Prediction of S parameters of hybrid meander structure with multilayer perceptron neural network

Abstract

Hybrid meander structures are usually investigated using numerical and analytical methods, which are based on Maxwell's equations. These methods require a lot of computational time. Especially it is important when it is necessary to repeat the calculations until the desired structure of meander is found. Therefore, the analysis of the hybrid meander structures using intelligent methods based on the feedforward backpropagation multilayer perceptron networks is presented in this paper. The selection of structure of the network is one of the most important tasks during the investigation. Therefore, the number of hidden layers and the number of neurons in every hidden layer were varied in order to find the optimal structure of the network for the most accurate prediction of S21 parameter. The obtained optimal construction of the network consisted of three hidden layers with 18, 18 and 9 neurons respectfully. The difference between calculated and predicted cut-off frequency results using the optimal configuration of the network do not exceed 63 MHz. The duration of the calculations has been shortened approximately for 180 times from 90 minutes till less than 0.5 s.