In recent years, evolutionary algorithms have been successfully adopted for the optimization of various electromagnetic problems. One of the most common electromagnetic application is in the framework of microstrip antennas, thanks to the advantage of being low cost and low profile. In order to reduce the computational effort of the electromagnetic optimization, a suitable equivalent model by ANN has been created in order to substitute the commercially available full-wave analysis solvers. With the aim of reducing committed error level, a new solution of multiple neural networks instead of one network is presented. In addition, efficiency of new training scheme is also shown in Numerical results section. The effectiveness of proposed techniques will be illustrated by optimizing a particular type of antenna, namely proximity coupled feed.
Splitted neural networks for better performance of antenna optimization
GRIMACCIA, FRANCESCO;MUSSETTA, MARCO;ZICH, RICCARDO
2014-01-01
Abstract
In recent years, evolutionary algorithms have been successfully adopted for the optimization of various electromagnetic problems. One of the most common electromagnetic application is in the framework of microstrip antennas, thanks to the advantage of being low cost and low profile. In order to reduce the computational effort of the electromagnetic optimization, a suitable equivalent model by ANN has been created in order to substitute the commercially available full-wave analysis solvers. With the aim of reducing committed error level, a new solution of multiple neural networks instead of one network is presented. In addition, efficiency of new training scheme is also shown in Numerical results section. The effectiveness of proposed techniques will be illustrated by optimizing a particular type of antenna, namely proximity coupled feed.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
