Triple band frequency selective surface design for 5G mm-wave communication with artificial neural networks

Abstract

High-performance frequency selective surfaces (FSSs) have gained attention for their spatial filtering characteristics in 5G communication systems. In this work, we propose an efficient and accurate design methodology for the FSS. Three different artificial neural network methods (ANN) are employed, and their performances are compared for analysis and synthesis purposes. Results show that GRNN has the highest performance for both training and test phase of ANN based FSS analysis and synthesis. A novel, compact, low-profile triple band FSS unit cell is introduced, and the working mechanism is described. By applying ANN based design procedure, the unit cell dimensions to resonate at the 5G mm-wave frequency band is extracted. A unit cell with the extracted physical dimensions is simulated with a full-wave analysis tool. The simulation results show that the FSS has the filtering feature at the predetermined mm-wave frequencies of the 5G communication. The prototype of the FSS is fabricated, as well. The simulations are verified experimentally with measurement results. The results show that proposed ANN based analysis and synthesis method can be an effective tool for the design of FSS band-pass filter for 5G applications.