Database enrichment for multi objective optimization

Patricia Klotz

Description

Database enrichment for multi-objective optimization Patricia Klotz a,?, Nathalie Bartoli a, Laurence Cornez a, Renaud Lecourt a and Nicolas Savary b aONERA, 2,avenue Edouard Belin, 31055 Toulouse, France bTURBOMECA, 64511 Bordes, France Abstract The problem of optimization of complex phenomena is studied by using response surface approximation. Design of experiments is elaborated for database genera- tion of samples, but it can be improved to obtain more adapted response surfaces through a sequential enrichment based on bootstrap techniques. A multi-objective optimization is performed on a two phase flow configuration for the optimization of injection parameters in a Low Premixed Prevaporized (LPP) injection system for a combustor chamber. Key words: multi-objective optimization, response surface, neural network, design of experiments, bootstrap 1 Introduction This paper outlines some optimiza- tion research done at ONERA within the context of an ONERA Internal project (PRF DOOM) and a Euro- pean project (INTELLECT D.M.). Because optimization requires a lot of computations, the use of response surface model (RSM) is a good so- lution to decrease CPU time. The set of points for fitting RSM to each objective function is provided by a ? Corresponding author Email address: (Patricia Klotz). design of experiments. We use here Latin Hypercube Sampling (LHS) which combines deterministic and random sampling in order to reduce the variance of the RSM.

• into account


• using response


• multi-objective optimization


• ac- count multi-objective


• lpp


• phase flow

Sujets

• Bartoli
• Lecourt
• Cornez
• Savary
• Multi-objective optimization

Informations

• Publié par: pefav
• Nombre de lectures: 15
• Langue: English

Extrait

Database enrichment for multi-objective optimization

a,  a a Patricia Klotz , Nathalie Bartoli , Laurence Cornez , a b Renaud Lecourt and Nicolas Savary a ONERA, 2,avenue Edouard Belin, 31055 Toulouse, France b TURBOMECA, 64511 Bordes, France

Abstract

The problem of optimization of complex phenomena is studied by using response surface approximation. Design of experiments is elaborated for database genera- tion of samples, but it can be improved to obtain more adapted response surfaces through a sequential enrichment based on bootstrap techniques. A multi-objective optimizationisperformedonatwophaseowcongurationfortheoptimizationof injection parameters in a Low Premixed Prevaporized (LPP) injection system for a combustor chamber.

Key words:multi-objective optimization, response surface, neural network, design of experiments, bootstrap

1

Introduction

Thispaperoutlinessomeoptimiza- tion research done at ONERA within the context of an ONERA Internal project (PRF DOOM) and a Euro- pean project (INTELLECT D.M.). Becauseoptimizationrequiresalot ofcomputations,theuseofresponse surface model (RSM) is a good so- lutiontodecreaseCPUtime.The setofpointsforttingRSMtoeach objective function is provided by a

 Corresponding Email address: (Patricia Klotz).

author [e-mail]

designofexperiments.Weusehere Latin Hypercube Sampling (LHS) whichcombinesdeterministicand randomsamplinginordertoreduce thevarianceoftheRSM.Thenum- ber of necessary points and their best locationcanbebothimproved.Our purposeistodeneaneconomicway to design a RSM which will be used foroptimization.Wecannoticethat dependingontypeofsurface,some RSMaremoreecientthanothers, but unfortunately it can’t be pre- dicted.Themethodologydeveloped canbeappliedtoallproblemswhere computationtimeisdeterminant. Inthesectionsthatfollow,atrst,

Preprint submitted to Aerospace Science and Technology

10 July 2007