Estimation of resonant frequency and bandwidth of compact unilateral coplanar waveguide‐fed flag shaped monopole antennas using artificial neural network

Abstract

Neural network based estimation of resonant frequency and bandwidth of compact unilateral coplanar waveguide (CPW)‐fed flag shaped printed monopole antennas is presented. The proposed antennas are similar to CPW‐fed antenna; however, by replacing unilateral CPW feed instead of CPW feed, compactness of about 48.615 percent is achieved. These are designed on an inexpensive FR4‐epoxy substrate with dielectric constant of 4.4 and thickness of 1.6 mm. Resonant frequencies and bandwidths of the flag shaped antennas with different dimensions are computed using method of moment electromagnetic solver IE3D 15.10 and they have been given as training and test data for the proposed multilayered perceptron feed forward neural network with Levenberg–Marquardt training algorithm to estimate the resonant frequency and bandwidth of the proposed antennas. The estimated values of resonant frequency and bandwidth with average percentage of error are 1.275 and 0.325, respectively. For verification, the proposed antenna is fabricated and measured with resonant frequency 5.78 GHz and bandwidth of 1.0 GHz for HiperLAN/2 and IEEE802.11a applications. It has impedance bandwidth from 5.0 to 6.0 GHz for return loss lower than −10 dB and good omnidirectional radiation performance over entire frequency range, with a compact size of 11.07 × 27.5 × 1.6 mm3. Various features such as compactness, simple geometry, and low cost, uniplanar structure make the antennas suitable for modern wireless communication systems. This approach replaces the use of very complicated analysis. © 2015 Wiley Periodicals, Inc. Microwave Opt Technol Lett 57:337–342, 2015

Prediction of Slot-Shape, Slot-Size and Inserted Air-Gap of a Microstrip Antenna Using Knowledge-Based Neural Network

“…The inputto-hidden layer weights are denoted as: W ji for 1 ≤ i ≤ m and 1 ≤ j ≤ n and the hidden-to-output layer weights as: W qj for 1 ≤ q ≤ k and 1 ≤ j ≤ n. The excitation and response of the model are represented as: e i for 1 ≤ i ≤ m and r q for 1 ≤ q ≤ k. The model within a box of dotted line is basically known as a MLP neural model which has extensively used in the literature [7][8][9][10][11][12]. The accuracy of this type of model depends on the number of training patterns generated by simulation/measurement, and it increases by increasing the number of training patterns.…”

“…Once the training is over, the trained KBNN model then predicts the slot-shape (i.e., LS-Antenna, TS-Antenna, CSAntenna or SS-Antenna), slot-size (i.e., x 1 , y 1 , x 2 and y 2 ) and thickness of the inserted air-gap (i.e., A g ) within a fraction of a second for any arbitrary set of parameters: 1.5 GHz ≤ (f 1 and f 2 ) ≤ 3.0 GHz, 6.2 dBi ≤ (G 1 and G 2 ) ≤ 9.6 dBi, 6.6 dBi ≤ (D 1 and D 2 ) ≤ 9.9 dBi, 83% ≤ (A 1 and A 2 ) ≤ 100% and 85% ≤ (R 1 and R 2 ) ≤ 100%. For better understanding of the KBNN modeling, a conventional MLP model of structural configuration 10 * 78 * 76 * 6 using the approach as described in the literature [7][8][9][10][11][12] is also created. The prediction of Slot-Shape (S), Slot-Size (x 1 , y 1 , x 2 , y 2 ) and inserted Air-Gap (A g ) is summarized using block-diagram in Figure 3.…”

“…Thus, these obligations in conventional synthesis methods lead to complexities and processing cost. For fulfilling such obligations, neural networks modeling could be a better alternative [6][7][8][9][10][11][12]. In the last decade, artificial neural networks (ANN) modeling has acquired enormous importance in the microwave community, especially in modeling of MSAs due to their ability and adaptability features.…”

“…In the last decade, artificial neural networks (ANN) modeling has acquired enormous importance in the microwave community, especially in modeling of MSAs due to their ability and adaptability features. The neural models [7][8][9][10][11][12] may not be very reliable if they are trained with small number of patterns. Also, even with sufficient training patterns, the reliability of these models is not guaranteed in extrapolation region, and in most cases, it fails.…”

“…In this paper, the concept of KBNN modeling has been used for predicting slot-shape, slot-size and inserted air-gap, simultaneously. To the best of the authors' knowledge, there is no published work for modeling such a complicated problem in the referenced literature [7][8][9][10][11][12] and [13][14][15][16][17][18][19][20][21][22][23]. Handling such a complicated problem using analytical/numerical techniques is still a challenging task facing the electromagnetic community.…”