific substrate material with a radiating element located on one side of the substrate and the other is a conductor layer that acts as a ground plane [4]. Microstrip antennas work on UHF frequency allocations (300 MHz -3000 MHz) up to X Band (5200 MHz -10900 MHz) so that microstrip antennas can be used for wireless communication.
Microstrip antennas have different components and dimensions to get different performance after what is needed, especially for microstrip antennas that have complex dimensions and components [5]. in consequence, needs to fill the demand for a new and the corresponding design as solvent to adaptive matching as tuner the frequency on antenna design that needs requires a complex mathematical method and simulation. This paper has the novel design to tune the performance of spade card microstrip antenna that can operate on the single, dual or multi-band and able to produce circular or linear polarization using Backpropagation Neural Network [6] [7]. The model has a good performance despite only using a few numbers of training data.

A. Antenna Design
The antenna constructed in this paper is a microstrip circuit antenna with the following component specifications: The dielectric value (εr) = 4.2 The dielectric components thickness (h) = 1.6 mm  Seam substrate copper : The conductor material thickness (t) = 0.000001 mm Copper conductivity (σ) =5.80x107mho m-1 Material size = 29.7 x 21 mm After mathematically calculated this research did a simulation and optimization to get the expected results using www.ijacsa.thesai.org CST software. The shape of the ground plane and patches are made to have different sizes to obtain the desired frequency [8]. Shape design and simulation of this antenna is important because it will affect the antenna performance. The main part of the antenna is a triangle shape joined with a double circle at the bottom that resembles the shape of a reversed heart, shown in Fig. 1. Part circle on the patch directly connect the feeder line as a link to the patch on the antenna connector [9]. Then the ground plane consists from one form and forms ground plane is basically that is rectangular in addition to the connection with the connector.
In this paper, basically the patch of the antenna has the circular shape, so the component of patch element dimensions with the equation: The notation above can be explained that a is the radius of the antenna patch component, h is the substrate thickness, r is the dielectric permittivity of the substrate and F is the logarithmic function. The transmission line circuit of microstrip element design is finished theoretical based on the matter and any literature [10] [11].

B. Datasets
The data used in this paper have been obtained from the CST design results. The antenna simulation using a different combination of ground plane height and the triangle and circle dimensions resulted in 123 samples which have been classified to be more precise. in this research, picked random 90 samples which have been measured and used it as the training data and the rest as the testing data [12] [13]. The frequency, return loss, gain and directivity will be saved for further use.

C. Network Achitecture
The implementation uses a three-layer of a Multilayer Perceptron in this paper. In Fig. 2, The input layer has only one node as the frequency, 16 nodes inside the hidden layer to add a non-linearity to model and 6 nodes in the output layer as the representation of the ground plane height, triangle side length, circle diameters, return loss, gain and directivity of the antenna [14][15]. The algorithm was experimented using a different-numbers of hidden layers and neurons [16] [17]. However, the architecture explained above is the best possible combination and has the best result in minimizing the loss and converge faster than any other combination.

D. Implementation
The training process of backpropagation network consists of two main step. First, training features will be fed to the input layer and will flow through all neuron in the hidden layer. The interconnection between neuron is called synapse. All synapse in the network has its own weight [18] [19]. The features will be multiplied with its synapse weight and add a bias value at the same time. After the features arrive at the output layer, the loss will be computed using Mean Average Loss (MAE) (1). Second, the loss value will be minimized using stochastic gradient descent. The loss value will be backpropagated to compute the gradient to minimize the loss and update all synapse weights.

∑| ̂ |
Neural network model was created using Python and Tensor flow framework. The activation function of all neuron inside the hidden layer are ReLU and the activation function of the output are Linear, because the method do regression task in this experiment [20][21]. Stochastic Gradient Descent (SGD) used as the optimizer with learning rate of 0.0001 and repeated until 20.000 epochs. Training data will feed into the network in batch of 10 data. Thus to run experiments, we compare the results of several experimental samples with the accuracy of the prediction algorithm used. Thus, to measure the success of www.ijacsa.thesai.org the implementation of the method used by comparing the results of experiments and comprehensive algorithm predictions.

III. RESULT AND ANALYSIS
After 20.000 Epochs, the analysis evaluates the model using the test data. The analysis evaluates the MAE value and compare that value with the loss of the training process. The training loss is around 0.044 and the testing loss is around 0.058.

IV. CONCLUSION
Backpropagation neural network has been used to design an antenna. By using a correct activation function and proper learning rate, backpropagation neural network shows a promising result for the tune of a spade card shape microstrip antenna: the range of frequency at 2.4 GHz until 3.6 GHz, Maximum Power of Gain value is 5.83 dB and the minimum value is 3 dB and Maximum Directivity Value is 6.22 and the minimum value is 3.32. The model has a good performance despite only using a few numbers of training data. This research can be pushed further to learn using a different shape, substrate, frequency range and other parameters of a microstrip antenna. For future work it is necessary to develop a wider range of antenna frequencies and the need for a combination of algorithms to improve the accuracy of the results. www.ijacsa.thesai.org

APPENDIX
The appendix contains sample results from the results of experiments that have been carried out. The sample provides an overview of the accuracy of the prediction algorithm used. Thus algorithmic implementation can be implemented more efficiently in the various antenna sizes that are proposed. From some of the samples displayed, all value is the good agreements that based on existing minimum standard parameters.