Wyniki wyszukiwania - BazTech

Ograniczanie wyników

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: surrogate models

Sortuj według:

Ogranicz wyniki do:

Recent developments in simulation-driven multi-objective design of antennas

Koziel S., Bekasiewicz A.

Bulletin of the Polish Academy of Sciences. Technical Sciences

2015

Vol. 63, nr 3

781--789

This paper addresses computationally feasible multi-objective optimization of antenna structures. We review two recent techniques that utilize the multi-objective evolutionary algorithm (MOEA) working with fast antenna replacement models (surrogates) constructed as Kriging interpolation of coarse-discretization electromagnetic (EM) simulation data. The initial set of Pareto-optimal designs is subsequently refined to elevate it to the high-fidelity EM simulation accuracy. In the first method, this is realized point-by-point through appropriate response correction techniques. In the second method, sparsely sampled high-fidelity simulation data is blended into the surrogate model using Co-kriging. Both methods are illustrated using two design examples: an ultra-wideband (UWB) monocone antenna and a planar Yagi-Uda antenna. Advantages and disadvantages of the methods are also discussed.

Forging preform shape optimization using surrogate models

Repalle J., Grandhi R. V.

Computer Assisted Mechanics and Engineering Sciences

2007

Vol. 14, No. 1

53-66

Forging of practical products from simple billet shapes is a complex and nonlinear process due to the multi-disciplinary phenomenon of material flow and processing conditions. General forgings are usually produced in a number of stages in order to avoid defects such as underfill, extra flash, voids, and folds. In spite of advancements in analysis techniques, forging process simulations do not provide function sensitivity information. Hence, the research focuses on exploring efficient non-gradient based preform shape optimization methods. In this research, an attempt is made to develop a preform shape design technique based on interpolative surrogate models, namely Kriging. These surrogate models yield insight into the relationship between output responses and input variables and they facilitate the integration of discipline-dependent analysis codes. Furthermore, error analysis and a comparison between Kriging and other approximation models (response surface and multi-point approximations) are presented. A discussion about what the results mean to a designer is provided. A case study of an automotive component preform shape design is presented for demonstration.