摘要

This paper proposes a modification of the hybrid Taguchi-genetic algorithm (HTGA) for solving global numerical optimization problems with continuous variables. The HTGA is a method that combines a conventional genetic algorithm (CGA), which has a powerful global exploration capability, with the Taguchi method, which can exploit the optimum offspring. The Taguchi method is utilized in the HTGA to help in selecting the best genes in the crossover operations. The new implementation proposed in this paper (nHTGA) involves producing, at each generation, a single offspring by Taguchi method, one of its parents being the best individual found so far, instead of repeatedly applying Taguchi to generate several individuals with both parents selected at random as HTGA does. Moreover, the efficiency of the algorithm is enhanced by only crossing via Taguchi individuals with a high enough number of different genes. The performance of the proposed HTGA is assessed by solving several benchmark problems of global optimization with large number of dimensions and very large numbers of local minima. The computational experiments show that the new algorithm causes a reduction, sometimes drastic, in the number of function calls, i.e. in computational time, for all the benchmark problems proposed. As an example of application of this novel algorithm to a real-world problem, the optimization of an ultra-broadband zigzag log-periodic antenna is carried out and discussed.

  • 出版日期2009-2