in the second case, s 2.2, H s 1.5 mm. We can use the r neural network independently of the optimization phase.
Now that the optimization of the antenna has begun, the GA can be used with the neural network in the variation range.
Parameters of the genetic algorithm:
ⅷ Population: 20 chromosomes ⅷ Generation: 100 ⅷ Coding parameters: 16 bits ⅷ Chromosome length: 48 bits.
A number of generations equal to 40 is sufficient to obtain f s 15 GHz; this case is represented in Figure 7. A good 0 combination of parameters is quickly obtained. The genetic algorithm coupled with a neural network provide excellent results.
The antenna parameters computed in this way are: L s W s 8.416 mm, H s 1.759 mm, s 1. r More complex applications are presently under development in our laboratory. Figure 7 Frequency desired versus generations CONCLUSION Ž An original approach, coupling the FDTD replaced by a
. neural network and genetic algorithm, has been presented in this paper. A simple example shows the feasibility of the method. Neural networks permit an advantage in computing resources. This tool can be used for other applications, and in addition to the FDTD code.