A novel hybrid global optimization algorithm having training strategy: hybrid Taguchi-vortex search algorithm


SAKA M. , Coban M., EKE İ., TEZCAN S. S. , TAPLAMACIOĞLU M. C.

TURKISH JOURNAL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCES, vol.29, no.4, pp.1908-1930, 2021 (Journal Indexed in SCI) identifier

  • Publication Type: Article / Article
  • Volume: 29 Issue: 4
  • Publication Date: 2021
  • Doi Number: 10.3906/elk-2004-193
  • Title of Journal : TURKISH JOURNAL OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCES
  • Page Numbers: pp.1908-1930
  • Keywords: Hybrid Taguchi-vortex search algorithm, Taguchi orthogonal arrays, vortex search algorithm, global optimization, engineering design problems with constraints, NUMERICAL FUNCTION OPTIMIZATION, PARTICLE SWARM OPTIMIZATION, DIFFERENTIAL EVOLUTION

Abstract

In this paper, a novel hybrid Taguchi-vortex search algorithm (HTVS) is proposed for solving global optimization problems. Taguchi orthogonal approximation and vortex search algorithm (VS) are hybridized in presenting method. In HTVS, orthogonal arrays in the Taguchi method are trained and obtained better solutions are used to find global optima in VS. Thus, HTVS has better relation between exploration and exploitation, and it exhibits more powerful approximation to find global optimum value. Proposed HTVS algorithm is applied to sixteen well-known benchmark optimization test functions with different dimensions. The results are compared with the Taguchi orthogonal array approximation (TOAA), vortex search algorithm, grey wolf optimizer (GWO), sine cosine algorithm (SCA), moth-flame optimization algorithm (MFO), whale optimization algorithm (WOA) and salp swarm algorithm (SSA). In order to compare the effectiveness of HTVS statistically, Wilcoxon signed-rank test (WSRT) is used in this study. Furthermore, HTVS is applied to two different real engineering problems having some constraints (tension/compression spring design and pressure vessel design). All obtained results suggested that HTVS can find optimal or very close to optimal results. Moreover, it has good computational ability and fast convergence behavior as well.