Multi Variate Analysis in Systematic Impeller Design Applying modeFRONTIER
EnginSoft
22-23 October 2012 Pacengo del Garda
(VR) - Italy

www.caeconference.com

2011 Conference Proceedings
2011 Conference Proceedings are now avaliable to download
2006-2010 Proceedings

download CAE proceedings

CHOOSE YOUR COUNTRY

Virtual Prototyping in Italy Virtual Prototyping in Germany Virtual Prototyping in France
Virtual Prototyping in Spain Virtual Prototyping in Norway Virtual Prototyping in United Kingdom
Other Countries...



EnginSoft - Conference Abstracts


EnginSoft International Conference 2009
CAE Technologies for Industry

Multi Variate Analysis in Systematic Impeller Design Applying modeFRONTIER

Maurer Wolfgang - Sulzer Pumps (Switzerland)
Krueger Susanne - Sulzer Pumps (Switzerland)
Fuligno Luca - EnginSoft
Linares Francesco - EnginSoft GmbH

Abstract

Designing centrifugal pumps is a challenging process since many parameters define the entire geometry. The most important component is the impeller which transforms the kinetic energy into pressure and generates the head required. To obtain an optimised impeller geometry, more than 20 parameters should be varied: this results in a complex design space and a variation of the parameters yields an extensive database for every single impeller. Applying this for an entire pump range requires the availability of advanced and highly sophisticated design tools. The decreasing time to market together with the increasing demands on product flexibility is putting pressure on product development. The usage of existing databases helps to speed up this process. This paper presents an approach based on a classification of existing impeller designs with Multi Variate Analysis through self organising maps (SOM). A proper classification of the available designs gives the developer a better understanding of the complex parameter correlation. Based on these structures, SOM enable the prediction of not yet available components by interpolating among the existing designs. This procedure not only gives a first parameter estimate for the new component but also properly conditions the variable ranges for an optimisation which is likely to follow.


Back to index

 


 

copyright © 2011 all rights reserved | terms of use | Download EnginSoft Logo | VAT nb IT00599320223