Convergence to local optima rather than the global optimum is a common shortcoming of directed search (gradient-based) optimization techniques. Successive convergence on better local optima is an efficient way of finding the global optimum and is achieved by the proposed hybrid global optimization strategy, which combines ideas from the improving hit-and-run and the tunneling global optimization methods. The searching technique of improving hit-and-run is adapted to find suitable starting points to converge on other (better) local optima. The suitability of candidate starting points is confirmed by the concept of the tunneling method. The strategy presented has unique features due to the underlying structural theory being exact, in the sense of differential equations being solved for the member (or element) stiffnesses so that they are transcendental functions of the load factor. Hence the buckling eigenproblem for the entire structure is a transcendental one, instead of the linear one associated with the more usual finite element method formulations. Constrained minimum-mass aerospace panel stability design problems are used to demonstrate the efficiency and robustness of the hybrid global optimization strategy.
Hybrid optimization strategy beyond local optima in aerospace panel designs
Hybrides Verfahren zur globalen Optimierung der Verkleidungselemente von Raumfahrzeugen
AIAA Journal ; 37 , 5 ; 588-593
1999
6 Seiten, 6 Bilder, 3 Tabellen, 18 Quellen
Aufsatz (Zeitschrift)
Englisch
Hybrid Optimisation Strategy Beyond Local Optima in Aerospace Panel Designs
British Library Conference Proceedings | 1998
|Singular and Local Optima in Structural Optimization
British Library Conference Proceedings | 1994
|