Abstract
This paper focuses on a metamodel-based design optimization algorithm. The intention is to improve its computational cost and convergence rate. Metamodel-based optimization method introduced here, provides the necessary means to reduce the computational cost and convergence rate of the optimization through a surrogate. This algorithm is a combination of a high quality approximation technique called Inverse Distance Weighting and a meta-heuristic algorithm called Harmony Search. The outcome is then polished by a semi-tabu search algorithm. This algorithm adopts a filtering system and determines solution vectors where exact simulation should be applied. The performance of the algorithm is evaluated by standard truss design problems and there has been a significant decrease in the computational effort and improvement of convergence rate.
Highlights
Employing exact simulations are very common in optimizing engineering design
This paper describes a surrogate-based optimization algorithm
Most of the meta-heuristic algorithms (e.g. GA, Harmony Search (HS), and ACO) employ a large initial population size, which leads to a large and costly number of function evaluation and suffer from slow convergence. These prohibitive factors are more pronounced when applied to exact simulation
Summary
The problem with some of the most common and robust optimization algorithms such as Genetic Algorithm, Ant Colony, and Harmony Search, is that they entail a large number of iterations for reaching the optimum solution [1,2,3,4]. This is a significant barrier when applying exact simulations to real life engineering optimization problems. M.M. Shahbazi et al / Improved version of Inverse Distance Weighting metamodel assisted Harmony Search algorithm for truss design optimization 285. The numerical results shows that the enhanced IDW+HS algorithm (IDW+HS+Tabu), in comparison to the pure HS algorithm as well as the other conventional meta-heuristic algorithms (GA and ACO), leads to a lower computational cost and higher convergence rate
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call