Modelling of Non-Planar Microstrip Lines via Artificial Neural Networks

In order to meet the ever increasing needs in communication systems, in addition to software developments, the circuit stages need to be composed of innovative and high performance sub-layers. The most widely used transmission line structure in microwave circuit designs is microstrip structures. In order to realize a high performance and efficient circuit design, a highly accurate, fast and reliable model is required. In this study, it is aimed to estimate the value of the scattering parameters of the non-planar microstrip lines with respect to the geometrical design parameters, by using the Multilayer Perceptron. For this purpose, firstly the design of the non-planar microstrip transmission line in the 3D electromagnetic simulation program CST has been modelled. After that, data sets have been obtained for the training and test phases using the 3D CST model. Afterwards, the Multilayer Perceptron structure with different architectures was trained with this data and the network model with the lowest mean and standard deviation error was obtained. As can be inferred from the results, not only the use of MLP is an appropriate choice for modeling a high-accuracy, fast and reliable non-planar microstrip transmission line but it also open up an opportunity for fast, and accurate design optimization process of complex microwave stages using this model.