“Second generation” metaheuristic algorithms such as harmony search (HS) and big bang−big crunch (BB−BC) are very efficient in truss optimization problems but computationally expensive. This paper presents two hybrid formulations of HS and BB−BC where metaheuristic search is hybridized by including gradient/pseudo-gradient information as the criterion to accept or reject new trial designs or to perform new explosions. Each new trial design is formed by combining a set of descent directions and then eventually corrected to improve it further. An improved local 1D search derived from simulated annealing is also performed. The new algorithms are tested in two weight optimization problems of truss structures: (i) the classical planar 200-bar truss optimized with 29 design variables; (ii) a large-scale space-tower with 1938 elements and 204 design variables. Optimization results prove the efficiency and robustness of the optimization algorithms developed in the research described in this paper.
|Titolo:||Comparison of hybrid metaheuristic algorithms for truss weight optimization|
|Data di pubblicazione:||2013|
|Nome del convegno:||3rd International Conference on Soft Computing Technology in Civil, Structural and Environmental Engineering|
|Digital Object Identifier (DOI):||http://dx.doi.org/10.4203/ccp.103.30|
|Appare nelle tipologie:||4.1 Contributo in Atti di convegno|