In the present paper, the performances of surrogate-assisted evolutionary algorithms for dynamic identification problems and damage detection are investigated. An improved algorithm is designed to limit the computational effort by introducing a proper infill sampling strategy in Differential Evolution (DE). The algorithm combines the robustness of DE with the computational efficiency due to a second-order surrogate approximation of the objective function. New individuals are selected trying to enhance both the accuracy in the region of the optimum predicted by the surrogate and the global exploration. The efficiency of the algorithm is tested by searching the global minimum of benchmark functions and by solving damage identifi-cation problems. Results are compared with those obtained adopting both the original DE algorithm and a previous proposal surrogate-based algorithm called DE-Q.
A surrogate-assisted evolutionary algorithm for dynamic structural identification / Gambarelli, Paola; Vincenzi, Loris. - ELETTRONICO. - 1:(2014), pp. 93-98. (Intervento presentato al convegno 4th International Conference on Engineering Optimization, ENGOPT 2014 tenutosi a Lisbon nel 8-11 september 2014) [10.1201/b17488-18].
A surrogate-assisted evolutionary algorithm for dynamic structural identification
GAMBARELLI, PAOLA;VINCENZI, Loris
2014
Abstract
In the present paper, the performances of surrogate-assisted evolutionary algorithms for dynamic identification problems and damage detection are investigated. An improved algorithm is designed to limit the computational effort by introducing a proper infill sampling strategy in Differential Evolution (DE). The algorithm combines the robustness of DE with the computational efficiency due to a second-order surrogate approximation of the objective function. New individuals are selected trying to enhance both the accuracy in the region of the optimum predicted by the surrogate and the global exploration. The efficiency of the algorithm is tested by searching the global minimum of benchmark functions and by solving damage identifi-cation problems. Results are compared with those obtained adopting both the original DE algorithm and a previous proposal surrogate-based algorithm called DE-Q.
Pubblicazioni consigliate
I metadati presenti in IRIS UNIMORE sono rilasciati con licenza Creative Commons CC0 1.0 Universal, mentre i file delle pubblicazioni sono rilasciati con licenza Attribuzione 4.0 Internazionale (CC BY 4.0), salvo diversa indicazione.
In caso di violazione di copyright, contattare Supporto Iris
