No abstract

The design of a reflectarray for satellite applications is accomplished in this letter. The use of full-wave electromagnetic methods to compute the reflectarray elements in the design process has been replaced by the characterization of the reflectarray elements by artificial neural networks (ANNs). The antenna is designed to provide coverage in Ku-band to a European region defined by EuTELSAT. The results obtained using ANN in the design are compared to those obtained from the method of moments (MoM), and a good agreement is achieved. The speedup factor between reflectarray optimization by both the ANN and the MoM is in the order of 10 . Hence, the use of ANN is presented here as a promising fast technique in the design of reflectarrays.Index Terms-Artificial neural networks (ANNs), contoured-beam reflectarray, reflectarray, reflectarray design.

Abstract-This paper describes the accurate characterization of the reflection coefficients of a multilayered reflectarray element by means of artificial neural networks. The procedure has been tested with different RA elements related to actual specifications. Up to 9 parameters were considered and the complete reflection coefficient matrix was accurately obtained, including cross polar reflection coefficients. Results show a good agreement between simulations carried out by the Method of Moments and the ANN model outputs at RA element level, as well as with performances of the complete RA antenna designed.

No abstract

No abstract