Neural network modeller for design optimisation of multilayer patch antennas

The development of a neural network modeller for the optimisation of a dual-band multilayer microstrip patch antenna is presented. The multilayer antenna is built on an in-house semi-insulating GaAs substrate and poses a challenge to electromagnetic-based optimisation modelling. The developed neural network model allows some of the optimisation processes of the design to be carried out swiftly by replacing repeated electromagnetic simulations whilst still retaining a good accuracy as compared with method of moments modelling.