An array antenna that conforms to some part of an airplane, high-speed train, or other vehicle is called a conformal array antenna. Usually, a conformal array antenna has cylindrical, spherical or some other shape, with the radiating elements mounted on the curved surface. The antenna becomes integrated with the structure and does not cause extra drag so that it has the advantage of the aerodynamic performance. In addition, the antenna is placed in a proper location according to the functional requirements of radar and electronic equipment.
As the need for conformal array antennas is consistently increased recently, various techniques including local optimization algorithm and global optimization algorithm on the pattern synthesis of conformal array antennas have been extensively studied. In particular, the evolutionary algorithms derived from biologic behavior have attracted more attention because of their simple structure, efficiency, and suitability for solving multi-parameter and multi-goal problems. Though the genetic algorithm(GA) and the particle swarm optimization(PSO) are well known evolutionary algorithms, these algorithms show the premature convergence and the slower convergence speed in high dimensional problems. To overcome these problems, Boeringer proposed the adaptive genetic algorithm(AGA) by using variable probability in algorithm parameters. However, when all individuals of the population show similar genetic characteristic due to the elite selection technique, the AGA can still suffer from the premature convergence in the high dimensional problem.
In this dissertation, investigations on an enhanced adaptive genetic algorithm(EAGA) for pattern synthesis of a conformal array are presented. The proposed algorithm is the modified AGA for fast convergence speed combined with the conventional IWO(Invasive Weed Optimization) technique for improved optimized fitness. In the IWO, not only the best but also the worst individuals are utilized in order to obtain more competitive or less similar genetic characteristic. In addition, in the proposed EAGA, the population size is increased instead of being fixed according to the cost value of various individuals in order to increase the solution-finding probability near the global optimum.
To compare the efficiency of the EAGA with that of the AGA, a typical test function, Rastrigrin function, which has a large number of local minima and global minima of zero, is applied, and values of the test function are compared. The superiority of the EAGA to the AGA is confirmed.
As examples for appling the proposed algorithm to various types of surfaces for synthesizing the radiation patterns, array antennas on two different curved surfaces are considered in this dissertation. One has a curved cylindrical surface with quadratic function, and the other has a rotated-type surface of quadratic function.
As a first example a 4×16 conformal array antenna realized on a curved cylindrical surface is designed. This antenna operates as a 1×16 array in the azimuth direction and is controlled by the 5-bit attenuators and 5-bit phase shifters. To verify the validity of the EAGA, the value of the cost function and the synthesized pattern in case of the EAGA is compared with those in case of the AGA. The pattern synthesized by the EAGA also meets the design specification. The measured results of the fabricated 4×16 array antenna are compared with the synthesized results simulated by MATLAB and CST’s MWS.
As a second example, rotated-type conformal array antennas having various coefficients of quadratic function are considered. The patterns of the rotated-type conformal arrays having three different coefficients of quadratic function are synthesized using the EAGA, respectively, which follows a concentric planar array synthesis. In order to simplify the pattern syntheses of the various rotated-type antennas, the transformed active element pattern of the concentric array is used instead of the active element pattern of each rotated-type antenna. The synthesized patterns approximatively meet the design specifications and the comparison between MATLAB pattern synthesis results and the MWS results.
All these results show that the enhanced algorithm proposed in this dissertation is useful for pattern synthesis of conformal array antenna and is expected to achieve the optimum design for various curved surfaces in an effective manner.