On the use of metamodel-assisted, multi-objective evolutionary algorithms

Karakasis, Marios K.; Giannakoglou, Kyriakos C.

doi:10.1080/03052150600848000

Cited by 96 publications

(61 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Regarding the uncertain variables (flow conditions), it is assumed that these follow a normal distribution around an a-priori known average value and standard deviation. The general purpose optimization platform EASY (Evolutionary Algorithms SYstem, [2]) undertakes the optimization through Metamodel-Assisted Evolutionary Algorithms, MAEAs [14,15]. The overall workflow is outlined in Fig.…”

Section: The Proposed Workflow and Its Constituentsmentioning

confidence: 99%

Aerodynamic Shape Optimization Under Flow Uncertainties Using Non-Intrusive Polynomial Chaos and Evolutionary Algorithms

Liatsikouras¹,

Asouti²,

Giannakoglou³

et al. 2017

Proceedings of the 2nd International Conference on Uncertainty Quantification in Computational Sciences and Engineering (UNCECO

Self Cite

View full text Add to dashboard Cite

Abstract. The existence of uncertainties, associated with the operating conditions or manufacturing imperfections, occur quite often in aerodynamic optimization problems. In this paper, a workflow for shape optimization in the presence of environmental uncertain flow conditions, varying stochastically around an average value with an a-priori known standard deviation, is presented. To do so, the Uncertainty Quantification (UQ) for the objective function needs to be carried out. This is based on the non-intrusive Polynomial Chaos Expansion (niPCE) method [1], which allows for a controllable number of calls to the CFD tool used to evaluate each candidate solution, compared to other sampling methods such as Monte-Carlo. Within the proposed workflow, PCE is combined with the optimization platform EASY (Evolutionary Algorithms SYstem) [2], which undertakes the optimization task. The overall process is fully automated. The shape under consideration is parameterized using CAD-free approaches, such as Radial Basis Function techniques or the combined use of two cages and the corresponding Harmonic Coordinates, which are responsible only for surface deformations whereas, as it will be explained below, other morphing/smoothing [3] tool undertakes the adaptation of the CFD mesh to the updated boundaries at each optimization cycle. The PCE method selects the points (Gaussian nodes) in the design space to be evaluated and the computed performance metrics are integrated with weights indicated by the Gauss integration rules, in order to compute the mean value and standard deviation of the objective function of the flow problem under uncertainties. The aforementioned tools are applied in two problems, in which OpenFOAM is the CFD evaluation software.

show abstract

Section: The Proposed Workflow and Its Constituentsmentioning

confidence: 99%

Aerodynamic Shape Optimization Under Flow Uncertainties Using Non-Intrusive Polynomial Chaos and Evolutionary Algorithms

Liatsikouras¹,

Asouti²,

Giannakoglou³

et al. 2017

Proceedings of the 2nd International Conference on Uncertainty Quantification in Computational Sciences and Engineering (UNCECO

Self Cite

View full text Add to dashboard Cite

show abstract

“…In this paper, without loss in generality, radial basis function (RBF) networks (Haykin 1999) are used as metamodels. For efficient training algorithms of metamodels during the evolution of a M AEA the reader should refer to (Karakasis and Giannakoglou 2005). Training the metamodels requires a pool (or database) of samples, shared among all demes.…”

Section: Distributed Hierarchical Evolutionary Algorithmmentioning

confidence: 99%

“…the metamodel. Metamodels (Giannakoglou 2002, Jin 2005, Karakasis and Giannakoglou 2005, Lim et al 2008, Zhou et al 2007, Emmerich et al 2006 interpolation or approximation methods such as polynomial regression, artificial neural networks, etc, which, after being trained on previously seen solutions, are used in place of E 1 to screen out non-promising candidate solutions at very low CPU cost. In this manner, only a few of the best offspring undergo evaluations on E i , i > 0.…”

Section: Distributed Hierarchical Evolutionary Algorithmmentioning

confidence: 99%

Distributed evolutionary algorithms with hierarchical evaluation

Kampolis

Giannakoglou

2009

Engineering Optimization

Self Cite

View full text Add to dashboard Cite

“…In MAEAs, the role of metamodels is to provide approximations to the fitness or cost of candidate solutions and, thus, save a great amount of evaluations which would otherwise be carried out by the CPU-demanding problem-specific tool. In Karakasis and Giannakoglou (2005), Giannakoglou et al (2001), they are used to pre-evaluate the population members, in the so-called Inexact Pre-Evaluation (IPE ) phase of each generation of the MAEA. IPE is based on metamodels trained on the fly, on a small number of already evaluated individuals in the neighborhood of each inividual (local metamodels).…”

Section: Introductionmentioning

confidence: 99%