Técnicas de inteligencia computacional para el diseño robusto de dispositivos radiantes

Laburpena

The doctoral thesis proposes a new antenna with a suitable structure for the main frequency bands of mobile services, whose optimization process has been done by an Evolutionary Programming algorithm. The growth of the demand for high speed data services has generated the necessity of increasing the antenna bandwidth. Due to its reduced size and low specific absorption rate (SAR), planar inverted-F antennas (PIFA) and patch antennas are typically employed in the wireless services. The main disadvantage of these devices is their narrow bandwidth, for this reason these antennas are not useful for the new generation mobile handsets. Nevertheless, when PIFA and patch antenna are employed in the same structure, the bandwidth is increased remarkably, and the dimensions are reduced with respect to their original sizes. The proposed antenna is implemented in a variation of microstrip line, which is called coplanar waveguide (CPW). A low-cost FR4 substrate has been used. As the behavior of this substrate is poor at high frequency, a theoretical and practical analysis of the prototype has been done to determine the adverse effects around 5.6GHz, since it is the frequency of the WIFI UNII service. In the theoretical study, the antenna has been simulated with different loss tangent materials. In the experimental analysis, two versions of the antenna have been made and compared, one on low cost FR4 substrate and the other on low loss Rogers 4350 material. The simulated and measured results show that the behavior of the antenna with a low cost substrate (FR4) and with a low loss tangent substrate (Ro4320) is very similar. In summary, this thesis proposed an antenna which is based on the hybridization of two single antennas: a PIFA and a circular patch antenna with a CPW feeding. As there is not an accurate expression for the proposed design, an optimization process must be carried out. A parametric study has been done before the optimization process in order to know the parameters that need to be modified to increase the bandwidth. This study shows that the variables to be changed during the optimization process are six: length (Lg) and width (Wg) of the ground plane, length (B) and width (A) of the narrow side in the PIFA radiation plane, patch antenna radius (ri) and radius of the circumference between patch antenna and ground plane (ro). Finally, the overall antenna dimensions are 57 x 24 x 6 mm and for 10 dB return loss the operating bandwidth are 49% in the low frequency band (1800MHz) and 11.25% in the high frequency band (5200MHz), the gain values are 3 and 4.2dBi respectively. Computational Intelligence techniques have been employed during the optimization process. The results obtained after the optimization show that the proposed antenna can be used in GSM, DCS, UMTS and WLAN, as well As in other wireless services like GPS and Bluetooth which are very common in the new generation handset. For comparison, other classical technique such as Nelder Mead Simplex method has been used during the optimization. This technique is included in the optimization section of the electromagnetic simulator employed to obtain the performance of the antenna (CST Microwave Studio). This is a very robust method but, the main problem of this algorithm is that it becomes very instable when using more than four parameters. With this method the bandwidth in the low frequency band is improved but in the high frequency it is deteriorated. The optimization criterion in the electromagnetic simulator doesn’t let us to adjust independently both bands. So, it is necessary to include an external optimization method to adjust both bands separately. The use of an external optimization method, like Evolutionary Programming algorithm, as a part of the antenna’s evaluation process, involves the definition of a communication protocol between the simulator and the program employed to implement the evolutionary algorithm. The proposed optimization process may be carried out with any other computational intelligence technique.