SDDOM 2014 Abstracts


Full Papers
Paper Nr: 1
Title:

A Catapult. Searching Optima Using Factorial Designs and 2D-Neural Network Mapping Technique - A Tutorial

Authors:

Natalja Fjodorova, Marjana Novic and Matej Hohnjec

Abstract: The goal of this paper is to represent the feed forward bottle neck neural network (FFBN NN) mapping technique in comparison with traditional statistical method like Factorial Design (FD). Application of both methods provides more information about studied process and enable to establish certificate limits more affectively reaching to best quality and selecting the less cost processes. The represented FFBN NN mapping technique is simple in use, not time consuming and gives 2D visualization of multiple optima in studied technological processes. A catapult design was applied to illustrate the cases and purposes where proposed method can be implemented. The FFBN NN mapping technique can be recommended for use in industries including application in Six Sigma improvement phase.
Download

Paper Nr: 3
Title:

Low-cost EM-Simulation-based Multi-objective Design Optimization of Miniaturized Microwave Structures

Authors:

Slawomir Koziel, Adrian Bekasiewicz, Piotr Kurgan and Leifur Leifsson

Abstract: In this work, a simple yet reliable technique for fast multi-objective design optimization of miniaturized microwave structures is discussed. The proposed methodology is based on point-by-point identification of a Pareto-optimal set of designs representing the best possible trade-offs between conflicting objectives such as electrical performance parameters as well as the size of the structure of interest. For the sake of computational efficiency, most operations are performed on suitably corrected equivalent circuit model of the structure under design. Model correction is implemented using a space mapping technique involving, among others, frequency scaling. Our approach is demonstrated using a compact rat-race coupler. For this specific example, a set of ten designs representing a Pareto set for two objectives (electrical performance and the layout area) is identified at the cost corresponding to less than thirty high-fidelity EM simulations of the structure.
Download

Paper Nr: 4
Title:

Trawl-door Shape Optimization with 3D CFD Models and Local Surrogates

Authors:

Elvar Hermannsson, Leifur Leifsson, Slawomir Koziel, Piotr Kurgan and Adrian Bekasiewicz

Abstract: Design and optimization of trawl-doors are key factors in minimizing the fuel consumption of fishing vessels. This paper discusses optimization of the trawl-door shapes using high-fidelity 3D computational fluid dynamic (CFD) models. The accurate 3D CFD models are computationally expensive and, therefore, the direct use of traditional optimization algorithms, which often require a large number of evaluations, may be prohibitive. The design approach presented here is a variation of sequential approximate optimization exploiting low-order local response surface models of the expensive 3D CFD simulations. The algorithm is applied to the design of modern and airfoil-shaped trawl-doors.
Download

Paper Nr: 6
Title:

Adaptive Kriging for Simulation-based Design under Uncertainty - Development of Metamodels in Augmeted Input Space and Adaptive Tuning of Their Characteristics

Authors:

Alexandros Taflanidis and Juan Camilo Medina

Abstract: This investigation focuses on design-under-uncertainty problems that employ a probabilistic performance as objective function and consider its estimation through stochastic simulation. This approach puts no constraints on the computational and probability models adopted, but involves a high computational cost especially for design problems involving complex, high-fidelity numerical models. A framework relying on kriging metamodeling to approximate the system performance in an augmented input space is considered here to alleviate this cost. A sub region of the design space is defined and a kriging metamodel is built to approximate the system response (output) with respect to both the design variables and the uncertain model parameters (random variables). This metamodel is then used within a stochastic simulation setting (addressing uncertainties in the model parameters) to approximate the system performance when estimating the objective function for specific values of the design variables. This information is then used to search for a local optimum within the previously established design sub domain. Only when the optimization algorithm drives the search outside this domain, a new metamodel is generated. The process is iterated until convergence is established and an efficient sharing of information across these iterations is established to adaptively tune characteristics of the kriging metamodel.
Download

Paper Nr: 7
Title:

Computationally Efficient Multi-Objective Optimization of and Experimental Validation of Yagi-Uda Antenna

Authors:

Adrian Bekasiewicz, Slawomir Koziel and Leifur Leifsson

Abstract: In this paper, computationally efficient multi-objective optimization of antenna structures is discussed. As a design case, we consider a multi-parameter planar Yagi-Uda antenna structure, featuring a driven element, three directors, and a feeding structure. Direct optimization of the high-fidelity electromagnetic (EM) antenna model is prohibitive in computational terms. Instead, our design methodology exploits response surface approximation (RSA) models constructed from sampled coarse-discretization EM simulation data. The RSA model is utilized to determine the Pareto optimal set of the best possible trade-offs between conflicting objectives. In order to alleviate the difficulties related to a large number of designable parameters, the RSA model is constructed in the initially reduced design space, where the lower/upper parameter bounds are estimated by solving appropriate single-objective problems resulting in identifying the extreme point of the Pareto set. The main optimization engine is multi-objective evolutionary algorithm (MOEA). Selected designs are subsequently refined using space mapping technique to obtain the final representation of the Pareto front at the high-fidelity EM antenna model level. The total design cost corresponds to less than two hundred of EM antenna imulations.
Download